Alzheimer’s disease (AD) remains a significant challenge in neurological disorders, necessitating advanced diagnostic techniques for early detection and intervention. This study presents a novel approach for AD classification utilizing a combination of electroencephalography (EEG) and functional near-infrared spectroscopy (fNIRS) signals. The distinctive characteristics of the cognitive tasks employed in data acquisition underscore the need for task-specific feature extraction insights. To this end, we propose an innovative event-specific feature extraction method that adapts to the unique attributes of each task and signal. By tailoring feature extraction to the inherent characteristics of each task, we achieve maximal information extraction, thereby elevating classification performance. Our methodology employs Recursive Feature Elimination with Cross-Validation, which progressively deletes features with low importance from the model. This iterative process generates the essential features after the feature extraction. The EEG-fNIRS feature fusion capitalizes on their complementary nature, enhancing the discriminatory power of the classification model. Also, in addition to the resting state data usually used for AD classification, we used data collected through three cognitive tasks to identify rich features of AD patients. Our method demonstrates significant promise in effective diagnosis with varying cognitive statuses - healthy controls, mild cognitive impairment, and AD patients.

Multimodal sentiment analysis (MSA) and multimodal emotion recognition (MER) tasks have gained a surge of attention in recent years. Although both tasks share common ground in many ways, they are often treated as a separate task. In this work, we propose, EASUM, a new training scheme for bridging the MSA and MER tasks. EASUM aims to bring mutual benefits to both tasks based on the premise that the sentiment and emotion are closely related; hence each information should provide deeper insight into one’s affective state to complement the other. We exploit this premise to further improve the performance of each task by 1) first training a domain general model using four benchmark datasets from the MSA and MER tasks: CMU-MOSI, CMU-MOSEI, MELD, and IEMOCAP. Depending on the dataset, the domain general model learns to predict sentiment or emotion values based on the domain invariant features. 2) Then these values are later used as auxiliary pseudo labels when training a domain specific model for each task. Our premise as well as new training scheme are validated through extensive experiments on the four benchmark datasets. The results also demonstrate that the proposed method outperforms the state-of-the-art on the CMU-MOSI, CMU-MOSEI, and MELD datasets, and performs comparable to the state-of-the-art on the IEMOCAP dataset while using approximately 40% fewer parameters.

Scene change detection (SCD) is a critical task for various applications, such as visual surveillance, anomaly detection, and mobile robotics. Recently, supervised methods for SCD have been developed for urban and indoor environments where input image pairs are typically unaligned due to differences in camera viewpoints. However, supervised SCD methods require pixel-wise change labels and alignment labels for the target domain, which can be both time-consuming and expensive to collect. To tackle this issue, we design an unsupervised loss with regularization methods based on the feature-metric alignment of input image pairs. The proposed unsupervised loss enables the SCD model to jointly learn the flow and the change maps on the target domain. In addition, we propose a semi-supervised learning method based on a distillation loss for the robustness of the SCD model. The proposed learning method is based on the student-teacher structure and incorporates the unsupervised loss of the unlabeled target data and the supervised loss of the labeled synthetic data. Our method achieves considerable performance improvement on the target domain through the proposed unsupervised and distillation loss, using only 10% of the target training dataset without using any labels of the target data.

Aspect-based sentiment analysis (ABSA) is a task of identifying fine-grained sentiment entities in a given sentence, which is generally formulated as a sequence labeling problem. Recently, advancements in large pre-trained language models (PLMs) led to generative ABSA, where the task is treated as text-to-text transition resolved by fine-tuning PLMs. Although the generative methods are designed to capture sentence-level semantic information, they are inappropriate for explicit comprehension of sentiment structure. In order to address this issue, we propose sentiment element named entity recognition (SENER) for ABSA. SENER integrates the concepts of named entity recognition (NER) and generative ABSA to retrieve the sentiment entities with predefined sentiment element names, leading to better semantic and sentiment structure understanding. Experimental results on several ABSA tasks show that the proposed SENER significantly outperforms previous works on ASQP and ASTE.

In this paper, we explore incremental few-shot object detection (iFSD), which incrementally learns novel classes using only a few examples without revisiting base classes. Previous iFSD works achieved the desired results by applying metalearning. However, meta-learning approaches show insufficient performance that is difficult to apply to practical problems. In this light, we propose a simple fine-tuning-based approach, the Incremental Two-stage Fine-tuning Approach (iTFA) for iFSD, which contains three steps: 1) base training using abundant base classes with the class-agnostic box regressor, 2) separation of the RoI feature extractor and classifier into the base and novel class branches for preserving base knowledge, and 3) fine-tuning the novel branch using only a few novel class examples. We evaluate our iTFA on the real-world datasets PASCAL VOC, COCO, and LVIS. iTFA achieves competitive performance in COCO and shows a 30% higher AP accuracy than meta-learning methods in the LVIS dataset. Experimental results show the effectiveness and applicability of our proposed method11Code is available at https://github.com/TMIU/iTFA.

The goal of intelligent embodied agents is to learn how to explore within the environment, interact with objects, and understand the environment in order to achieve task objectives. There are two main approaches to training such agents: one is to train an action policy that performs the task goal through end-to-end learning, and the other is to construct a policy by implementing the necessary abilities according to the task goal in a modular manner. For complex and long-horizon tasks, such as visual room rearrangement, a modular approach that infers task sequence by identifying the causality of actions through prior knowledge shows higher performance. Based on this insight, we propose an Online Subtask Prediction Network (OSPNet) that determines the subtask to be performed at each moment based on the environment information and past subtask inference history to train an embodied agent for long-horizon tasks through an end-to-end manner, and also propose a Subtask Aware Policy Network (SAPNet) as the action policy that decides actions based on the reasoning of the OSPNet. We implement an embodied agent that performs visual room rearrangement using the proposed SAPNet and train it through imitation learning, demonstrating similar or better performance with much fewer training steps than previous works.

Recent face swapping frameworks have achieved high-fidelity results. However, the previous works suffer from high computation costs due to the deep structure and the use of off-the-shelf networks. To overcome such problems and achieve real-time face swapping, we propose a lightweight one-stage framework, FastSwap. We design a shallow network trained in a self-supervised manner without any manual annotations. The core of our framework is a novel decoder block, called Triple Adaptive Normalization (TAN) block, which effectively integrates the identity and pose information. Besides, we propose a novel data augmentation and switch-test strategy to extract the attributes from the target image, which further enables controllable attribute editing. Extensive experiments on VoxCeleb2 and wild faces demonstrate that our framework generates high-fidelity face swapping results in 123.22 FPS and better preserves the identity, pose, and attributes than other state-of-the-art methods. Furthermore, we conduct an in-depth study to demonstrate the effectiveness of our proposal.

Accurately predicting pedestrian trajectories requires a human-like socio-physical understanding of movement, nearby pedestrians, and obstacles. However, traditional methods struggle to generate multiple trajectories in the same situation based on socio-physical understanding and are computationally intensive, making real-time application difficult. To overcome these limitations, we propose SPU-BERT, a fast multi-trajectory prediction model that incorporates two non-recursive BERTs for multi-goal prediction (MGP) and trajectory-to-goal prediction (TGP). First, MGP predicts multiple goals through generative models, followed by TGP generating trajectories that approach the predicted goals. SPU-BERT can simultaneously understand movement, social interaction, and scene context from trajectories and semantic maps using a single Transformer encoder, providing explainable results as evidence of socio-physical understanding. In experiments, SPU-BERT accurately predicted future trajectories (with 0.19 m and 7.54 pixels of ADE20 for the ETH/UCY datasets and SDD) with over 100 times faster computation (0.132 s) than the state-of-the-art method. The code is available at https://github.com/kina4147/SPUBERT.

Accurate change detection enables a wide range of tasks in visual surveillance, anomaly detection and mobile robotics. However, contemporary change detection approaches assume an ideal matching between the current and stored scenes, whereas only coarse matching is possible in real-world scenarios. Thus, contemporary approaches fail to show the reported performance in real-world settings. To overcome this limitation, we propose SimSaC. SimSaC concurrently conducts scene flow estimation and change detection and is able to detect changes with imperfect matches. To train SimSaC without additional manual labeling, we propose a training scheme with random geometric transformations and the cut-paste method. Moreover, we design an evaluation protocol which reflects performance in realworld settings. In designing the protocol, we collect a test benchmark dataset, which we claim as another contribution. Our comprehensive experiments verify that SimSaC displays robust performance even given imperfect matches and the performance margin compared to contemporary approaches is huge.

Intelligent agents need to understand the surrounding environment to provide meaningful services to or interact intelligently with humans. The agents should perceive geometric features as well as semantic entities inherent in the environment. Contemporary methods in general provide one type of information regarding the environment at a time, making it difficult to conduct high-level tasks. Moreover, running two types of methods and associating two resultant information requires a lot of computation and complicates the software architecture. To overcome these limitations, we propose a neural architecture that simultaneously performs both geometric and semantic tasks in a single thread: simultaneous visual odometry, object detection, and instance segmentation (SimVODIS). SimVODIS is built on top of Mask-RCNN which is trained in a supervised manner. Training the pose and depth branches of SimVODIS requires unlabeled video sequences and the photometric consistency between input image frames generates self-supervision signals. The performance of SimVODIS outperforms or matches the state-of-the-art performance in pose estimation, depth map prediction, object detection, and instance segmentation tasks while completing all the tasks in a single thread. We expect SimVODIS would enhance the autonomy of intelligent agents and let the agents provide effective services to humans.

An intelligent robot requires episodic memory that can retrieve a sequence of events for a service task learned from past experiences to provide a proper service to a user. Various episodic memories, which can learn new tasks incrementally without forgetting the tasks learned previously, have been designed based on adaptive resonance theory (ART) networks. The conventional ART-based episodic memories, however, do not have the adaptability to the changing environments. They cannot utilize the retrieved task episode adaptively in the working environment. Moreover, if a user wants to receive multiple services of the same kind in a given situation, the user should repeatedly command multiple times. To tackle these limitations, in this article, a novel hierarchical clustering resonance network (HCRN) is proposed, which has a high clustering performance on multimodal data and can compute the semantic relations between learned clusters. Using HCRN, a semantic relation-aware episodic memory (SR-EM) is designed, which can adapt the retrieved task episode to the current working environment to carry out the task intelligently. Experimental simulations demonstrate that HCRN outperforms the conventional ART in terms of clustering performance on multimodal data. Besides, the effectiveness of the proposed SR-EM is verified through robot simulations for two scenarios.

Surface mount technology (SMT) is a process for producing printed-circuit boards. The solder paste printer (SPP), package mounter, and solder reflow oven are used for SMT. The board on which the solder paste is deposited from the SPP is monitored by the solder paste inspector (SPI). If SPP malfunctions due to the printer defects, the SPP produces defective products, and then abnormal patterns are detected by SPI. In this article, we propose a convolutional recurrent reconstructive network (CRRN), which decomposes the anomaly patterns generated by the printer defects, from SPI data. CRRN learns only normal data and detects the anomaly pattern through the reconstruction error. CRRN consists of a spatial encoder (S-Encoder), a spatiotemporal encoder and decoder (ST-Encoder-Decoder), and a spatial decoder (S-Decoder). The ST-Encoder-Decoder consists of multiple convolutional spatiotemporal memories (CSTMs) with a spatiotemporal attention (ST-Attention) mechanism. CSTM is developed to extract spatiotemporal patterns efficiently. In addition, an ST-Attention mechanism is designed to facilitate transmitting information from the spatiotemporal encoder to the spatiotemporal decoder, which can solve the long-term dependency problem. We demonstrate that the proposed CRRN outperforms the other conventional models in anomaly detection. Moreover, we show the discriminative power of the anomaly map decomposed by the proposed CRRN through the printer defect classification.

Multiple motion estimators such as an interacting multiple model (IMM) have been utilized to track target objects such as cars and pedestrians with diverse motion patterns. However, the standard IMM has limitations in combining motion models with different state definitions, so it cannot contain a complementary set of models that accurately work for all motion patterns. In this paper, we propose IMM-based adaptive target tracking with heterogeneous velocity representations and linear/curvilinear motion models. It can integrate four motion models with different state definitions and dimensions to be completely complimentary for all types of motions. We experimentally demonstrate the effectiveness of the proposed method with accuracy for various motion patterns using two types of datasets: synthetic datasets and real datasets. Experimental results show that the proposed method achieves the adaptive target tracking for diverse types of motion and also for various objects such as cars, pedestrians, and drones in the real world.

Semantic segmentation and depth estimation lie at the heart of scene understanding and play crucial roles especially for autonomous driving. In particular, it is desirable for an intelligent self-driving agent to discern unexpected obstacles on the road ahead reliably in real-time. While existing semantic segmentation studies for small road hazard detection have incorporated fusion of multiple modalities, they require additional sensor inputs and are often limited by a heavyweight network for real-time processing. In this light, we propose an end-to-end Real-time Obstacle Detection via Simultaneous refinement, coined RODSNet (https://github.com/SAMMiCA/RODSNet) which jointly learns semantic segmentation and disparity maps from a stereo RGB pair and refines them simultaneously in a single module. RODSNet exploits two efficient single-task network architectures and a simple refinement module in a multi-task learning scheme to recognize unexpected small obstacles on the road. We validate our method by fusingCityscapesandLost and Founddatasets and show that our method outperforms previous approaches on the obstacle detection task, even recognizing the unannotated obstacles at 14.5 FPS on our fused dataset ($2048\times 1024$resolution) using RODSNet-$2\times $. In addition, extensive ablation studies demonstrate that our simultaneous refinement effectively facilitates contextual learning between semantic and depth information.

Contemporary soft keyboards possess limitations: the lack of physical feedback results in an increase of typos, and the interface of soft keyboards degrades the utility of the screen. To overcome these limitations, we propose an Invisible Mobile Keyboard (IMK), which lets users freely type on the desired area without any constraints. To facilitate a data-driven IMK decoding task, we have collected the most extensive text-entry dataset (approximately 2M pairs of typing positions and the corresponding characters). Additionally, we propose our baseline decoder along with a semantic typo correction mechanism based on self-attention, which decodes such unconstrained inputs with high accuracy (96.0%). Moreover, the user study reveals that the users could type faster and feel convenience and satisfaction to IMK with our decoder. Lastly, we make the source code and the dataset public to contribute to the research community.

We present a challenging dataset, ChangeSim, aimed at online scene change detection (SCD) and more. The data is collected in photo-realistic simulation environments with the presence of environmental non-targeted variations, such as air turbidity and light condition changes, as well as targeted object changes in industrial indoor environments. By collecting data in simulations, multi-modal sensor data and precise ground truth labels are obtainable such as the RGB image, depth image, semantic segmentation, change segmentation, camera poses, and 3D reconstructions. While the previous online SCD datasets evaluate models given well-aligned image pairs, ChangeSim also provides raw unpaired sequences that present an opportunity to develop an online SCD model in an end-to-end manner, considering both pairing and detection. Experiments show that even the latest pair-based SCD models suffer from the bottleneck of the pairing process, and it gets worse when the environment contains the non-targeted variations. Our dataset is available at https://sammica.github.io/ChangeSim/.

Text entry takes an important role of effectively delivering the intention of users to computers, where physical and soft keyboards have been widely used. However, with the recent trends of developing technologies like augmented reality and increasing contactless services due to COVID-19, a more advanced type of text entry is required. To tackle this issue, we propose Air-Text which is an intuitive system to write in the air using fingertips as a pen. Unlike previously suggested air-writing systems, Air-Text provides various functionalities by the seamless integration of air-writing and text-recognition modules. Specifically, the air-writing module takes a sequence of RGB images as input and tracks both the location of fingertips (5.33 pixel error in 640x480 image) and current hand gesture class (98.29% classification accuracy) frame by frame. Users can easily perform writing operations such as writing or deleting a text by changing hand gestures, and tracked fingertip locations can be stored as a binary image. Then the text-recognition module, which is compatible with any pre-trained recognition models, predicts a written text in the binary image. In this paper, examples of single digit recognition with MNIST classifier (96.0% accuracy) and word-level recognition with text recognition model (79.36% character recognition rate) are provided.

Games have been used as excellent testbeds for research on artificial intelligence (AI) and computational intelligence for their diversity and complexity. In this article, we present AI World Cup, a set of AI competitions based on the game of soccer. We provide an introduction to the three challenges that concern a robot soccer match using both value-based and image-based state representations. AI Soccer runs the robot soccer match by participants managing each team of five two-wheeled robots. AI Commentator and AI Reporter observe the AI Soccer match and output real-time commentary and a summary article, respectively. Also, we introduce the AI World Cup platform along with rationale behind notable design choices. The official international AI World Cups held in 2018 and 2019 and the AI Masters competition held in 2019 as a part of the World Cyber Games are briefly discussed. Technical aspects of the strategies developed by participants are also discussed.

Text entry aims to provide an effective and efficient pathway for humans to deliver their messages to computers. With the advent of mobile computing, the recent focus of text-entry research has moved from physical keyboards to soft keyboards. Current soft keyboards, however, increase the typo rate due to a lack of tactile feedback and degrade the usability of mobile devices due to their large portion on screens. To tackle these limitations, we propose a fully imaginary keyboard (I-Keyboard) with a deep neural decoder (DND). The invisibility of I-Keyboard maximizes the usability of mobile devices and DND empowered by a deep neural architecture allows users to start typing from any position on the touch screens at any angle. To the best of our knowledge, the eyes-free ten-finger typing scenario of I-Keyboard which does not necessitate both a calibration step and a predefined region for typing is first explored in this article. For the purpose of training DND, we collected the largest user data in the process of developing I-Keyboard. We verified the performance of the proposed I-Keyboard and DND by conducting a series of comprehensive simulations and experiments under various conditions. I-Keyboard showed 18.95% and 4.06% increases in typing speed (45.57 words per minute) and accuracy (95.84%), respectively, over the baseline.

Anomaly detection identifies anomaly samples that deviate significantly from normal patterns. Usually, the number of anomaly samples is extremely small compared to the normal samples. To handle such imbalanced sample distribution, one-class classification has been widely used in identifying the anomaly by modeling the features of normal data using only normal data. Recently, recurrent autoencoder (RAE) has shown outstanding performance in the sequential anomaly detection compared to the other conventional methods. However, RAE, which has a long-term dependency problem, is optimized only to handle the fixed-length inputs. To overcome the limitations of RAE, we propose recurrent reconstructive network (RRN) as a novel RAE, with three functionalities for anomaly detection of streaming data: 1) a self-attention mechanism; 2) hidden state forcing; and 3) skip transition. The designed self-attention mechanism and the hidden state forcing between the encoder and decoder effectively manage the input sequences of varying length. The skip transition with the attention gate improves the reconstruction performance. We conduct a series of comprehensive experiments on four datasets and verify the superior performance of the proposed RRN in the sequential anomaly detection tasks.

Adaptive resonance theory (ART) networks, including developmental resonance network (DRN), basically use a vigilance parameter as a hyperparameter to determine whether a current input can belong to any existing categories or not. The problem here is that the clustering quality of those networks is sensitive to the vigilance parameter so that the users are required to fine-tune the parameter delicately beforehand. Another problem is that those networks only deal with a hyperrectangular decision boundary, which means they cannot learn categories of arbitrary shape. In addition, the order of data processing is a critical factor to categorize clusters correctly because each category can expand its boundary into the areas of other categories erroneously. To deal with these problems, we propose an advanced version of DRN, Adaptive DRN (A-DRN), which learns the vigilance parameters assigned for individual category nodes as well as category weights. The proposed A-DRN combines close categories to construct a cluster that contains the categories identifying a cluster boundary of arbitrary shape. Our A-DRN also employs a sliding window. The sliding window buffers sequential data points to presume the data distribution roughly, which helps our network to have a robust and consistent performance to a random order of input data. Through the experiments, we empirically demonstrate the effectiveness of A-DRN in both synthetic and real-world benchmark data sets.

Convolutional neural networks (CNNs) are one of the most successful deep neural networks. Indeed, most of the recent applications related to computer vision are based on CNNs. However, when learning new tasks in a sequential manner, CNNs face catastrophic forgetting: they forget a considerable amount of previously learned tasks while adapting to novel tasks. To overcome this main barrier to continual learning with CNNs, we introduce developmental memory (DM) into a CNN, continually generating submemory networks to learn important features of individual tasks. A novel training method, referred to here as guided learning (GL), guides the newly generated submemory to become an expert on the new task, eventually improving the performance of the overall network. At the same time, the existing submemories attempt to preserve the knowledge of old tasks. Experiments on image classification tasks show that compared with the state-of-the-art algorithms, the proposed CNN with DM not only improves the classification performance on the new image task but also leads to less forgetting of previous image tasks to facilitate continual learning.

This paper presents Reoriented Short-Cuts (RSC): A modification of the traditional Short-Cut technique, allowing almost sure, single homotopy class, asymptotic convergence in high degree of freedom (DoF) problems. An additional Informed Gaussian Sampling (IGS) technique is also introduced for convergence comparison. Traditionally, Short-Cut methods are used as a final technique to further optimize an initially found path. Typical Short-Cut methods fail as a single DoF may converge faster than the remaining, creating a zero-volume region between path segments and objects, halting further improvements. Previous attempts to solve this separate DoFs individually, drastically increasing collision checking computation. RSC and IGS control the shifting of the vertex to be Short-Cut, moving vertex positions by reorienting the line segments, removing the zero-volume convergence region. These methods are compared to similar strategies in a variety of problems including random worlds, and robot manipulation, to show the convergence across both translation and rotation oriented problems.

Detection and classification of road-objects like cars, pedestrians, and cyclists is the first step in autonomous driving. In particular, point-wise object segmentation for 3D point clouds is essential to estimate the precise appearances of the road-objects. In this paper, we propose SPriorSeg, a fast and accurate point-level object segmentation for point clouds by integrating the strengths of deep convolutional auto-encoder and region growing algorithm. Semantic segmentation using the light-weighted convolutional auto-encoder generates semantic prior by labeling a spherical projection image of point clouds pixel-by-pixel with classes of road-objects. The region growing algorithm achieves pixel-wise instance segmentation by taking into account semantic prior and geometric features between neighboring pixels. We build a well-balanced, pixel-level labeled dataset for all classes using 3D bounding boxes and point clouds from the KITTI object dataset. The dataset is employed to train our light-weighted neural network for semantic segmentation and demonstrate the performance of both semantic and instance segmentation of SPriorSeg.

The automated home referred to as Smart Home is expected to offer fully customized services to its residents, reducing the amount of home labor, thus improving human beings' welfare. Service robots and Internet of Things (IoT) play the key roles in the development of Smart Home. The service provision with these two main components in a Smart Home environment requires: 1) learning and reasoning algorithms and 2) the integration of robot and IoT systems. Conventional computational intelligence-based learning and reasoning algorithms do not successfully manage dynamic changes in the Smart Home data, and the simple integrations fail to fully draw the synergies from the collaboration of the two systems. To tackle these limitations, we propose: 1) a stabilized memory network with a feedback mechanism which can learn user behaviors in an incremental manner and 2) a robot-IoT service provision framework for a Smart Home which utilizes the proposed memory architecture as a learning and reasoning module and exploits synergies between the robot and IoT systems. We conduct a set of comprehensive experiments under various conditions to verify the performance of the proposed memory architecture and the service provision framework and analyze the experiment results.

Intelligent agents gather information and perceive semantics within the environments before taking on given tasks. The agents store the collected information in the form of environment models that compactly represent the surrounding environments. The agents, however, can only conduct limited tasks without an efficient and effective environment model. Thus, such an environment model takes a crucial role for the autonomy systems of intelligent agents. We claim the following characteristics for a versatile environment model: accuracy, applicability, usability, and scalability. Although a number of researchers have attempted to develop such models that represent environments precisely to a certain degree, they lack broad applicability, intuitive usability, and satisfactory scalability. To tackle these limitations, we propose 3-D scene graph as an environment model and the 3-D scene graph construction framework. The concise and widely used graph structure readily guarantees usability as well as scalability for 3-D scene graph. We demonstrate the accuracy and applicability of the 3-D scene graph by exhibiting the deployment of the 3-D scene graph in practical applications. Moreover, we verify the performance of the proposed 3-D scene graph and the framework by conducting a series of comprehensive experiments under various conditions.

Objects can be described in hierarchical semantics, and people also perceive them this way. It leads to the need for hierarchical classification in machine learning. On the other hand, when a new data that belongs to a new class is given, the existing classification methods should be retrained for all data including the new data. To deal with these issues, we propose an adaptive resonance theory-supervised predictive mapping for hierarchical classification (ARTMAP-HC) network that allows incremental class learning for raw data without normalization in advance. Our proposed ARTMAP-HC is composed of hierarchically stacked modules, and each module incorporates two fuzzy ARTMAP networks. Regardless of the level of the class hierarchy and the number of classes for each level, ARTMAP-HC is able to incrementally learn sequentially added input data belonging to new classes. By using a novel online normalization process, ARTMAP-HC can classify the new data without prior knowledge of the maximum value of the dataset. By adopting the prior labels appending process, the class dependency between class hierarchy levels is reflected in ARTMAP-HC. The effectiveness of the proposed ARTMAP-HC is validated through experiments on hierarchical classification datasets. To demonstrate the applicability, ARTMAP-HC is applied to a multimedia recommendation system for digital storytelling.

A GPS is usually installed at the center of a vehicle because with this configuration its data become the position of the vehicle without complex calibration and transformation. In this paper, however, we verify that the off-center arrangement of the GPS improves the position and orientation accuracy theoretically and experimentally. The extended Kalman filter (EKF) formulation and its observability and uncertainty analysis present that the off-centered GPS is better observable and less uncertain. Moreover, our experiments with synthetic data and real sensor dataset confirm the improvement of localization accuracy. The synthetic data generated in various situations reveal the important properties of localization with the off-centered GPS. The public outdoor dataset with real GPS data also supports the advantage of the off-centered GPS.

In human-robot interaction (HRI), classification is one of the most important problems, and it is essential particularly when the robot recognizes the surroundings and chooses a reaction based on a certain situation. Each interaction is different since new people appear or the environment changes, and the robot should be able to adapt to different situations during a brief interaction. Thus, it is imperative that the classification is performed incrementally in real time. In this sense, we propose an online incremental classification resonance network (OICRN) that enables incremental class learning in multi-class classification with high performance online. In OICRN, a scale-preserving projection process is introduced to use the raw input vectors online without a normalization process in advance. The integrated network of the convolutional neural network (CNN) for feature extraction and the OICRN for classification is applied to a robotic system that learns human identities through HRIs. To demonstrate the effectiveness of our network, experiments are carried out on benchmark data sets and on a humanoid robot, Mybot, developed in the Robot Intelligence Technology Laboratory, KAIST.

Adaptive resonance theory (ART) networks deal with normalized input data only, which means that they need the normalization process for the raw input data, under the assumption that the upper and lower bounds of the input data are known in advance. Without such an assumption, ART networks cannot be utilized. To solve this problem and improve the learning performance, inspired by the ART networks, we propose a developmental resonance network (DRN) by employing new techniques of a global weight and node connection and grouping processes. The proposed DRN learns the global weight converging to the unknown range of the input data and properly clusters by grouping similar nodes into one. These techniques enable DRN to learn the raw input data without the normalization process while retaining the stability, plasticity, and memory usage efficiency without node proliferation. Simulation results verify that our DRN, applied to the unsupervised clustering problem, can cluster raw data properly without a prior normalization process.

To perform a home service task through cooperation with a human in a real environment, a robot needs to deal with the environmental changes and accordingly plan appropriate behavior sequence. For this purpose, in this paper, we propose an adaptive task planner which is based on memory and reasoning. A robot perceives user behaviors and objects using an RGB-depth and thermal sensor. The robot stores a temporal sequence of behaviors for performing a task in its episodic memory that is realized by a sequence to sequence network. When the user command is given, the episodic memory is used to retrieve the behavior sequence to carry out the command. On the other hand, when the robot perceives user behaviors, the robot postpones its behavior till his/her behavior is stopped. Once stopped, the episodic memory retrieves the behavior sequence to conduct a task that the user has intended. A task scheduler schedules the behavior sequence from the memory and sends it to an internal simulator. The internal simulator confirms the behavior sequence to be executable and then if executable, it sends the next executable behavior to the execution module. If a behavior fails in the internal simulation test, fast forward planner generates an alternative behavior sequence to resolve the failed behavior problem. The effectiveness and applicability of the proposed planner is demonstrated by a wheel-based humanoid robot.

Robots are expected to perform smart services and to undertake various troublesome or difficult tasks in the place of humans. Since these human-scale tasks consist of a temporal sequence of events, robots need episodic memory to store and retrieve the sequences to perform the tasks autonomously in similar situations. As episodic memory, in this paper we propose a novel Deep adaptive resonance theory (ART) neural model and apply it to the task performance of the humanoid robot, Mybot, developed in the Robot Intelligence Technology Laboratory at KAIST. Deep ART has a deep structure to learn events, episodes, and even more like daily episodes. Moreover, it can retrieve the correct episode from partial input cues robustly. To demonstrate the effectiveness and applicability of the proposed Deep ART, experiments are conducted with the humanoid robot, Mybot, for performing the three tasks of arranging toys, making cereal, and disposing of garbage.

Memory modeling has been a popular topic of research for improving the performance of autonomous agents in cognition related problems. Apart from learning distinct experiences correctly, significant or recurring experiences are expected to be learned better and be retrieved easier. In order to achieve this objective, this paper proposes a user preference-based dual-memory adaptive resonance theory network model, which makes use of a user preference to encode memories with various strengths and to learn and forget at various rates. Over a period of time, memories undergo a consolidation-like process at a rate proportional to the user preference at the time of encoding and the frequency of recall of a particular memory. Consolidated memories are easier to recall and are more stable. This dual-memory neural model generates distinct episodic memories and a flexible semantic-like memory component. This leads to an enhanced retrieval mechanism of experiences through two routes. The simulation results are presented to evaluate the proposed memory model based on various kinds of cues over a number of trials. The experimental results on Mybot are also presented. The results verify that not only are distinct experiences learned correctly but also that experiences associated with higher user preference and recall frequency are consolidated earlier. Thus, these experiences are recalled more easily relative to the unconsolidated experiences.

This paper proposes a new architecture Deep Convolutional and Recurrent writer (DCRW) for image generation by adapting the deep Recurrent attentive writer (DRAW) architecture which is a sequential variational auto-encoder with a sequential attention mechanism for image generation. The main difference between DRAW and DCRW is that in DCRW we have replaced RNN in encoder with CNN and after replacement attention mechanism have been used for CNN. The reason behind this modification is that CNNs are the state of the art for image processing in deep learning and their basic architecture is inspired from the visual cortex. Further, for the testing of proposed architecture experiments are performed on MNIST handwritten digits data set for generation of images and results are analyzed.

Recurrent neural networks (RNNs) have been widely used for complex data modeling. However, when it comes to long-term time-dependent complex sequential data modeling with stochasticities, RNNs seem to fail because of vanishing gradients problem. Hence, in this paper, we propose a new architecture, stochastic long short term memory (S-LSTM), along with its forward and backward dynamics equations. S-LSTM models stochasticities using Bayesian brain hypothesis, which is a probabilistic model that makes predictions against which samples are tested to update the conclusions about their causes. This is the same as minimizing the difference between inference and posterior densities for suppressing the free energy. During training of S-LSTM, it predicts the mean as well as variance at each time step. The prediction error is minimized by the predicted variance which acts as an inverse weighting factor for prediction error and tries to optimize the maximum likelihood. Our proposed model is evaluated through numerical experiments on noisy Lissajous curves. In the experiments, S-LSTM is found to predict and preserve more stochasticities in the noisy Lissajous curves as compared to LSTM.

Online sequential extreme learning machine (OS-ELM) is an online learning algorithm training single-hidden layer feedforward neural networks (SLFNs), which can learn data one-by-one or chunk-by-chunk with fixed or varying data size. Due to its characteristics of online sequential learning, OS-ELM is popularly used to solve time-series prediction problem, such as stock forecast, weather forecast, passenger count forecast, etc. OS-ELM, however, has two fatal drawbacks: Its input weights cannot be adjusted and it cannot be applied to learn recurrent neural network (RNN). Therefore we propose a modified version of OS-ELM, called online recurrent extreme learning machine (OR-ELM), which is able to adjust input weights and can be applied to learn RNN, by applying ELM-auto-encoder and a normalization method called layer normalization (LN). Proposed method is used to solve a time-series prediction problem on NewYork City passenger count dataset, and the results show that R-ELM outperforms OS-ELM and other online-sequential learning algorithms such as hierarchical temporal memory (HTM) and online long short-term memory (online LSTM).

Episodic memory which can store and recall episodes has been modeled by various research. Those models focus on encoding and retrieving the same sequence of events of episodes. In this paper, we propose context preference-based deep adaptive resonance theory (CPD-ART). CPD-ART uses a new approach in encoding and retrieving a temporal sequence of events considering subjects, preference criteria such as weather, and object contexts such as beverage. A new layer, context preference field, is added to the encoding and retrieval processes for decision making. Context preference field encodes and stores the knowledge of criteria and object contexts, along with their relations in probability weight vectors. Simulation results demonstrate that CPD-ART is able to conduct decision making analysis and retrieve the sequence of events of an episode correctly through decision making analysis based on subjects, preference criteria, and the object contexts.

One of the major functions of intelligent robots such as social or home service robots is to interact with users in natural language. Moving on from simple conversation or retrieval of data stored in computer memory, we present a new Human-Robot Interaction (HRI) system which can understand and reason over environment around the user and provide information about it in a natural language. For its intelligent interaction, we integrated Dynamic Memory Networks (DMN), a deep learning network for Visual Question Answering (VQA). For its hardware, we built a robotic head platform with a tablet PC and a 3 DOF neck. Through an experiment where the user and the robot had question answering interaction in our customized environment and in real time, the feasibility our proposed system was validated, and the effectiveness of deep learning application in real world as well as a new insight on human robot interaction was demonstrated.

Episodic memory is the memory of personal experiences as episodes with subjective time. Task memory is defined as a memory for storing the knowledge of sequential procedures to perform tasks. Rather than encoding and retrieving such a temporal sequence of events or procedures, respectively, it is more efficient to implement both memories into a single memory model together. For this purpose, this paper proposes an integrated adaptive resonance theory (I-ART) neural model for episodic memory with task memory. The performance of the proposed episodic memory model is confirmed through comparison study with the other methods. And the proposed task memory is applied to perform tasks by Mybot-KSR2, developed in RIT Lab., KAIST.

Various kinds of evolutionary algorithms have been developed to solve multi-objective optimization problems. One of them is multi-objective quantum-inspired evolutionary algorithm (MQEA) which utilizes quantum computing concepts to search the solution space effectively. MQEA used nondominated sorting and crowding distance calculation as the selection operator. This paper proposes MQEA with another kind of selection operator. The proposed RN-MQEA uses reference point-based nondominated sorting approach as the selection operator, which is adopted from NSGA-III. In the computer simulations, RN-MQEA is found to provide more diverse solutions compared to MQEA and NSGA-III in solving the DTLZ test problems.

This paper proposes a novel algorithm, named dual multiobjective quantum-inspired evolutionary (DMQEA) algorithm augmented fuzzy cognitive map (FCM). DMQEA was developed to help users select preferable solutions out of the non-dominated ones and has been proven to be an effective way compared to other multi-objective optimization methods, such as MQEA, MQEA-PS, etc. DMQEA, in this paper, has been coupled with decision supporting tool, fuzzy cognitive maps (FCMs) to support selecting best models which can reflect users' preferences. Even though the attempts with single optimization such as genetic algorithms (GAs) or particle swarm optimization (PSO) have been frequently carried out, there have been only few attempt to incorporate FCM with multicriteria decision making algorithm, especially to reflect user's preference. This study aims to integrate DMQEA with FCM to build models according to user's preference. In robotics field, the interaction with human operators is an important issue and DMQEA-FCM can aid robots in their decision making process in the context of the interaction.

Due to the change in life style and diet, modern people suffer from obesity, diabetes, and other types of diseases. Regular practice of exercise can alleviate the negative effects from the diseases and even cure the diseases in certain cases. In addition, regular practice of exercise improves the quality of life. These facts have drawn much attention and people nowadays recognize the importance of exercise. As a result, more and more people hope to start exercising but they lack the knowledge of how and what to exercise. Professional counseling costs relatively expensive and thus it is difficult for ordinary people to access a counselor. To tackle these issues we propose a fuzzy expert system that designs a workout program. The system receives user's body information, preference on exercise style, and available time. Then, the system generates a customized workout program based on fuzzy reasoning. We conduct experiments to verify the performance of the proposed system. The participants enter their body condition, preference and available time and receive customized workout programs from the system. The experiments verifies the applicability of the system. The future research includes the extension of the system to meet various user demands and to reflect a number of expert knowledge sources.

Biologically inspired episodic memory is able to store time sequential events, and to recall all of them from partial information. Because of the advantages of episodic memory, the biological concepts of episodic memory have been utilized to many applications. In this research, we propose a new memory model, called Deep ART (Adaptive Resonance Theory), to make a robust memory system for learning episodic memory. Deep ART has an attribute field in the bottom layer, which is newly designed to get semantic information of inputs. After encoding all inputs with their features, events are categorized in the event field using specified inputs. Since an episode is made of a temporal sequence of events, Deep ART makes event sequences with proposed sequence encoding and decoding processes. They can encode any temporal sequence of events, even if there are duplicated events in the episode. Moreover, based on the result of the analysis of retrieval error, Deep ART does not use the complement coding for partial inputs to enhance the accuracy of episode retrieval from partial cues. The simulation results demonstrate the effectiveness of Deep ART as the long term memory.

Multimedia recommendation technology has been developed in various fields these days. In order to provide multimedia in addition to dialog, it is essential to select appropriate multimedia associated with a certain situation for more delivery effects of digital storytelling, which enables story telling agents to share their stories with users using digital multimedia in an effective way. For this purpose, we propose a multimedia recommendation system for software agents of smart devices to select multimedia that is appropriate to the given situation in storytelling to users or interacting with users. The fusion ART network is employed for the multimedia recommendation system that selects an appropriate digital media file for individual multimedia features. The system is learned incrementally based on feedback from users. The proposed system is purposed to select multimedia to be conveyed in addition to the dialog between the user and the digital creature on a smartphone. The applicability is verified through experiments with a smartphone application implemented for demonstration.

This paper proposes a novel hierarchical behavior planner with a multi-layered confabulation based behavior selection structure for robots to perform tasks. The proposed planner integrates a STRIPS based behavior selection approach and cogent confabulation approach. The STRIPS based behavior selection approach is a goal tree search that induces goal-oriented sequences of behaviors, while the cogent confabulation approach is based on conditional probabilities between input symbols and target behaviors, aims to model human thinking mechanism. Our planner is applied with a set of behaviors defined in a multi-layered structure to show that it can plan a hierarchical sequences of behaviors to perform given tasks. The effectiveness and applicability of the proposed scheme is demonstrated through the experiments with the robot Mybot, developed in the Robot Intelligence Technology Lab. at KAIST.

This paper proposes a novel scheme of integrating episodic memory into semantic memory based task planner. Task planners have taken an important role in AI research along with semantic memory to better perform tasks for robots. Episodic memory memorizes and retrieves temporal sequence of situated behaviors by which temporal relationship between behaviors can be defined. None of any research, however, has implemented it into their work for task planning. By introducing episodic memory into task planner, the temporal causal relationship between situated behaviors, which are stored in semantic memory, is taken into consideration. The integrated architecture proves its effectiveness by notably reducing the number of nodes traversed in finding solutions. Robots can reduce time complexity in solving given problems by retrieving previous memories. Deep Adaptive Resonance Theory (Deep-ART) neural model and cogency-based hierarchical behavior planner are used for the episodic memory and the task planner, respectively. Cogency-based hierarchical behavior planner proves its capability of solving given problems in experiment with humanoid robot Mybot, and Deep-ART is augmented to the planner and tested in simulations. Therefore, the contribution of this approach lies on developing a framework which takes advantage of implementing episodic memory and planner in one place.

Episodic memory can store time sequential events and retrieve them anytime with specific cues. However, if the episodic memory only stores events comprised of actions and objects, execution of episodes may fail if current situation is different from the settings it learned in. As a solution, we propose Deep C-ART (Context-Adaptive Resonance Theory) which considers not only time sequential events but also their contexts. In addition to the learning process of Deep ART, Deep C-ART stores context information such as situation of objects, states of robots, place, and time of episodes. Since context changes over each event in an episode, Deep C-ART forms an episode with an event sequence and a context sequence. During retrieval and execution of episode, it compares the current situation with the learned one to verify that it is executable or in an anomaly situation. The effectiveness of Deep C-ART is demonstrated through computer simulations.

To prepare for the anticipated age of human-robot symbiosis, robots should be able to interact and cooperate with humans effectively by understanding the meaning and intention of human behavior. In this paper, we define human intention as “desired behavior of the human using objects.” To infer the defined human intention, a robot should learn the object affordance along with a behavior hierarchy structure. Thus, in this paper, we propose a behavior hierarchy-based affordance network (BHAN) and a behavior hierarchy-based affordance map (BHAM) to represent the object affordance, behavior hierarchy structure, and object hierarchy structure, simultaneously. Autonomous and interactive BHAN/BHAM learning algorithms are also proposed to make a robot develop the BHAN and BHAM by itself, as well as by interacting with a human. Based on the newly developed BHANs and BHAM, a robot could infer the human intention from information observed in context and from human behavior. The effectiveness of the proposed method was demonstrated through experiments on human-robot interaction with building blocks using a simulated differential wheel robot and a real human-sized humanoid robot.

This paper proposes a robust background model-based dense-visual-odometry (BaMVO) algorithm that uses an RGB-D sensor in a dynamic environment. The proposed algorithm estimates the background model represented by the nonparametric model from depth scenes and then estimates the ego-motion of the sensor using the energy-based dense-visual-odometry approach based on the estimated background model in order to consider moving objects. Experimental results demonstrate that the ego-motion is robustly obtained by BaMVO in a dynamic environment.

285 million people are estimated to be visually impaired worldwide. Visually impaired people typically use a white cane or a guide dog or both of them to walk down the street. However, such as a cane and/or a dog are not enough to secure them from being collided with obstacles in a dynamic environment. This paper proposes a navigational assistance system based on fuzzy integral-based gaze control for visually impaired people in a dynamic indoor environment. It largely consists of an RGB-D camera and a vibrotactile vest interface. The RGB-D camera detects static and dynamic obstacles and obtains obstacle information on their center positions, sizes, and velocities. The fuzzy integral-based gaze control for obstacle detection is proposed to reduce a blind spot of the camera and obtain more information of the environment. The vibrotactile vest interface notifies a direction to avoid the obstacle using a fuzzy integral-based imminent-obstacle selection algorithm. To confirm the performance of the proposed assistance system for visually impaired people, experiments are carried out in an indoor dynamic environment.

A robot is expected to carry out a task autonomously with its own knowledge system. Using the knowledge system, the robot can recognize current situation and recall a proper sequence for performing an appropriate task in that situation. To build such knowledge system, the robot learns the knowledge from user demonstrations as if a child learns through interactions with parents and teachers. User demonstration is captured by an RGBD camera embedded the robot. The robot needs to segment each execution from continuous RGB-D streams. In this paper, each execution is composed of an object and an action performed on the object. The sequence of executions, or the procedure, should be stored in the robot's memory for the the robot to retrieve and execute the procedure in a similar situation later. Such a procedural memory is developed based on an adaptive resonance system. Using the procedural memory learned, the robot can perform the full sequences of tasks with only partial information given on executions. The effectiveness of the proposed scheme is demonstrated for four tasks through computer simulations.

This paper proposes a resource-oriented, decentralized auction algorithm (RODAA) for multirobot task allocation considering multiple resources of the robots and limited robot communication range. The resources that this paper focuses on are the expendable supplies that a robot consumes and recharges while performing tasks, such as energy. In the proposed algorithm, each robot generates its cost for the task in a probabilistic manner considering multiple paths that visit none or different combinations of refill stations for performing the task based on the robot's residual resources. For robust and time-efficient task allocation with limited robot communication range in a dynamic network, a multihop-based auction algorithm is proposed. This paper also introduces a solar panel cleaning mission as a new application for multirobot systems and the proposed algorithm is implemented in the simulation of the mission. The simulation results demonstrate that the proposed algorithm is capable of completing the panel cleaning mission faster than other auction-based task allocation algorithms and has lower overall resource consumption.

During the last few decades, as a part of effort to enhance natural human robot interaction (HRI), considerable research has been carried out to develop human-like gaze control. However, most studies did not consider hardware implementation, real-time processing, and the real environment, factors that should be taken into account to achieve natural HRI. This paper proposes a fuzzy integral-based gaze control algorithm, operating in real-time and the real environment, for a robotic head. We formulate the gaze control as a multicriteria decision making problem and devise seven human gaze-inspired criteria. Partial evaluations of all candidate gaze directions are carried out with respect to the seven criteria defined from perceived visual, auditory, and internal inputs, and fuzzy measures are assigned to a power set of the criteria to reflect the user defined preference. A fuzzy integral of the partial evaluations with respect to the fuzzy measures is employed to make global evaluations of all candidate gaze directions. The global evaluation values are adjusted by applying inhibition of return and are compared with the global evaluation values of the previous gaze directions to decide the final gaze direction. The effectiveness of the proposed algorithm is demonstrated with a robotic head, developed in the Robot Intelligence Technology Laboratory at Korea Advanced Institute of Science and Technology, through three interaction scenarios and three comparison scenarios with another algorithm.

Since multi-objective optimization algorithms (MOEAs) have to find exponentially increasing number of nondominated solutions with the increasing number of objectives, it is necessary to discriminate more meaningful ones from the other nondominated solutions by additionally incorporating user preference into the algorithms. This paper proposes dual multi-objective particle swarm optimization (DMOSPO) by introducing secondary objectives of maximizing both user preference and diversity to the nondominated solutions obtained for primary objectives. The proposed DMOSPO can induce the balanced exploration of the particles in terms of user preference and diversity through the dual-stage of nondominated sorting such that it can generate preferable and diverse nondominated solutions. To demonstrate the effectiveness of the proposed DMOPSO, empirical comparisons with other state-of-the-art algorithms are carried out for benchmark functions. Experimental results show that DMOPSO is competitive with the other compared algorithms and properly reflects the user's preference in the optimization process while maintaining the diversity and solution quality.

This paper proposes multiobjective particle swarm optimization with preference-based sort (MOPSO-PS), in which the user's preference is incorporated into the particle swarm optimization (PSO) update process to determine the relative merits of nondominated solutions while handling the mutual dependences and priorities of objectives. In MOPSO-PS, the user's preference is represented as the degree of consideration for each objective using the fuzzy measure. The global evaluation of a particle, which represents the quality of the particle according to the user's preference, is carried out by the fuzzy integral, which integrates the partial evaluation value of each objective with respect to the degree of consideration. Since the global best attractor of each particle in the population is randomly chosen among the nondominated particles having a relatively higher global evaluation value in each PSO update iteration, the optimization is gradually guided by the user's preference. After the optimization, the most preferable particle can be chosen for practical use by selecting the particle with the highest global evaluation value. The effectiveness of the proposed MOPSO-PS is demonstrated by the application of path, following footstep optimization for humanoid robots in addition to empirical comparison with the other algorithms. The footsteps optimized by the MOPSO-PS were verified by simulation. The results indicate that the user's preference is properly reflected in optimized solutions without any loss of overall solution quality or diversity.

This paper proposes a multi-objective evolutionary algorithm-based optimal posture controller to generate an optimal trajectory of humanoid robots against external disturbance using an iterative linear quadratic regulator (ILQR) and concurrently optimize multiple performance criteria. As the dimensionality of nonlinear system increases, it is difficult to find the weighting matrices of cost function in ILQR. In the proposed method, this problem is solved by employing a multi-objective quantum-inspired evolutionary algorithm (MQEA) to obtain nondominated solutions of the weighting matrices generating various optimal trajectories that satisfy multiple performance criteria. Among numerous nondominated solutions generated from MQEA, fuzzy measure and fuzzy integral are employed for global evaluation by integrating the partial evaluation of each of them over criteria with respect to user's degree of consideration for each criterion. The effectiveness of the proposed method is verified by computer simulations for the problem of balancing the posture of a humanoid robot against external impulse force, where the robot is modeled by a four-link inverted pendulum.

Multiobjective quantum-inspired evolutionary algorithm (MQEA) employs Q-bit individuals, which are updated using rotation gate by referring to nondominated solutions in an archive. In this way, a population can quickly converge to the Pareto optimal solution set. To obtain the specific solutions based on user's preference in the population, MQEA with preference-based selection (MQEA-PS) is developed. In this paper, an improved version of MQEA-PS, MQEA-PS2, is proposed, where global population is sorted and divided into groups, upper half of individuals in each group are selected by global evaluation, and selected solutions are globally migrated. The global evaluation of nondominated solutions is performed by the fuzzy integral of partial evaluation with respect to the fuzzy measures, where the partial evaluation value is obtained from a normalized objective function value. To demonstrate the effectiveness of the proposed MQEA-PS2, comparisons with MQEA and MQEA-PS are carried out for DTLZ functions.

This paper proposes an optimal control method to generate a minimum-torque change trajectory of humanoid arm by using a differential dynamic programming (DDP). Since DDP is a locally optimal feedback controller, the convergence is not guaranteed unless DDP starts with a good reference trajectory for high-dimensional nonlinear dynamical systems. The reference trajectory is generated by using the minimum-jerk trajectory method, and then the corresponding torque profile is obtained by using the computed-torque method. This reference trajectory is not optimal because it is solely based on the kinematics of the system. In the proposed method, the rate of torque change is used as control input in DDP to generate the optimal trajectory, which concurrently minimizes the torque variations and considers the dynamics of the system. The effectiveness of the proposed method is verified by computer simulations for generating the optimal trajectory of a 7 degrees-of-freedom (DOF) MyBot humanoid arm in Webots simulator.

Since multiobjective evolutionary algorithms (MOEAs) provide a set of nondominated solutions, decision making of selecting a preferred one out of them is required in real applications. However, there has been some research on MOEA in which the user's preferences are incorporated for this purpose. This paper proposes preference-based solution selection algorithm (PSSA) by which user can select a preferred one out of nondominated solutions obtained by any one of MOEAs. The PSSA, which is a kind of multiple criteria decision making (MCDM) algorithm, represents user's preference to multiple objectives or criteria as a degree of consideration by fuzzy measure and globally evaluates obtained solutions by fuzzy integral. The PSSA is also employed in each and every generation of evolutionary process to propose multiobjective quantum-inspired evolutionary algorithm with preference-based selection (MQEA-PS). To demonstrate the effectiveness of PSSA and MQEA-PS, computer simulations and real experiments on evolutionary multiobjective optimization for the fuzzy path planner of mobile robot are carried out. Computer simulation and experiment results show that the user's preference is properly reflected in the selected solution. Moreover, MQEA-PS shows improved performance for the DTLZ problems and fuzzy path planner optimization problem compared to MQEA with dominance-based selection and other MOEAs like NSGA-II and MOPBIL.

When a humanoid robot moves in a dynamic environment, a simple process of planning and following a path may not guarantee competent performance for dynamic obstacle avoidance because the robot acquires limited information from the environment using a local vision sensor. Thus, it is essential to update its local map as frequently as possible to obtain more information through gaze control while walking. This paper proposes a fuzzy integral-based gaze control architecture incorporated with the modified-univector field-based navigation for humanoid robots. To determine the gaze direction, four criteria based on local map confidence, waypoint, self-localization, and obstacles, are defined along with their corresponding partial evaluation functions. Using the partial evaluation values and the degree of consideration for criteria, fuzzy integral is applied to each candidate gaze direction for global evaluation. For the effective dynamic obstacle avoidance, partial evaluation functions about self-localization error and surrounding obstacles are also used for generating virtual dynamic obstacle for the modified-univector field method which generates the path and velocity of robot toward the next waypoint. The proposed architecture is verified through the comparison with the conventional weighted sum-based approach with the simulations using a developed simulator for HanSaRam-IX (HSR-IX).

This paper proposes particle swarm optimization based central pattern generator (CPG) to generate rhythmic signals for fish-like locomotion of robotic fish. The robotic fish's wave form approximates fish's traveling wave. Since each joint angle of the robotic fish is modeled by a periodic function, it can be easily produced by a CPG. A CPG consists of biological neural oscillators, which can produce coordinated rhythmic signals by using simple input signals. The proposed CPG uses a neural oscillator for each joint of a robotic fish. To optimize the parameters of the CPG which determine the output signals, particle swam optimization (PSO) is employed. The effectiveness of the proposed CPG is demonstrated by computer simulation and real experiment with the robotic fish Fibo, developed in the Robot Intelligence Technology Lab., KAIST.

Abstract-Multi-objective particle swarm optimization (MOPSO) provides a set of nondominated solutions and the number of nondominated solutions increases exponentially when the number of objectives increases. To select a desired solution out of them, preference-based solution selection algorithm (PSSA) was proposed by incorporating user's preference into multi-objective evolutionary algorithms. In this paper, multi objective particle swarm optimization with preference-based sorting (MOPSO-PS) is proposed, where a global best position is randomly selected from the archive of nondominated solutions sorted by global evaluation considering user's preferences for multiple objectives. The user's preference is represented as a degree of consideration for the objectives by the fuzzy measures. The global evaluation of the solutions is carried out by the fuzzy integral of partial evaluation with respect to the fuzzy measures, where the partial evaluation of each solution is obtained as a normalized objective function value. To demonstrate the effectiveness of the proposed MOPSO-PS, empirical comparisons to NSGA-II, MQEA, and MOPSO are carried out for the DTLZ functions. Experimental results show that the user's preference is properly reflected in the selected solutions without any loss of overall quality and diversity.

This paper presents a novel interval type-2 fuzzy logic control architecture for flocking system when the system has noisy sensor measurements. The traditional type-1 fuzzy logic controller (FLC) using precise type-1 fuzzy sets cannot fully model and handle the uncertainties of sensor data. However, type-2 FLC using type-2 fuzzy sets with a footprint of uncertainty (FOU) produces better performances under noisy environments. In this paper, therefore, we present a reactive control architecture for flocking algorithm that is based on interval type 2 FLC to implement the flocking behaviors consisting separation, obstacle avoidance, and velocity matching behaviors. The type-2 based control system could cope with the uncertainties of noisy sensor measurements and resulted in good performances that outperformed the type-1 FLC.

Conventional methods produced composite facial expressions by interpolating the representative facial expressions under the assumption that the transitions between facial expressions are linear. Considering the nonlinear property, this paper proposes a fuzzy integral-based method for generating composite facial expressions. Fuzzy measures represent the relationship among emotions and the partial evaluation of current emotion state is obtained from a predefined error function of the ideal basic emotion states and the current emotion state. Fuzzy integral of the partial evaluation with respect to fuzzy measures is employed to globally evaluate the current emotion state for generating composite facial expressions. The effectiveness of the proposed method for generating composite facial expressions is demonstrated through the experiments with a robotic head with 19 degrees of freedom, developed in RIT Laboratory, KAIST.

The standard fuzzy logic controllers, also known as type-1 fuzzy logic controllers, have often been criticized for their inability to handle uncertainties in the control processes. Therefore, a lot of attention is being focused on type-2 fuzzy logic controllers, especially, the interval type-2 fuzzy logic controllers. This paper aims at developing both type-1 and type-2 fuzzy logic controllers for an airplane altitude control problem and comparing their performances. Both the controllers have similar knowledge bases, and both of them are tested in two simulation setups, an ideally modeled setup and a setup with uncertainties. The results show that the type-2 fuzzy logic controller outperforms the type-1 fuzzy logic controller, especially, in the environment with uncertainties. Therefore, this research seems to validate the superiority of type-2 fuzzy logic control in making the controllers independent of uncertainties in intricate model details.

Artificial creatures need an intelligent behavior selection method to be used as an intermediate interface for natural interaction with users. For this purpose, the mechanisms of thought were proposed based on the probability and the degree of consideration. However, they were time-consuming when applying to an intelligent artificial creature with large numbers of wills, contexts and behaviors. Moreover, since the context-based evaluation only considers the behaviors short-listed by the will-based evaluation, some generated behaviors are inappropriate over the perceived contexts. To solve these problems, this paper proposes a novel behavior selection method for the intelligent artificial creatures using the degree of consideration-based mechanism of thought (DoC-MoT). The behaviors are short-listed by current dominant wills and perceived contexts and then they are globally evaluated by the fuzzy integral of the partial evaluation values of behaviors over artificial creature's wills and external contexts, with respect to the fuzzy measure values representing its degrees of consideration. The effectiveness of the proposed behavior selection method is demonstrated by experiments carried out with a synthetic character “DD” in the 3D virtual environment. The results show that the generated behaviors were appropriate both to the current wills and the perceived contexts. Moreover, the computation time to select a behavior was decreased in the proposed method than the behavior selection methods using the probability-based MoT and DoC-MoT without the behavior short-listing.

This paper proposes a novel evolutionary multiobjective footstep planner for humanoid robots. First, a footstep planner using a univector field navigation method is proposed to provide a command state (CS), which is to be an input of a modifiable walking pattern generator (MWPG) at each footstep. Then, the MWPG generates corresponding trajectories for every leg joint of the humanoid robot at each footstep to follow the CS. Second, a multiobjective evolutionary algorithm (MOEA) is employed to optimize the univector fields satisfying multiple objectives in navigation. Finally, a preference-based selection algorithm based on a fuzzy measure and fuzzy integral is proposed to select the preferred one out of various nondominated solutions obtained by the MOEA. The effectiveness of the proposed evolutionary multiobjective footstep planner is demonstrated through computer simulations for a simulation model of a small-sized humanoid robot, HanSaRam-VIII.

The wireless sensor network is a decentralized and self-organized system. Each sensor node in the sensor network should be intelligent enough to carry out its task of monitoring the environment. There would be numerous ways for deploying the sensor nodes in the environment. In this paper, swarm intelligence-based sensor network deployment strategy is proposed. To make a reference point for each sensor node, fuzzy integral is utilized as a multi-criteria decision making process. Three criteria, such as sensor value, crowdedness and confidence, are used for partial evaluation and the degree of consideration for each criterion is represented by fuzzy measure. Global evaluation by fuzzy integral determines the best position for each sensor node independently. To show the effectiveness of the proposed strategy, it is compared with the SPSO07-based deployment strategy through computer simulations in a simulation environment. The results show that the proposed strategy covers much wider area with sensor nodes than the SPSO07-based one.

This paper proposes a method to generate an optimal trajectory of nonlinear dynamical system and concurrently optimize multiple performance criteria. As the dimensionality of system increases, it is difficult to find values of cost/reward function of conventional optimal controllers. In order to solve this problem, the proposed method employs iterative linear quadratic regulator and multi-objective quantum-inspired evolutionary algorithm to generate various optimal trajectories that satisfy multiple performance criteria. Fuzzy measure and fuzzy integral are also employed for global evaluation by integrating the partial evaluation of each solution over criteria with respect to user's degree of consideration for each criterion. Effectiveness of the proposed method is verified by computer simulation carried out for the problem of stabilizing two-link inverted pendulum model.

Central pattern generator (CPG) is used to control the locomotion of vertebrate and invertebrate animals, such as walking, running or swimming. It consists of biological neural networks that can produce coordinated rhythmic signals by using simple input signals. In this paper, a full-body joint trajectory generator is proposed for stable bipedal walking by using an evolutionary optimized CPG. Sensory feedback pathways are proposed in the CPG structure, which uses force sensing resistor (FSR) signals. In order to optimize the parameters of CPG, quantum-inspired evolutionary algorithm is employed. Then, controller is developed to control the position of both ankles and pelvis and the pitching angles of shoulders. The proposed trajectory generator controls the position of the center of pelvis along lateral direction, and the pitching angle of both shoulders in addition to the position of both ankles for stable biped locomotion. The stability of biped locomotion along lateral direction is improved by controlling the position of the center of pelvis along lateral direction. To reduce yawing momentum, the pitching angle of both shoulders are controlled. The effectiveness is demonstrated by simulations with the Webot model of a small-sized humanoid robot, HSR-IX and real experiments with HSR-IX.

This paper proposes a navigation framework for humanoid robots, which integrates gaze control and modified univector field-based path planning to cope with moving obstacles. To make navigation robust, obstacles are modeled according to their relative velocities and positions. Moreover, partial evaluation values for gaze control architecture are also considered for modifying their virtual size and moving trajectory. In addition, gaze control architecture is proposed, which estimates the size of local map confidence area, self-localization error, surrounding obstacles and obstacle-free distance against those obstacles in the local map. The proposed framework is verified through computer simulations by using a developed simulator for HanSaRam-VIII.

This paper proposes a multiobjective quantum-inspired evolutionary algorithm (MQEA) to design efficient fuzzy path planner of mobile robot. MQEA employs the probabilistic mechanism inspired by the concept and principles of quantum computing. As the probabilistic individuals are updated by referring to nondominated solutions in the archive, population converges to Pareto-optimal solution set. In order to evaluate the performance of proposed MQEA, robot soccer system is utilized as a mobile robot system. Three objectives such as elapsed time, heading direction and posture angle errors are designed to obtain robust fuzzy path planner in the robot soccer system. Simulation results show the effectiveness of the proposed MQEA from the viewpoint of the proximity to the Pareto-optimal set. Moreover, various trajectories by the obtained solutions from the proposed MQEA are shown to verify the performance and to see its applicability.

This paper proposes a novel hybrid algorithm of particle swarm optimization (PSO) and evolutionary programming (EP), named particle swarm optimization driven by evolving elite group (PSO-EEG) algorithm. The hybrid algorithm combines the movement update property of canonical PSO with the evolutionary characteristics of EP. It is processed in two stages; elite group stage by EP and ordinary group stage by PSO. For the former group, a novel concept of evolving elite group (EEG) is introduced, which consists of relatively superior particles in a population. The elite particles are evolved by mutation and selection scheme of EP. The other ordinary particles refer to the closest elite particle as well as the global best and the personal best, to update their location. Simulation results demonstrate the proposed PSO-EEG is highly competitive in terms of robustness, accuracy and convergence speed for five well-known complex test functions.

In this paper, an artificial creature is designed to have its own genome in which a specific personality is encoded. The genome is composed of 14 chromosomes each of which consists of three kinds of genes such as fundamental genes, internal-state-related genes, and behavior-related genes. To represent various types of personality, a large number of genes are needed. In this case, if gene values are assigned manually for the individual genome, it becomes increasingly difficult and time-consuming to generate a desired personality reliably and consistently. Considering this problem, this paper proposes an evolutionary process that generates a genome encoding a specific personality of an artificial creature. The process evolves a population of genomes such that it customizes the genome, which meets a simplified set of personality traits desired by the user. The evaluation procedure for each genome of the population is carried out in a virtual environment using a tailored perception scenario and a dedicated fitness function. An artificial creature, Rity, is developed in the virtual 3-D world created in a PC to demonstrate the effectiveness of the proposed process.

Mass-spring-damper motion dynamics-based particle swarm optimization (MMD-PSO) is a novel optimization paradigm based on motion dynamic model which consists of mass, spring and damper. In MMD-PSO some particles, which are located fitter places than other particles, drop their anchor and connect springs and dampers between the anchors and all the particles. These connections influence the movements of the particles so as to proceed to fitter places attracted by the anchors. To demonstrate the effectiveness of MMD-PSO, several experiments are carried out on numerical optimization problems with complex test functions. The results show that proposed MMD-PSO is more powerful than original PSO and PSO mass-spring analogy in terms of robustness and convergence speed with no tuning parameters.

Genetic robot is one of artificial creatures and has its own genome in which each chromosome consists of many genes that contribute to defining its personality. By using the concept of genetic robot, this paper proposes personalized robotic doll by applying evolutionary process to generate unique propensity, defined by its genome. A genome population is evolved such that it customizes the genome satisfying a propensity desired by user based on Big Five personality dimensions. Robotic doll has emotion and motivation to reflect its internal state and to provide human friendly interaction. To demonstrate the effectiveness of this scheme, a bear-like robotic doll, GomDoll, is developed and the evolved genome is implanted to it to see its manner of internal and external responses to stimuli.

This paper proposes a multiobjective evolutionary algorithm (MOEA) for the problem with many objectives, where each objective is more strengthened. In the real world applications, satisfying as many objectives as possible somewhat at the same time can be less preferred than optimizing each specific objective individually. To solve this kind of problems, this paper proposes the complement of (1−k) dominance and the pruning method considering objective deviation to get a set of nondominated solutions with specifically optimized objectives. Promoting the specificity of objective improves the optimization performance on problems with many objectives. In experimental results, proposed algorithm shows improved performance compared with the state-of-the-art MOEAs such as SPEA, SPEA2 and NSGA2. The performance is measured in terms of the solution set coverage and the closeness to the true Pareto front. Also, diversity metric is applied to verify the spread of nondominated set.

The Particle Swarm Optimization (PSO) algorithm is a robust stochastic evolutionary computation technique based on the movement and intelligence of swarms. This paper proposes a novel PSO algorithm, based on the potential field and the motion dynamics model. It is assumed that particles form potential fields and each particle has its own mass. The potential filed and mass are modeled by the particles’ fitness value. By using these fitness based models, the proposed algorithm performs well, in particular, in avoiding the local minima compare to the original PSO. The proposed PD-PSO successfully solves minimization problems of complex test functions.

This paper proposes a new concept of the genetic robot which has its own robot genome, in which each chromosome consists of many genes that contribute to defining the robotpsilas personality. The large number of genes also allows for a highly complex system, however it becomes increasingly difficult and time-consuming to ensure reliability, variability and consistency for the robotpsilas personality while manually initializing values for the individual genes. To overcome this difficulty, this paper proposes an evolutionary algorithm for a genetic robotpsilas personality (EAGRP). EAGRP evolves a gene pool that customizes the robotpsilas genome so that it closely matches a simplified set of features desired by the user. It does this using several new techniques. It acts on a 2 dimensional individual upon which a new masking method, the Eliza-Meme scheme, is used to derive a plausible individual given the restricted preference settings desired by the user. The proposed crossover method allows reproduction for the 2-dimensional genome. Finally, the evaluation procedure for individuals is carried out in a virtual environment using tailored perception scenarios.

Data may be wrongly measured or come from other sources. Such data is a big problem in regression, which retrieve parameters from data. Random sample consensus (RANSAC) and maximum likelihood estimation sample consensus (MLE-SAC) are representative researches, which focused on this problem. However, they do not cope with varying data distribution because they need to tune variables according to given data. This paper proposes user-independent parameter estimator, u-MLESAC, which is based on MLESAC. It estimates variables necessary in probabilistic error model through expectation maximization (EM). It also terminates adaptively using failure rate and error tolerance, which can control trade-off between accuracy and running time. Line fitting experiments showed its high accuracy and robustness in varying data distribution. Its results are compared with other estimators. Its application to landmark-based localization also verified its performance compared with other estimator.

This paper proposes a multi-layered architecture of middleware for ubiquitous robots. Ubiquitous robots consist of different platforms with various functions and interfaces. Without a middleware, software agents have to hold information of all ubiquitous robots in advance to access other sensors or mobile robots. This decreases modularity and scalability of the entire system, therefore makes it difficult to develop and maintain software agents. The proposed architecture uses context information to decouple hardware and software agents as a bridge between them. Providing context-awareness to the middleware, software agents is able to get context information without accessing physical sensors directly, so that modularity and scalability are increased. Not only providing context information to software agents, the middleware uses mobile robots to gather as much sensor information as possible to generate context information actively. A middleware architecture is proposed with five layers classified due to device/environment dependencies, which are physical layer, device management layer, context provider and mobile robot(Mobot) scheduler layer, software robot(Sobot) management layer and software agent layer. The proposed middleware is implemented and simulated with virtual sensors in the virtual environment.

Accelerated Q-learning algorithm is proposed for environment having both goal and fail states. It extends Q-learning, a well-known scheme in reinforcement learning. Unlike this conventional Q-learning, the proposed algorithm keeps track of the past failure experiences as a separate fail state-action value, QF. Agent uses this value along with a goal state-action value, QN, which is calculated and updated using conventional Q-learning, to modify the exploratory behavior during learning phase. Effectiveness of the proposed accelerated Q-learning algorithm is verified in a grid world environment. The proposed algorithm significantly reduces a convergence speed to find out the optimal path from start state to goal state while maximizing its receiving rewards.

In order to handle complex navigational commands, this paper proposes a novel algorithm that can modify a walking period and a step length in both sagittal and lateral planes. By allowing a variation of zero moment point (ZMP) over the convex hull of foot polygon, it is possible to change the center of mass (CM) position and velocity independently throughout the single support phase. This permits a range of dynamic walking motion, which is not achievable using the 3-D linear inverted pendulum mode (3D-LIPM). In addition, the proposed algorithm enables to determine the dynamic feasibility of desired motion via the construction of feasible region, which is explicitly computed from the current CM state with simple ZMP functions. Moreover, adopting the closed-form functions makes it possible to calculate the algorithm in real time. The effectiveness of the proposed algorithm is demonstrated through both computer simulation and experiment on the humanoid robot, HanSaRam-VII, developed at the Robot Intelligence Technology (RIT) laboratory, Korea Advanced Institute of Science and Technology (KAIST).

This paper proposes a novel two-layered confabulation architecture for an artificial creature to select a proper behavior considering the internally generated will and the context of the external environment consecutively. The architecture is composed of seven main modules for processing perception, internal state, context, memory, learning, behavior selection, and actuation. The two-layered confabulation in a behavior module is processed by a will-based confabulation and a context-based confabulation consecutively by referring to confabulation probabilities in a memory module. An arbiter in the behavior module chooses a proper behavior among the suggested ones from the two confabulations, which is to be put into an action. To demonstrate the effectiveness of the proposed architecture, experiments are carried out for an artificial creature, implemented in the 3-D virtual environment, which behaves as per its will considering the context in the environment.

This paper proposes the evolutionary generation of an artificial creature’s personality by using the concept of multi-objective optimization. The artificial creature has its own genome and in which each chromosome consists of many genes that contribute to defining its personality. The large number of genes allows for a highly complex system, however it becomes increasingly difficult and time-consuming to ensure reliability, variability and consistency for the artificial creature’s personality while manually assigning gene values for the individual genome. Moreover, there needs user’s preference to obtain artificial creature’s personality by using evolutionary generation. Preference is strongly depend on each user and most of them would have difficulty to define their preference as a fitness function. To solve this problem, this paper proposes multi-objective generating process of an artificial creature’s personality. Genome set is evolved by applying strength Pareto evolutionary algorithm (SPEA). To facilitate the individuality of generated artificial creature, complement of (1-k) dominance and pruning method considering deviation are proposed. Ob tained genomes are tested by using an artificial creature, Rity in the virtual 3D world created in a PC.

This paper presents the components and overall architecture of the ubiquitous robot (Ubibot) system developed to demonstrate ubiquitous robotics, a new paradigm for integrated services. The system has been developed on the basis of the definition of the ubiquitous robot as that of encompassing the software robot Sobot, embedded robot Embot and the mobile robot Mobot. This tripartite partition, which independently manifests intelligence, perception and action, enables the abstraction of intelligence through the standardization of sensory data and motor or action commands. The Ubibot system itself is introduced along with its component subsystems of Embots, the position Embot, vision Embot and sound Embot, the Mobots of Mybot and HSR, the Sobot, Rity, a virtual pet modeled as an artificial creature, and finally the middleware which seamlessly enables interconnection between other components. Three kinds of experiments are devised to demonstrate the fundamental features, of calm sensing, context awareness and seamless service transcending the spatial limitations in the abilities of earlier generation personal robots. The experiments demonstrate the proof of concept of this powerful new paradigm which shows great promise.

This study investigates the nonlinear dynamics of traction for an omniwheel mobile robot platform. A nonlinear slip model is incorporated into the dynamics of the system and the resulting equations of motion are derived using Euler-Lagrange formulation. These are additionally transformed into slip-space, where the dynamical equations lend themselves to a convenient analysis of the slip dynamics. The conventional assumptions for ideal rolling are also explored and a reduced expression for the nonlinear dynamics is generated for such situations. Preliminary explorations toward a comprehensive analysis of the dynamics for omniwheel platforms under various control schemes is also initiated.

Complex navigational commands require a walk ing pattern generator that is able to modify the pattern at any point in the walking gait. This paper utilizes the 3D-LIPM(Linear Inverted Pendulum Model) for generating a walking pattern, but introduces a method that allows for manipulation of the ZMP over the convex hull of the foot polygon whilst in single support phase. This permits a range of dynamic walking states that are not achievable using the conventional 3D-LIPM. These walking states are defined as a feasible region. A real-time algorithm is then developed, which follows the cue from a complex navigational command exactly when the desired walking state is in the feasible region and chooses the nearest feasible motion when it lies outside the feasible region. The proposed scheme is both simulated and implemented on the humanoid robot HanSaRam-VII developed at RIT laboratory, KAIST.

Confabulations, where millions of items of relevant knowledge are applied in parallel in the human brain, are typically employed in thinking. This paper proposes a novel behavior selection architecture and memory-based learning method for an artificial creature based on two-layered confabulation. A behavior is selected by considering both internally generated will and the context of the external environment. Proposed behavior selection using a confabulation scheme is a parallel process in which a number of behaviors are considered simultaneously. An arbitration mechanism is employed to choose a proper behavior which is to be put into an action. Also memory-based behavior learning is proposed, w here the memory has the selection probabilities of behaviors based on will and context. The learning module updates the contents of memory according to the user-given reward or penalty signal. To demonstrate the effectiveness of the proposed scheme, an artificial creature is implemented in the 3D virtual environment such that it can interact with a human being considering its will and context.

In this paper, a new architecture is proposed for the efficient human interaction with software robot (Sobot) in a PDA and the Sobot transmission between PDAs. Sobot can move to any mobile device through a wireless communication network. This ability is required to follow its user by moving itself to a device in his/her new location. Sobot as an artificial creature, has genetic code which is a set of computerized codes representing the personality. It has an internal state which consists of motivation, homeostasis, and emotion. The data set of genetic code and current internal state are transmitted through the network when it moves to the other device. A main server manages IP addresses of registered devices, Sobot's data set, and Sobot's location for the transmission. To demonstrate the effectiveness of the proposed scheme, Sobot is implemented in windows mobile environment of PDA such that it can interact with a human being and move to the other mobile device without spatial limitation.

This paper discusses the personality of genetic robot and its evolving algorithm within the purview of the broader ubiquitous robot framework. Ubiquitous robot systems blends mobile robot technology (Mobot) with distributed sensor systems (Embot) and overseeing software intelligence (Sobot), for various integrated services. The Sobot is a critical question since it performs the dual purpose of overseeing intelligence as well as user interface. The Sobot is hence modelled as an artificial creature with autonomously driven behavior. The artificial creature has its own genome and in which each chromosome consists of many genes that contribute to defining its personality. This paper proposes evolving the personality of an artificial creature. A genome population is evolved such that it customized the genome satisfying a set of personality traits desired by the user. Evaluation procedure for each genome of the population is carried out in a virtual environment. Effectiveness of this scheme is demonstrated by using an artificial creature, Rity in the virtual 3D world created in a PC.

This paper presents a robotic doll with emotional and reflexive behaviors. The robotic doll imitates an animal's appearance to provide comfort in human interaction. Emotion is considered to show more natural behaviors and to interact with user more intimately. Behavior is selected based on reflex-ness and emotion. A bear-like robotic doll, GomDoll is developed with the full use of the available degrees of freedom, sensors, and emotional and reflexive architecture implemented on a micro-controller. Experimental results demonstrate the effectiveness of the proposed architecture.

This paper presents the development of generating dance performances of humanoid robot, HSR (HanSaRam)-VII, synchronized with Robonova, for entertainment. This heterogeneous team, RoboBees, participated in 'Robots at Play Award 2006' and nominated as one of top six teams in the world. A method of generating and combining both periodic motion (on-line pattern generator) and aperiodic motion (off-line pattern generator) is investigated along with the control architecture for HSR-VII. Detailed system descriptions of two types of humanoid robots, HSR-VII from KAIST and Robonova from miniROBOT Corp., are included with the experimental results. In addition, the time domain passivity compliance control system is introduced to guarantee the stable periodic motion by changing the initial planned trajectories. Video clip of RoboBees' performance is available at http://rit.kaist.ac.kr.

This paper proposes a control method to absorb the landing force or the ground reaction force for a stable dynamic walking of a humanoid robot. Humanoid robot may become unstable during walking due to the impulsive contact force of the sudden landing of its foot. Therefore, a control method to decrease the landing force is required. In this paper, time-domain passivity control approach is applied for this purpose. Ground and the foot of the robot are modeled as two one-port network systems that are connected, and exchange energy with each other. The time-domain passivity controller with admittance causality is implemented, which has the landing force as input and foot's position to trim off the force as output. The proposed landing force controller can enhance the stability of the walking robot from simple computation. The small-sized humanoid robot, HanSaRam-VII that has 27 DOFs, is developed to verify the proposed scheme through dynamic walking experiments.

This paper discusses the reason why QEA works and verifies how QEA works. The theoretical analysis of the simplified model of the segment process of QEA shows that QEA with a single individual for ONEMAX problem guarantees the global solution in terms of expected running number of generations. The analysis for exploration shows clearly that QEA starts with a global search scheme and changes automatically into a local search scheme as generation advances because of its inherent probabilistic mechanism, which leads to a good balance between exploration and exploitation. For comparison purpose, simulated annealing is considered with three test functions. The results support the conclusions derived from the theoretical analysis of QEA with a single individual.

This paper proposes a multiobjective evolutionary algorithm (MOEA) inspired by quantum computing, which is named quantum-inspired multiobjective evolutionary algorithm (QMEA). In the previous papers, quantum-inspired evolutionary algorithm (QEA) was proved to be better than conventional genetic algorithms for single-objective optimization problems. To improve the quality of the nondominated set as well as the diversity of population in multiobjective problems, QMEA is proposed by employing the concept and principles of quantum computing such as uncertainty, superposition, and interference. Experimental results pertaining to the multiobjective 0/1 knapsack problem show that QMEA finds solutions close to the Pareto-optimal front while maintaining a better spread of nondominated set.

This paper proposes a genetic representation method to evolve artificial creature’s personality by using artificial genome and evolutionary generative algorithm. The genome consists of computer-coded chromosomes. Based on the internal architecture, the chromosomes are designed for the genetic representation. They are composed of the fundamental genes, internal state related genes, and behavior related genes as essential components, which represent the personality and are used for the animal-like evolution in the simulated environment. To get a desired personality (genome), a proper fitness function is designed and genetic operators are applied to the population (genomes). The artificial creature, Rity, is developed in a virtual world of PC to test the effectiveness of this representation scheme.

In this paper, a method to reduce the landing impact force is proposed for a stable dynamic walking of a humanoid robot. To measure the meaningful landing impact force, a novel foot mechanism, which uses FSRs (force sensing resistors), is introduced as well. Humanoid robot might become unstable during the walking due to the impulsive contact force from the sudden landing of its foot. Therefore a new control method to decrease the landing impact force has been required. In this paper, time domain passivity control approach is applied for this purpose. Ground and the foot of the robot are modeled as two one-port network systems which are connected and exchanging energy each other. And, the time domain passivity controller which has the landing impact force as input and foot's position to trim off the force as output, is implemented. Unlike previous works, the proposed controller can guarantee the stability of the robot system without any dynamic model information at all. The small sized humanoid robot, HanSaRam-VI which has 25 DOFs, with the proposed foot mechanism is developed to verify the proposed approach through dynamic walking experiments

One of the key components of an ubiquitous robot (Ubibot) is the software robot (Sobot) which can communicate with embedded robot (Embot) and mobile robot (Mobot). Sobot is a virtual robot, which has the ability to move to any place or connect to any device through a network in order to overcome spatial limitations. Embot has the capability to sense the surroundings, interpret the context of the environment, and can communicate with Sobot and Mobot. On the other hand, a Mobot provides integrated mobile service, though it has spatial limitations. To incorporate Sobot, Embot, and Mobot reliably as an ubibot, middle layer is needed to arbitrate different protocols among them. This paper focuses on incorporating Sobot and Mobot for providing seamless and context-aware services to human beings. To implement the incorporation of them, the basic concept and structure of the middle layer are presented. The effectiveness of the middle layer for Sobot and Mobot is demonstrated through the real experiments.

For the purpose of a humanoid robot's stable walking or running, it is important to absorb landing force or ground reaction force which is generated when the robot's foot lands on the ground surface. The force can make the robot unstable, and the problem becomes serious if the robot runs. This paper proposes a control system, which can absorb the landing force of a humanoid robot. Time-domain passivity control approach is applied for this purpose. Ground and the robot's foot are modeled as two one-port network systems, which are connected and exchange energy with each other. The time-domain passivity controller has the landing force as input and controls the foot's position to reduce the force. The proposed controller can guarantee the stability of the robot system without need of any dynamic model information or control parameters. Using small sized humanoid robot, dynamic walking experiments are performed to verify the proposed scheme, and its efficiency is shown from the comparison with the other scheme.

Humanoids require interaction with their environment through careful placement and utilisation of its contacts to realise a desired motion. In most cases however, contact placement is constrained and there subsequently exists many motions which are not realisable. When contacts are restricted to a horizontal ground plane, a zero moment point (ZMP) analysis is sufficient for determining the readability of a specified motion. In this paper we briefly review the fundamentals of a ZMP analysis and then extend these principles so that a rigourous analysis concerning the realisability of humanoid motion on a complex topology (uneven surface) can be explored.

When evolutionary algorithms are used for solving numerical constrained optimization problems, how to deal with the relationship between feasible and infeasible individuals can directly influence the final results. This paper proposes a novel ecology-inspired EA to balance the relationship between feasible and infeasible individuals. According to the feasibility of the individuals, the population is divided into two groups, feasible group and infeasible group. The evaluation and ranking of these two groups are performed separately. The number of parents from feasible group has a sigmoid relation with the number of feasible individuals, which is inspired by the ecological population growth in a confined space. The proposed method is tested using (/spl mu/, /spl lambda/) evolution strategies with 13 benchmark problems. Experimental results show that the proposed method is capable of improving performance of the dynamic penalty method for constrained optimization problems.

This work proposes a new face detection system using quantum-inspired evolutionary algorithm (QEA). The proposed detection system is based on elliptical blobs and principal component analysis (PCA). The elliptical blobs in the directional image are used to find the face candidate regions, and then PCA and QEA are employed to verify faces. Although PCA related algorithms have shown outstanding performance, there still exist some problems such as optimal decision boundary or learning capabilities. By PCA, we can obtain the optimal basis but they may not be the optimal ones for discriminating faces from non-faces. Moreover, a threshold value should be selected properly considering the success rate and false alarm rate. To solve these problems, QEA is employed to find out the optimal decision boundary under the predetermined threshold value which distinguishes between face images and non-face images. The proposed system provides learning capability by reconstructing the training database, which means that system performance can be improved as failure trials occur.

Most of the optimization problems in the real world have constraints. In recent years, evolutionary algorithms caught a lot of researchers' attention for solving constrained optimization problems. Infeasible individuals are often underrated by most of the current evolutionary algorithms when evolutionary algorithms are used for solving constraint optimization problems. This paper proposes an approach to balance the feasible and infeasible individuals. Feasible and infeasible individuals are divided into two groups: feasible group and infeasible group. The evaluation and ranking of these two groups are performed separately. Parents for reproduction are selected from the two groups by a parent selection method. Objective function and bubble sort method are selected as the fitness function and ranking method for the feasible group. One existing evolutionary algorithm: stochastic ranking method, is modified to evaluate and rank the infeasible group. The new method is tested using a (/spl mu/, /spl lambda/)-ES on 13 benchmark problems. The results show that the proposed method is capable of improving the searching performance of the stochastic ranking method.

From recent research on combinatorial optimization of the knapsack problem, quantum-inspired evolutionary algorithm (QEA) was proved to be better than conventional genetic algorithms. To improve the performance of the QEA, this paper proposes research issues on QEA such as a termination criterion, a Q-gate, and a two-phase scheme, for a class of numerical and combinatorial optimization problems. A new termination criterion is proposed which gives a clearer meaning on the convergence of Q-bit individuals. A novel variation operator H/sub /spl epsi// gate, which is a modified version of the rotation gate, is proposed along with a two-phase QEA scheme based on the analysis of the effect of changing the initial conditions of Q-bits of the Q-bit individual in the first phase. To demonstrate the effectiveness and applicability of the updated QEA, several experiments are carried out on a class of numerical and combinatorial optimization problems. The results show that the updated QEA makes QEA more powerful than the previous QEA in terms of convergence speed, fitness, and robustness.

In this paper, some guidelines for setting the parameters of quantum-inspired evolutionary algorithm (QEA) are presented. QEA is based on the concept and principles of quantum computing, such as a quantum bit and superposition of states. However, QEA is not a quantum algorithm, but a novel evolutionary algorithm. Like other evolutionary algorithms, QEA is also characterized by the representation of the individual, the evaluation function, and the population dynamics. From recent research on the knapsack problem, the results of QEA are better than those of CGA (conventional GA). Although the performance of QEA is excellent, there is relatively little or no research on the effects of different settings for its parameters. This paper describes some guidelines for setting these parameters. The guidelines are drawn up based on extensive experiments carried out for a class of combinatorial and numerical optimization problems. Through the guidelines, the performance of QEA can be maximized.

The stability augmentation system (SAS) is designed to improve the stability while parallel evolutionary optimization based on Lagrangian II (PEvolian II) is successfully applied to satisfy several constraints and to minimize the rising time. A controller to stabilize F-16 aircraft flying with a steady state around the altitude of 25,000 ft is described. The nonlinear pitching motion model of F-16 is linearized in the range of the flight envelope of velocity vs. altitude. Then applying feedback linearization stabilizes the statically unstable system. As the gain-scheduling method is introduced at various operating points within the flight envelope, the optimized controller is designed all over the envelope. parallel evolutionary optimization based on Lagrangian II is used to optimize the proportional and integral gains of the controller, satisfying complex nonlinear constraints.

In this paper, a two-phase evolutionary optimization scheme is proposed for obtaining optimal structure of fuzzy control rules and their associated weights, using evolutionary programming (EP) and the principle of maximum entropy (PME). The scheme consists of two phases: in the first phase, the rule structure and the scale factors for error, change of error and input are found by EP. The rule structure and the scale factors are encoded by integer and real number string, then varied by the proposed adjacent mutation and Gaussian mutation, respectively. In the second phase, the PME is employed to determine the weights of each rule so that all the fuzzy control rules can be utilized to the greatest extent. The optimization of the second phase can be regarded as fine tuning for the output response of the controlled system. Only several decades of generation is needed for determining the weights in the second phase, so the time-varying plant or online adjustment can be dealt with. The effectiveness of the proposed scheme is demonstrated by computer simulations.

In this paper, a new two mode Q-learning using both the success and failure experiences of an agent is proposed for the fast convergence, which extends Q-learning, a well-known scheme used for reinforcement learning. In the Q-learning, if the agent enters into the "fail" state, it receives a punishment from environment. By this punishment, the Q value of the action which generated the failure experience is decreased. On the other hand, the proposed two mode Q-learning is based on both the normal and failure Q values for the selection of the action in a state-action space. To determine the failure Q value using the previous failure experience of the agent, it employs a failure Q value module. To demonstrate the effectiveness of the proposed method, it is compared with the conventional Q-learning in a goalie system to perform goalkeeping in robot soccer.

Most of the existing evolutionary algorithms for constrained problems derate the importance of the infeasible individuals. In these algorithms, feasible individuals might get more possibility to survive and reproduce than infeasible individuals. To recover the utility of infeasible individuals, a grouping-based evolutionary algorithm (GEA) for constrained problems is proposed in this paper. Feasible population and infeasible individuals are separated as two groups. Evaluation, rank and reproduction of these groups are performed separately. The only chance for the two groups to exchange information happens when the offspring replace the parents. Thus, the designer could pay more attention to the evolutionary process inside the group. The simulation results of four benchmark problems show the effectiveness of the proposed algorithm.

A controller to stabilize F-16 aircraft flying with a steady state around the altitude of 25,000 ft is considered. The nonlinear pitching motion model of F-16 is linearized in the range of the flight envelope of velocity vs. altitude. Then the statically unstable system is stabilized by applying feedback linearization. As the gain-scheduling method is introduced at various operating points within the flight envelope, the optimized controller is designed all over the envelope. Evolutionary Optimization based on Lagrangian (Evolian II) is used to optimize the controller gains P and I satisfying complex nonlinear constraints.

Describes the implementation of an Internet-based personal robot with three control modes, and experimental results on the remote control between KAIST, Korea and UC Davis, USA. The idea is to control a personal robot at the remote site (KAIST) by using a simulator provided at the local site (UC Davis). However, if the information of the current absolute position of the robot at its remote site cannot be estimated, the simulator may be useless. The absolute position of the robot can be determined by comparing a reference map with sensor information from sonar sensors and an electronic compass. A user can use three control modes-direct control mode, supervisory control mode and job scheduling mode, and monitor the current status of the robot using a graphic user interface of the simulator implemented with Java.

This paper proposes a novel evolutionary algorithm inspired by quantum computing, called a quantum-inspired evolutionary algorithm (QEA), which is based on the concept and principles of quantum computing, such as a quantum bit and superposition of states. Like other evolutionary algorithms, QEA is also characterized by the representation of the individual, evaluation function, and population dynamics. However, instead of binary, numeric, or symbolic representation, QEA uses a Q-bit, defined as the smallest unit of information, for the probabilistic representation and a Q-bit individual as a string of Q-bits. A Q-gate is introduced as a variation operator to drive the individuals toward better solutions. To demonstrate its effectiveness and applicability, experiments were carried out on the knapsack problem, which is a well-known combinatorial optimization problem. The results show that QEA performs well, even with a small population, without premature convergence as compared to the conventional genetic algorithm.

This paper proposes a new parallel evolutionary algorithm called parallel quantum-inspired genetic algorithm (PQGA). Quantum-inspired genetic algorithm (QGA) is based on the concept and principles of quantum computing such as qubits and superposition of states. Instead of binary, numeric, or symbolic representation, by adopting the qubit chromosome as a representation, QGA can represent a linear superposition of solutions due to its probabilistic representation. QGA is suitable for parallel structures because of rapid convergence and good global search capability. That is, QGA is able to possess the two characteristics of exploration and exploitation simultaneously. The effectiveness and the applicability of PQGA are demonstrated by experimental results on the knapsack problem, which is a well-known combinatorial optimization problem. The results show that PQGA is superior to QGA as well as other conventional genetic algorithms.

This paper proposes a hybrid parallelization of evolutionary algorithms (EAs) utilizing PC clustering environments to solve constrained numerical optimization problems. In the proposed parallel structure, the coarse-grained parallel EAs (PEAs) were implicated in upper level and the fine-grained PEAs were used in lower level. The design of effective evolutionary algorithms (EAs) is to obtain a proper balance between exploration and exploitation. The balance can be controlled by the spread rate and the migration of the best individuals. In the hybrid structure, the spread rate is high in lower level coarse-grained structure and low in upper level globally structure. The diversity is promoted by dividing individuals to several groups and migrating individual between them. By utilizing large number of processors, the optimization performance as well as the computation time were improved. Simulation results indicate that hybrid parallel EAs using the proposed structure have better performance in constrained numerical optimization problems than coarse-grained, or fine-grained parallel EAs, which are dedicated parallelization methods in previous work.

Describes the implementation of an Internet-based personal robot with direct Internet control architecture which is insensitive to the inherent Internet time delay. The personal robot can be controlled by using a simulator provided at a local site. However, a large Internet time delay may make some control inputs distorted. Moreover, since it is affected by the number of the Internet nodes and loads, this delay is variable and unpredictable. The proposed control architecture guarantees that the personal robot can reduce the path error and the time difference between a virtual robot at the local site and a real robot at the remote site. Simulations and experimental results in the real Internet environment demonstrate the effectiveness and applicability of the Internet-based personal robot with the proposed Internet control architecture.

This paper proposes a novel evolutionary computing method called a genetic quantum algorithm (GQA). GQA is based on the concept and principles of quantum computing such as qubits and superposition of states. Instead of binary, numeric, or symbolic representation, by adopting qubit chromosome as a representation GQA can represent a linear superposition of solutions due to its probabilistic representation. As genetic operators, quantum gates are employed for the search of the best solution. Rapid convergence and good global search capability characterize the performance of GQA. The effectiveness and the applicability of GQA are demonstrated by experimental results on the knapsack problem, which is a well-known combinatorial optimization problem. The results show that GQA is superior to other genetic algorithms using penalty functions, repair methods and decoders.

This paper proposes a complete binary tree topology and two efficient migration methods in fine-grained parallel evolutionary algorithms (FGPEAs) to solve constrained numerical optimization problems. The design of effective evolutionary algorithms (EAs) is to obtain a proper balance between exploration and exploitation. The balance can be controlled by the spread rate and the migration of the best individuals. A complete binary tree topology, which slows down the spread rate, is used for exploration to solve the heavily constrained problems. Two migration methods are also employed to prevent a superior individual from taking almost all the subpopulations and to facilitate the possibility of global search. One is the restriction of migration according to the migration times and the other is the modified individual migration by the mutation operators. The simulation results indicate that FGPEA using the proposed migration methods has better performance in constrained numerical optimization problems, and the FGPEA with the tree topology and the proposed migration methods shows good performance on heavily constrained numerical optimization problems.

Two evolutionary programming (EP) methods are proposed for handling nonlinear constrained optimization problems. The first, a hybrid EP, is useful when addressing heavily constrained optimization problems both in terms of computational efficiency and solution accuracy. But this method offers an exact solution only if both the mathematical form of the objective function to be minimized/maximized and its gradient are known. The second method, a two-phase EP (TPEP) removes these restrictions. The first phase uses the standard EP, while an EP formulation of the augmented Lagrangian method is employed in the second phase. Through the use of Lagrange multipliers and by gradually placing emphasis on violated constraints in the objective function whenever the best solution does not fulfill the constraints, the trial solutions are driven to the optimal point where all constraints are satisfied. Simulations indicate that the TPEP achieves an exact global solution without gradient information, with less computation time than the other optimization methods studied here, for general constrained optimization problems.

In this paper, a time-varying two-phase (TVTP) optimization neural network is proposed based on the two-phase neural network and the time-varying programming neural network. The proposed TVTP algorithm gives exact feasible solutions with a finite penalty parameter when the problem is a constrained time-varying optimization. It can be applied to system identification and control where it has some constraints on weights in the learning of the neural network. To demonstrate its effectiveness and applicability, the proposed algorithm is applied to the learning of a neo-fuzzy neuron model.