In this study, we present a method to grasp diverse unseen real-world objects using an off-policy actor-critic deep reinforcement learning (RL) with the help of a simulation and the use of as little real-world data as possible. Actor-critic deep RL is unstable and difficult to tune when a raw image is given as an input. Therefore, we use state representation learning (SRL) to make actor-critic RL feasible for visual grasping tasks. Meanwhile, to reduce visual reality gap between simulation and reality, we also employ a typical pixel-level domain adaptation that can map simulated images to realistic ones. In our method, as the SRL model is a common preprocessing module for simulated and real-world data, we perform SRL using real and adapted images. This pixel-level domain adaptation enables the robot to learn grasping skills in a real environment using small amounts of real-world data. However, the controller trained in the simulation should adapt to the real world efficiently. Hence, we propose a method combining a typical pixel-level domain adaptation and the proposed SRL model, where we perform SRL based on a feature-level domain adaptation. In evaluations of vision-based robotics grasping tasks, we show that the proposed method achieves a substantial improvement over a method that only employs a pixel-level or domain adaptation.

For robotic grasping tasks with diverse target objects, some deep learning-based methods have achieved state-of-the-art results using direct visual input. In contrast, actor-critic deep reinforcement learning (RL) methods typically perform very poorly when applied to grasp diverse objects, especially when learning from raw images and sparse rewards. To render these RL techniques feasible for vision-based grasping tasks, we used state representation learning (SRL), in which we encode essential information for subsequent use in RL. However, typical representation learning procedures are unsuitable for extracting pertinent information for learning grasping skills owing to the high complexity of visual inputs for representation learning, in which a robot attempts to grasp a target object. We found that the proposed preprocessed input image is the key to capturing effectively a compact representation. This enables deep RL to learn robotic grasping skills from highly varied and diverse visual inputs. Further, we demonstrate the effectiveness of the proposed approach with varying levels of preprocessing in a realistic simulated environment. We also describe how the resulting model can be transferred to a real-world robot and also demonstrate a 68% success rate on real-world grasp attempts.

Abstract We developed a new framework to generate hand and finger grasping motions. The proposed framework provides online adaptation to the position and orientation of objects and can generate grasping motions even when the object shape differs from that used during motion capture. This is achieved by using a mesh model, which we call primitive object grasping (POG), to represent the object grasping motion. The POG model uses a mesh deformation algorithm that keeps the original shape of the mesh while adapting to varying constraints. These characteristics are beneficial for finger grasping motion synthesis that satisfies constraints for mimicking the motion capture sequence and the grasping points reflecting the shape of the object. We verify the adaptability of the proposed motion synthesizer according to its position/orientation and shape variations of different objects by using motion capture sequences for grasping primitive objects, namely, a sphere, a cylinder, and a box. In addition, a different grasp strategy called a three‐finger grasp is synthesized to validate the generality of the POG‐based synthesis framework.

For smart manufacturing, an automated robotic assembly system built upon an autoprogramming environment is necessary to reduce setup time and cost for robots that are engaged in frequent task reassignment. This article presents an approach to the autoprogramming of robotic assembly tasks with minimal human assistance. The approach integrates “robotic learning of assembly tasks from observation” and “robotic embodiment of learned assembly tasks in the form of skills.” In the former, robots observe human assembly operations to learn a sequence of assembly tasks, which is formalized into a human assembly script. The latter transforms the human assembly script into a robot assembly script in which a sequence of robot-executable assembly tasks are defined based on action planning supported by workspace modeling and simulated retargeting. The assembly tasks, in the form of the robot assembly script, are then implemented via pretrained robot skills. These skills aim to enable robots to execute difficult tasks that involve inherent uncertainties and variations. We validate the proposed approach by building a prototype of the automated robotic assembly system for a power breaker and an electronic set-top box. The results verify that the proposed automated robotic assembly system is not only feasible but also viable, as it is associated with a dramatic reduction in the human effort required for automating robotic assembly.

We present a novel feature matching algorithm that systematically utilizes the geometric properties of image features such as position, scale, and orientation, in addition to the conventional descriptor vectors. In challenging scenes, in which repetitive structures and large view changes are present, it is difficult to find correct correspondences using conventional approaches that only use descriptors, as the descriptor distances of correct matches may not be the least among the candidates. The feature matching problem is formulated as a Markov random field (MRF) that uses descriptor distances and relative geometric similarities together. Assuming that the layout of the nearby features does not considerably change, we propose the bidirectional transfer measure to gauge the geometric consistency between the pairs of feature correspondences. The unmatched features are explicitly modeled in the MRF to minimize their negative impact. Instead of solving the MRF on the entire features at once, we start with a small set of confident feature matches, and then progressively expand the MRF with the remaining candidate matches. The proposed progressive approach yields better feature matching performance and faster processing time. Experimental results show that the proposed algorithm provides better feature correspondences in many challenging scenes, i.e., more matches with higher inlier ratio and lower computational cost than those of the state-of-the-art algorithms. The source code of our implementation is open to the public.

This paper presents a 3-DOF linkage-driven underactuated finger for achieving several different grasp modes. This finger mechanism is constructed by stacking one five-bar mechanism over one double parallelogram. This special architecture allows for installing all of the actuators on the base. A 2-finger robotic gripper having three actuators is developed for performing self-adaptive power grasping, parallel precision grasping, grasping during contact with the environment, and other grasping tasks through changing the orientation of the distal phalanx actively. The performance of the gripper is confirmed through both simulation and practical grasping experiment.

In this study, we aim at grasping a single target object in a cluttered environment using a robotic arm. While dexterous grasp for various shapes of objects is not considered in this work, we focus on developing the method to mitigate clutter near the target object as soon as quickly. For this purpose, we propose a method to generate nonlinear pushing motions for object singulation based on an off-the-shelf machine learning algorithm and a typical semantic segmentation algorithm. Through experiments, we show that the success rate of robotic grasping is considerably improved by the proposed pushing behavior. And notably, the nonlinear pushing trajectories allows the robot to perform singulation of the target object in a cluttered environment with fewer trials than linear pushing usually pursued in related works.

In this paper, we propose a method to model social interaction between a human and a virtual avatar. To this end, two human performers fist perform social interactions according to the Learning from Demonstration paradigm. Then, the relative relevance of all joints of both performers should be reasonably modeled based on human demonstrations. However, among all possible combinations of relative joints, it is necessary to select only some of the combinations that play key roles in social interaction. We select such significant features based on the joint motion significance, which is a metric to measure the significance degree by calculating both temporal entropy and spatial entropy of all human joints from a Gaussian mixture model. To evaluate our proposed method, we performed experiments on five social interactions: hand shaking, hand slapping, shoulder holding, object passing, and target kicking. In addition, we compared our method to existing modeling methods using different metrics, such as principal component analysis and information gain.

Abstract In this paper, we propose a novel motion controller for the online generation of natural character locomotion that adapts to new situations such as changing user control or applying external forces. This controller continuously estimates the next footstep while walking and running, and automatically switches the stepping strategy based on situational changes. To develop the controller, we devise a new physical model called an inverted‐pendulum‐based abstract model (IPAM). The proposed abstract model represents high‐dimensional character motions, inheriting the naturalness of captured motions by estimating the appropriate footstep location, speed and switching time at every frame. The estimation is achieved by a deep learning based regressor that extracts important features in captured motions. To validate the proposed controller, we train the model using captured motions of a human stopping, walking, and running in a limited space. Then, the motion controller generates human‐like locomotion with continuously varying speeds, transitions between walking and running, and collision response strategies in a cluttered space in real time.

Feature-based image retrieval and feature matching have been used together in many applications, but they have been treated as two separate problems. We propose an unified approach which, for a query image, finds a set of candidate images together with feature matching results. By considering the local geometric consistency of neighboring features, we can find more and better feature matches even in challenging situations. Since the proposed forward/backward matching and non-linear diffusion run very efficiently, they can be used in the candidate image selection and improve the image retrieval performance significantly. Through quantitative comparisons we show that the proposed approach performs better than the recent state-of-the-art feature matching algorithms and image retrieval algorithms.

In this study, we propose a motion generation technique which generates natural motions based on a double inverted pendulum model (DIPM) and motion capture data (Mocap). While generating the motions, the proposed controller keeps the balance of the character. A DIPM uses a hip strategy to maintain the character's stability so that the zero moment point (ZMP) stays inside the support area, composed by the feet. The naturalness of the generated motion is inherited from mocap data by aligning the motion capture sequence with the DIPM. We match the DIPM with the motion capture data in order to satisfy both the character's stability and the naturalness. To validate the proposed motion generation technique, we use two kinds of motion capture data: a balancing motion under external forces and a grasping motion with a box.

In this paper, we propose an approach for a robot to adapt time scale of motions to guarantee temporal constraints of robot tasks. To this end, an selective Gaussian mixture model is first presented based on motion significance. Also, an selective Gaussian mixture regression is then presented to select time indexes for regression of robot motions from Gaussian mixtures based on motion significance. Motion significance indicates the relative significance of every motion frame, which is defined as a set of data points acquired at a time index of multiple motion trajectories. Such motion significance is measured by combining both temporal entropy and spatial entropy of a motion frame, based on the analysis of Gaussian mixtures. To evaluate our proposed method, we perform the experiment in which a robot draws a figure on a whiteboard. In addition, we discuss how to develop the adaptive time scaling by adjusting piecewise Gaussian distributions involved in Gaussian mixtures.

This paper proposes a passivity-based nonlinear disturbance observer (DOB) design for a powered upper-limb robot control. The proposed DOB allows for the nonlinearities of the robot dynamics, whereas the typical DOB designs cannot. Moreover, by virtue of the passivity property, human operator and environmental interactions can be embedded in the control loop. As a DOB, the proposed approach has a disturbance observation property that makes the actual robot behave like a nominal model selected by the user. Performance analysis proposes a gain tuning rule. In experimental validation, actual powered upper-limb robot is used to perform unknown payload carrying applications.

We present a real-time framework for planning natural and smooth grasping motions in an online manner based on the interaction with human. The proposed framework is able to change its grasp strategies agilely according to the interaction with human. Given human demonstrations, we develop a motion field graph consisting of nodes and edges where the nodes contains reference finger poses and their time derivatives, and the edges indicates the similarity between a pair of nodes. Based on the graph, a new grasping motion can be planned that adapts to the changes of the environment and the interaction. The motion field guarantees smooth motions by integrating the velocities obtained from the demonstrations. To validate the framework, we build a demo system where a human can hand a cup over to a tele-operated robot or a virtual humanoid avatar which are controlled by an another person at a remote location. Also, the virtual avatar can grasp and manipulate a cola can.

We propose a topology-based Bayesian visual navigation framework, with which we represent the world environment as a collection of scenes. The proposed topological map consists of nodes represented as a bag of visual line words and edges represented as both an adjacency list and relative motion information to perform the transition between topological nodes. Our proposed Bayesian localization framework uses two measurement models: a visual line word-based place model and a path-matching model. To enable a mobile robot to reach to a desired destination, a coastal path is planned in such a way that maximizes the possibility of following the reference path defined as a sequence of scenes encountered when building the topological map, while avoiding possible collisions by leveraging a local grid map constructed in real time from Kinect depth information. To show the validity of our proposed framework, we provide several extensive experimental results in several indoor environments under different conditions, such as illumination changes and visual occlusions.

In this paper, we propose a keyframe-based online object learning and detection method. To manage appearance changes of target objects, the proposed method incrementally updates an object database using detection results. One of the major problems in updating the appearance model is that the object model can gradually be degraded by accumulated errors and biased to specific views. To solve this problem, our object model is updated according to the selected keyframes, which not only help memorize important views of target objects, but also prevent the holistic appearance model from overfitting. The database is represented as a graph of the registered images, and the importance of the database images is measured by analyzing the constructed graph. Then, the redundant or less important images are discarded from the database. As a result, the database is efficiently maintained while new views of the objects are gradually added. The experimental results demonstrate that the proposed algorithm efficiently maintains the object database and improves the detection performance compared to previous incremental object learning and detection algorithms.

In this paper, we propose a scene clustering algorithm which uses straight line features. Scenes are represented as nodes in the graph, and each connectivity between nodes is calculated by a pre-trained vocabulary tree. By applying a spectral clustering algorithm to the constructed graph, the scenes are partitioned into k groups where k is determined by the proposed estimation method. Instead of using the standard eigenvalue analysis, the optimal k is computed so that the partitioned graph becomes to have strong intra-class correlations while inter-class correlations are relatively weak. As a result, scenes are adaptively clustered according to the environmental changes. The clustering performance of the proposed method is quantitatively evaluated with three image sequences captured in challenging environments. Experimental comparisons demonstrate that our line-based algorithm outperforms existing algorithms utilizing other types of features as well as the produced scene clustering results are more human-like.

Dynamic motion of human shows kinematic aspects related to storing elastic energy in skeletal muscle. This results from joint stiffness modulation and as a consequence, countermovement which is opposite to the intended motion is observed. We propose a segmentation algorithm based on a hidden semi-Markov model that infers dynamic motion phases probabilistically from sEMG observations during countermovement jump. In addition, parameter re-estimation of both left-right state transition and restriction of state duration is applied to reduce frequent state transition due to large variation of sEMG observation probability. In experiments, the segmentation of motion phases using sEMG identified the phases of the vertical position of torso successfully and the parameter re-estimation reduced both the error rate and the transition occurrence.

In this paper, we propose a novel online multiobject learning and detection algorithm. From single seed images of the target objects, our algorithm detects these objects in the input sequence, and incrementally updates the databases with the detection results. Reasonably sized databases are maintained as graphs of the registered images, while new views of the objects are added as the detection proceeds. The importance of the registered images is computed using our ranking algorithm, and redundant images are pruned from the database. The proposed method fully utilizes graphical representation to detect and recognize objects. A 3D model of a candidate object is built on-the-fly using the retrieved images, and initially undetected features are hallucinated for further matching and verification. This process improves the detection performance compared to the baseline algorithm. Object/background feature classification and object-likelihood maps effectively keep noisy background features from being added to the databases. The experimental results demonstrate that the proposed algorithm efficiently maintains the object databases and achieves better performance.

This paper presents a graph-based visual simultaneous localization and mapping (SLAM) system using straight lines as features. Compared with point features, lines provide far richer information about the structure of the environment and make it possible to infer spatial semantics from the map. Using a stereo rig as the sole sensor, our proposed system utilizes many advanced techniques, such as motion estimation, pose optimization, and bundle adjustment. We use two different representations to parameterize 3-D lines in this paper: Plücker line coordinates for efficient initialization of newly observed line features and projection of 3-D lines, and orthonormal representation for graph optimization. The proposed system is tested with indoor and outdoor sequences, and it exhibits better reconstruction performance against a point-based SLAM system in line-rich environments.

In this paper, we propose a novel method for learning motor skills based on grossness and fineness of movements involved in daily-life tasks. Grossness and fineness depend on the degrees of complexity (i.e., linear combinations between basis vectors) and repeatability (i.e., repeat accuracies between multiple trials) of such movements. In such a daily-life task, a robot's movements are usually related to a task-relevant object. Therefore, the complexity and the repeatability should be acquired from datasets that include the spatial and temporal relationships between a robot and a task-relevant object. To measure the degree of complexity, correlations are first obtained from each data by canonical correlation analysis. To measure the degree of repeatability, variations are then obtained from covariances between datasets acquired by multiple trials. The grossness and fineness are finally acquired by combining the correlations and the variations. To learn a motor skill, a Gaussian Mixture Model (GMM) is estimated using well-known methods as Principal Component Analysis (PCA), k-means, Bayesian Information Criterion (BIC), and Expectation-Maximization (EM) algorithms. First, initial parameters of a GMM are estimated by weighting a conventional k-means algorithm with the grossness and fineness. Based on PCA, BIC, and EM algorithms, the GMM is then estimated using the initial parameters and a robot's motion trajectories. To validate our proposed methods, the GMM is evaluated in terms of reproduction and recognition using a robot arm that performs two daily-life tasks: cookie-decorating and constrained-delivering tasks.

In this paper, we propose a visual place recognition algorithm which uses only straight line features in challenging outdoor environments. Compared to point features used in most existing place recognition methods, line features are easily found in man-made environments and more robust to environmental changes such as illumination, viewing direction, or occlusion because they are more likely to be extracted from structures. Candidate matches are found using a vocabulary tree and their geometric consistency is verified by a motion estimation algorithm using line segments. The proposed algorithm operates in real-time, and it is tested with a challenging real-world dataset with more than 10,000 database images acquired in urban driving scenarios.

In this paper, we propose a complexity-based motion feature learning method and hierarchical spatio-temporal naïve Bayes classifier for human action recognition. As a motion feature learning method, we developed a complexity-based subsequence of time series clustering (C-STSC) method to learn time series codewords from a human motion trajectory. The key to the C-STSC method is to measure the importance of each subsequence in time series data through to use of a complexity measure. Next, time series codewords are learned on the basis of the important subsequences by using a clustering algorithm. Moreover, we also propose a hierarchical spatio-temporal naïve Bayes classifier (HST-NBC) to classify the C-STSC features, where both the codeword-type and its spatio-temporal information is explicitly represented as a composite node in a Bayesian network framework. To validate the proposed method, we present experimental results of the proposed approach with respect to several open datasets.

A Layered Hidden Markov Model (LHMM) has been usually used for recognizing various human activities. In such a LHMM, the performance tends to be improved than that of a single layered HMM. To further enhance the performance of such a LHMM, in this paper, we propose a brain-inspired feedback mechanism. For this achievement, the LHMM is first modeled using a set of training data that the semantic information (i.e., labels of data) is attached. In the inference phase, the semantic information is produced from the HMMs associated with the upper layers of the LHMM, and then the semantic information is used to improve the performances of the lower layers in the next inference step. Consequently, these interactive feed-forward and feedback information can dramatically improve the performance of the LHMM. To validate our proposed method, we compare the performance of our LHMM (i.e., with feedback mechanism) with that of a standard LHMM (i.e., with no feedback mechanism) using twenty-four human activities, which occur frequently when a human cooks.

Abstract This study aims to develop a controller for use in the online simulation of two interacting characters. This controller is capable of generalizing two sets of interaction motions of the two characters based on the relationships between the characters. The controller can exhibit similar motions to a captured human motion while reacting in a natural way to the opponent character in real time. To achieve this, we propose a new type of physical model called a coupled inverted pendulum on carts that comprises two inverted pendulum on a cart models, one for each individual, which are coupled by a relationship model. The proposed framework is divided into two steps: motion analysis and motion synthesis. Motion analysis is an offline preprocessing step, which optimizes the control parameters to move the proposed model along a motion capture trajectory of two interacting humans. The optimization procedure generates a coupled pendulum trajectory which represents the relationship between two characters for each frame, and is used as a reference in the synthesis step. In the motion synthesis step, a new coupled pendulum trajectory is planned reflecting the effects of the physical interaction, and the captured reference motions are edited based on the planned trajectory produced by the coupled pendulum trajectory generator. To validate the proposed framework, we used a motion capture data set showing two people performing kickboxing. The proposed controller is able to generalize the behaviors of two humans to different situations such as different speeds and turning speeds in a realistic way in real time.

For a robot to interact with a person effectively, it needs to predict future events that will be caused by the person to occur. By predicting events, a robot can take some preparative actions to reduce the waiting time and greatly improve the interaction. To select the best proactive actions and the best times for those actions, we propose a hybrid temporal influence diagram that can make proactive plans that minimize expected waiting times in a human-robot interaction. To validate our proposed method, we show experimental results for a robotic assistant in a manual assembly task.

In manipulation tasks, skills are usually modeled using the continuous motion trajectories acquired in the task space. The motion trajectories obtained from a human's multiple demonstrations can be broadly divided into four portions, according to the spatial variations between the demonstrations and the time spent in the demonstrations: the portions in which a long/short time is spent, and those in which the spatial variations are large/small. In these four portions, the portions in which a long time is spent and the spatial variation is small (e.g., passing a thread through the eye of a needle) are usually modeled using a small number of parameters, even if such portions represent the movement that is essential for achieving the task. The reason for this is that these portions are slightly changed in the task space as compared with the other portions. In fact, such portions should be densely modeled using more parameters (i.e., overfitting) to improve the performance of the skill because the movements of those portions must be accurately executed to achieve the task. In this paper, we propose a method for adaptively fitting these skills based on the temporal and the spatial entropies calculated by a Gaussian mixture model. We found that it is possible to retrieve accurate motion trajectories as compared with those of well-fitted models, whereas the estimation performance is generally higher than that of an overfitted model. To validate our proposed method, we present the experimental results and evaluations when using a robot arm that performed two tasks.

Most visual simultaneous localization and mapping systems use point features as their landmarks and adopt point-based feature descriptors to recognize them. Compared to point landmarks, however, lines have strength in conveying the structural information of the environment. Despite the benefit, they have not been widely used because lines are more difficult in detecting, tracking, and recognizing, and this delayed the use of lines as landmarks. In this paper, we propose a place recognition algorithm using straight line features, which enables reliable loop closure detections in large complex environments under significant illumination changes. A vocabulary tree trained with mean standard-deviation line descriptor is used in finding the candidate matches between keyframes, and a Bayesian filtering framework enables reliable keyframe matching for large-scale loop closures. The proposed algorithm is compared with state-of-the-art point-based methods using scale-invariant feature transform or speeded up robust features. The experimental results show that the proposed method outperforms the others in challenging indoor environments.

We propose Bayesian approaches for semantic mapping, active localization and local navigation with affordable vision sensors. We develop Bayesian model of egocentric semantic map which consists of spatial object relationships and spatial node relationships. Our topological-semantic-metric (TSM) map has characteristic that a node is one of the components of a general topological map that contains information about spatial relationships. In localization part, view dependent place recognition, reorientation and active search are used for robot localization. A robot estimates its location by Bayesian filtering which leverages spatial relationships among observed objects. Then a robot can infer the head direction to reach a goal in the semantic map. In navigation part, a robot perceives navigable space with Kinect sensor and then moves to goal location while preserving reference head direction. If obstacles are founded in front, then a robot changes the head direction to avoid them. After avoiding obstacles, a robot performs active localization and finds new head direction to goal location. Our Bayesian navigation program provides how a robot should select either an action for following line of moving direction or action for avoiding obstacles. We show that a mobile robot successfully navigates from starting position to goal node while avoiding obstacles by our proposed semantic navigation system with TSM map.

To achieve a certain task, a skilligent robot should be able to learn the skills embedded in that task. Furthermore, the robot should be able to infer such skills to handle uncertainties and perturbations, since most robot tasks are usually daily-life tasks that include many unexpected situations. Therefore, we propose a unified skill learning and inference framework. The framework includes six processing modules: 1) a human demonstration process, 2) an autonomous segmentation process, 3) a dynamic movement primitive learning process, 4) a Bayesian network learning process, 5) a motivation graph construction process, and 6) a skill-inferring process. Based on the framework, the robot learns and infers situation-adequate and goal-oriented skills to handle uncertainties and human perturbations. To show the validity of our framework, some experimental results are illustrated using a robot arm that performs a `tea service' task.

This paper proposes a robot recommender system, which uses a hybrid filtering method based on n-gram affective event model. Nowadays there are strong tendency to utilize a robot for educational services, which can provide educational contents to enhances individual student's motivation. However, the current service robots can be more holistic systems to offer personalized robotic services to satisfy every individuals by reflecting their preferences. Here, robotic service can be another field to meet personal need. Hybrid approaches of personalization technology that combine collaborative filtering approaches and content-based approaches are proposed over the last decade. Especially, n-gram based approaches are proposed to utilize sequential information from very large data sets. This paper suggests an extends affective event model and its n-gram model combining fact semantic knowledge, event episodic knowledge and emotion. To show the validity of the proposed approach, we applied the scenario of English learning. The experiment results shows that an educational service robot recommend two students as different content types, even though they miss same question.

Knowledge plays significant roles in real environments to successfully complete service tasks. Almost all knowledge inference mechanisms assume complete and correct knowledge about environments. However, these environments are said to be partially observable and uncertain with respect to service robots. Thus, a robot cannot often complete the given service tasks. To cope with such a difficulty, we propose a novel weighted action-coupled semantic network (wASN), which represents the world by symbolic relationships and their weights, and its association mechanism using bi-directional spreading activation model which recommends goal-oriented actions from action-coupled knowledge. The improvisational actions provide more opportunities for a service robot to achieve goals.

For effective human-robot interaction, a robot should be able to make prediction about future circumstance. This enables the robot to generate preparative behaviors to reduce waiting time, thereby greatly improving the quality of the interaction. In this paper, we propose a novel probabilistic temporal prediction method for proactive interaction that is based on a Bayesian network approach. In our proposed method, conditional probabilities of temporal events can be explicitly represented by defining temporal nodes in a Bayesian network. Utilizing these nodes, both temporal and causal information can be simultaneously inferred in a unified framework. An assistant robot can use the temporal Bayesian network to infer the best proactive action and the best time to act so that the waiting time for both the human and the robot is minimized. To validate our proposed method, we present experimental results for case in which a robot assists in a human assembly task.

Robots need to have knowledge of their environment to be able to successfully complete service tasks. Most knowledge inference mechanisms assume complete and correct knowledge about the environment. Real world environments are often uncertain and only partially observable. Thus, intelligent service robots may have an incomplete knowledge base which includes true positives as well as false negatives and false positives. False negatives and false positives can prevent service robots from completing their service tasks. In the field of logical inference, false positives are a more significant problem compared to false negatives. A weighted ontology and association mechanism was proposed in previous research which recommended improvisational goal-oriented actions that could be applied in the case of false negatives. However, false positives are not usually matched in the existing ontological semantic network. Consequently, the association mechanism does not work. To deal with false positives, the weighted ontology and association mechanism were extended by adding additional nodes which are associated with epistemic actions. The proposed method was successfully evaluated and verified through experiments; results show that almost all problems associated with false positives and false negatives were resolved.

Object recognition is one of the most important applications of robotics. For object recognition, edge orientation is widely used as a primitive visual feature. However, a classical filter-based approach passes not only edges inside target orientation band but also edges outside. This can thus cause problem in the estimation of the true orientation of edge. This study proposes a filtering scheme to reduce the false-positive responses, i.e. edges outside target orientation band, and investigate a solution inspired by biological vision. Motivated by several psychophysical and neuro-physiological findings, we present a computational framework based on the basic mechanisms of cortical processing, i.e. feed-forward, lateral and feedback stages. In the feed-forward stage, our model uses a classical filter-based method to allow as many true orientation edges to pass through as possible. False responses in orientation channels are then inhibited by lateral interaction. The remaining undesired responses are suppressed through the feedback stage. We evaluated the performance of our model against classical filter-based methods such as Gabor and Neumann filtering using several artificial and natural images. The results validated the effectiveness of our approach.

In this paper, we present a vanishing point-based two-step loop closure method in a line-based monocular simultaneous localization and mapping (SLAM) system. Vanishing points can provide absolute directional landmarks for mobile robots. This guiding ability is important in that the observation of the vanishing points is invariant with respect to the robot pose. In our system, loop closure is performed in two steps: first, the accumulated heading error is reduced using an observation of previously registered vanishing points, and second, the observation of known floor lines allows for further pose correction. In this paper, we apply this method to solve the loop closure problem for a line-based SLAM within a corridor environment where the vanishing points are usually easily detectable. Experimental results show that our method is very efficient in a structured indoor environment.

Disparity computation in occluded or texture-less regions is considered to be a fundamental issue in dense stereo matching, but there is another practical issue that must be resolved before it can be used effectively in various robotics applications. This issue is the problem of intensity difference between corresponding pixels of an image pair. To tackle such problems, we present a dependable stereo matching algorithm using two-layer recurrent process and chaining search. Two-layer process integrates pixel and region-levels information through recurrent interaction. To estimate the precise disparities in occluded regions, reliable disparities in non-occluded region are propagated to occluded regions by the proposed chaining search. To test our algorithm, it was compared with two outstanding algorithms in Middlebury benchmark using Gaussian noisy images. The results validated the effectiveness of our approach.

For a robot to learn complex movement skills, programming by demonstration and/or learning by trial and error is necessary. Measuring the complexity of such movement skills is important to decide the appropriate learning model, and the required size of dataset and additional prior knowledge. To deal with measuring the complexity of movement skills for robots, we propose an information-theoretic complexity measure. By modeling proprioceptive as well as exteroceptive sensory data as a multivariate Gaussian distribution, movement skills can be modeled as a probabilistic model. Next, complexity of the movement skills is measured by using neural complexity. In addition to the original neural complexity measure, endogeneous changes in time of the movement skills are modeled by sampling in time and modeling as individual random variables. To evaluate our proposed complexity measure, several experiments are performed on real robotic movement skills.

In this paper, we propose a proactive assistant robot for human assembly tasks. In order to predict future events of human activities such as requesting parts for assembly, we use the temporal Bayesian network that can infer both causal probability and temporal distribution of an conditional event. Based on the temporal Bayesian network model of an human assembly task, we also show that the proactive assistant robot make human-robot-interaction quickly by temporal prediction of an event.

In this work, we propose methods for automatically generating primitive skills from demonstration of a task. Additionally, we propose methods for improving the existing primitive skills and adding the new primitive skills incrementally and automatically. To validate our proposed methods, we present experimental results of a human-like robot handling three gestures and a task for making coffee.

Two HMM-based threshold models are suggested for recognition and incremental learning of scenario-oriented human behavior patterns. One is the expected behavior threshold model to discriminate if a monitored behavior pattern is normal or not. The other model is the registered behavior threshold model to detect whether such behavior pattern is already learned. If a behavior patten is detected as a new one, an HMM is generated to represent the pattern, and then the HMM is used to update behavior clusters by hierarchical clustering process.

In this paper, we present a monocular SLAM method which uses vertical and the floor line features as sensory input. A line-based partial 3D map was built, in which many structural properties of the environment can be included. The vertical line and the floor line can be represented as simple 2D objects in the floor plane. These two heterogeneous types of line features were developed as independent SLAM features and combined to represent the environment in a more complete fashion. Although the vertical and floor line use different parameterization and initialization methods, their measurement models are integrated into a unified EKF framework. Experimental results show that this method is practical in a structured indoor environment.

This paper presents tracking algorithms of a mobile robot that follows a human body moving at high speed in unstructured outdoor environment. The mobile robot, which is a differential-driven wheel type, is equipped with one laser range finder to perform high-speed human tracking and obstacle avoidance. Initially, the waist part of the human body is found the best part of the human body for robust tracking by a laser range finder. Next an obstacle avoidance algorithm considering the relative velocity between the robot and obstacles is proposed. It is shown that the proposed human tracking algorithms based on the Kalman filter performs well even in the case of the target occlusions and temporary losses. These tracking algorithms have been applied to a marathoner following robot that carries the marathoner's personal belongings in many different outdoor environments.

A significant obstacle for service robots is the execution of complex tasks in real environments. For example, it is not easy for service robots to find objects that are partially observable and are located at a place which is not identical but near the place where the robots saw them previously. To overcome the challenge effectively, robot knowledge represented as a semantic network can be extremely useful. This paper presents an ontology-based unified robot knowledge framework that integrates low-level data with high-level knowledge for robot intelligence. This framework consists of two sections: knowledge description and knowledge association. Knowledge description includes comprehensively integrated robot knowledge derived from low-level knowledge regarding perceptual features, part objects, metric maps, and primitive behaviors, as well as high-level knowledge about perceptual concepts, objects, semantic maps, tasks, and contexts. Knowledge association uses logical inference with both unidirectional and bidirectional rules. This characteristic enables reasoning to be performed even when only a partial information is available. The experimental results that demonstrate the advantages of using the proposed knowledge framework are also presented.

In this paper, to provide a robot with information relative to structure of its environment, we propose a method to recognize types of structural corridor landmarks such as T-junction, L-junction, end of the corridor, using vanishing points-based visual image features and hidden Markov models. Several experimental results are illustrated to demonstrate the validity of the proposed approach in a real environment.

This paper presents a method of topological localization in a home environment using only low-cost sonar sensors. The proposed method utilizes a relative motion model to obtain a prior information of node probability. Moreover, the size of a local gridmap used for gridmap matching is determined adaptively by an entropy test of node probability. The proposed method results in a reliable and convergent localization result even under the uncertainty and sparsity of sonar sensros. Experimental results verify performance of the proposed topological localization in a real home environment.

In this paper, we propose a novel monocular SLAM method in corridor environment which employs Segments-on-Floor (SoF) as feature data. Given that the height of the camera and the angle between the camera and the floor are known, an image of the SoF can be efficiently distinguished from the other space-lines by a simple data-association method, deriving the line correspondence from a simplified homography matrix of two sequentially gathered images. Furthermore, use of SoF simplifies the analysis of the geometrical property of the camera projection matrix. Therefore, we can reconstruct SoF by using a one-step inverse projection. Once SoF is calculated from visual data processing, they are then used in a normal SLAM process as feature data. We employ a simple particle filter in our corridor SLAM. Experimental results show that it is sufficient for mapping a moderately sized building environment.

Understanding situations has been regarded as a highly difficult task due to its complexity, especially in case of human-augmented ones where the contexts of situations are largely influenced by human activity. Even though the complexity of situations can be solved by modeling the real environment, it is not easy to model which covers uncertain problems of real world effectively. For this, this paper proposes a fusion technology for service-oriented context reasoning combining low-level sensory patterns and high-level semantic knowledge using Hidden Markov Model (HMM) and ontology. Integrated temporal reasoning using our approach enables a service robot to understand human-augmented situations immediately whenever the critical situation happens. Experimental results show that the proposed method can successfully extract situations from continuous sensory signals with 80% percent reliability.

This paper presents a predictive visual recognition method to identify types of structural corridor landmark such as T-junctions, L-junctions and ends of the corridors, using vanishing point-based visual features and a two-layer HMM framework. The main advantage of the approach is that a mobile robot can recognize the types of structural landmark before reaching them. This is important for efficient robot navigation because it allows the robot to make navigation decision in advance. Several experimental results are provided to demonstrate the validity of the proposed approach in a real environment.

We first generate a probabilistic affordance to select an action based on motivation values. The affordance is designed as a multilayer naïve Bayesian classifier with respect to uncertainties and equivalence classes. The multilayer naïve Bayesian classifier is a probabilistic model with multiple layers of conditional probability tables and/or probability distributions to represent the equivalence classes. The affordances are arranged based on goal-orientedness, since achieving a task usually requires actions performed in a sequence. Additionally, motivation values are generated using the arranged affordances and a motivation value propagation algorithm. A robot selects a goal-oriented as well as a situation-adequate action based on the motivation values. To validate our proposed methods, we present experimental results of an entertainment robot called AIBO, handling three tasks.

A Skilligent robot must be able to learn skills autonomously to accomplish a task. ldquoskilligencerdquo is the capacity of the robot to control behaviors reasonably, based on the skills acquired during run-time. Behavior control based on Bayesian networks is used to control reasonable behaviors. To accomplish this, subgoals are first discovered by clustering similar features of state transition tuples, which are composed of current states, actions, and next states. Here, features used in clustering are produced using changes of the states in the state transition tuples. Parameters of Bayesian networks and utility functions are learned separately using state transition tuples belonging to each subgoal. To select the best action while executing a task, the expected utility of each subgoal is calculated by the expected utility function and the robot chooses the action that maximizes expected utility calculated by the maximum expected utility (MEU) function. The MEU function is based on the conditional probabilistic distributions of Bayesian networks and utility functions. We also propose a method for reconstructing learned networks and increasing subgoals by incremental learning. To show the validities of our proposed methods, a task using dribbling-box-into-a-goal (DBIG) and obstacle-avoidance-while-dribbling-box (OAWDB) skills is simulated and experimented.

In the task of robot localization, Bayes filters use two processes: the prediction step and the measurement-update step. Briefly, the state transition model is responsible for prediction, and the sensor model is responsible for measurement updates. This paper presents a new approach to the sensor model, called the predictive sensor model, which utilizes a prediction mechanism to improve the efficiency of measurement updates in Bayes filters. By adding sensorial anticipation, we extend the original Bayes filter to an anticipatory Bayes filter. We also propose an entropy-based place-segmentation method for automatic segmentation of sequentially collected vision-sensor data. Our place segmentation technique is most useful for node clustering in the process of constructing topological maps. Our work was verified by experiments using observed data.

This work considers robot localization with an action-associated sparse appearance-based map, under conditions with dynamic change in the environment. In this case, two significant problems must be solved for robust localization. The first involves variations in the environment caused by dynamic objects and changes in illumination, and the second arises from the nature of sparse appearance-based map. That is, a robot must be able to recognize observations taken at slightly different positions and angles within a certain region as identical. In this paper, we address a possible solution to these problems on the basis of a probabilistic model called the Bayes filter. Here, we propose an observation model based LeTO2function and an action-associated sparse appearance-based map to be used for prediction, update, and final localization steps. In addition, multiple visual features are used to increase the reliability of the observation model. We performed experiments to demonstrate the validity of the proposed approach under various conditions with regard to dynamic objects, illumination, and viewpoint. The results clearly demonstrated the value of our approach.

We propose a semantic representation and Bayesian model for robot localization using spatial relations among objects that can be created by a single consumer-grade camera and odometry. We first suggest a semantic representation to be shared by human and robot. This representation consists of perceived objects and their spatial relationships, and a qualitatively defined odometry-based metric distance. We refer to this as a topological-semantic distance map. To support our semantic representation, we develop a Bayesian model for localization that enables the location of a robot to be estimated sufficiently well to navigate in an indoor environment. Extensive localization experiments in an indoor environment show that our Bayesian localization technique using a topological-semantic distance map is valid in the sense that localization accuracy improves whenever objects and their spatial relationships are detected and instantiated.

Prediction is a very important element of human intelligence and plays a major role in human behavior, perception, and learning. This paper presents the development of a mathematical model of the prediction mechanism in the context of a Bayes filter, which is the predominant schema used for integrating temporal data in the field of robot mapping and localization problems. We propose a generalized anticipatory Bayes filter that uses revised sensor values obtained from the prediction process at the measurement-update step to enhance the performance of the sensor model. The development of a generalized anticipatory Bayes filter is not only an extension of the original Bayes filter, but also a mathematical model of the human prediction mechanism of sensory processing. This work was verified by experiments using observed data.

This study addressed the problem of active localization, which requires massive computation. To solve the problem, we developed abstracted measurements that consist of qualitative metrics estimated by a single camera. These are contextual representations consisting of perceived landmarks and their spatial relations, and they can be shared by humans and robots. Next, we enhanced the Markov localization method to support contextual representations with which a robot's location can be sufficiently estimated. In contrast to passive methodologies, our approach actively uses the greedy technique to select a robot's action and improve localization results. The experiment was carried out in an indoor environment, and results indicate that the proposed active-semantic localization yields more efficient localization.

There are many difficulties to remotely control a robot over Internet such as transmission time delay, limitation of bandwidth, and packet loss. In this paper, we propose a novel remote control architecture and a communication architecture for Internet-based robots, which is insensitive to the transmission time delay and packet loss. The proposed architecture guarantees enhanced remote control by providing the perfect copy of the status of remote robot at real-time. The communication protocol has a flexible packet format and various packet sizes at runtime for robust and reliable control mechanism. The architecture together with the protocol provides improved control performance. Experiments shows that the effectiveness and applicability of the internet-based robot with the proposed architecture.

A service robot has to identify objects for given tasks by using the camera mounted on its body. In general, previously proposed object-recognition systems do not show enough performance to apply them to service robots. Since, the object-recognition system for service robots must be robust for environmental conditions and fast enough to support its tasks. Changes of Illumination, viewpoint and scale level are examples for those changes. In addition, the system must be robust for partially occluded cases. To satisfy those requirements, we propose the system with novel modeling and matching methods. Moreover, our system can show better performance by adjusting parameter based on environmental conditions. Through lots of tests for different environmental changes, our system is shown to be partially invariant to illumination and viewpoints. Moreover, our system verifies its robustness for various scale levels and partially occluded cases. Consequently, service robots can identify objects efficiently in environmental changes by adopting the recognition system we proposed.

We present a novel filtering method for building a grid map using sonar sensors. Recognizing that conflicting information from sonar sensor readings is the main reason of poor map quality, we established a proposition that put the conflicting information in order. Based on the proposition, we effectively filtered sonar readings that cause conflict. In addition, some information of filtered readings were not discarded but reprocessed to enhance the quality of the map. We call this filtering and reprocessing operation the conflict evaluation method. We evaluated the proposed method by experiments under various conditions.

Semantic knowledge especially for robots is often used as a form of knowledge representation which includes ontological knowledge to support Description Logics and rule language. However, it is difficult to design rules which provide some goal-associated Context, Object and Space (COS)-action complexes to be used to infer alternative attempts for success of a goal, when current attempt is not satisfactory. To cope with such a difficulty, we propose a novel weighted Action-coupled Semantic Network (wASN) which is used to recursively update and infer goal-associated COS-action complexes. To be specific, as in the semantic networks, wASN is a graph consisting of nodes and edges, respectively, to represent concepts-such as perception, object, space and context, and to represent semantic relation between the concepts. And, those concepts are designed to be all coupled with their associated actions. On the other hand, as in neural networks, our proposed wASN is designed to have weighted values of the connections among semantic nodes. And those weighted values are summed at each node, and are used to adopt a node by “winner-takes-all” activation strategy. It is shown that wASN is useful to recursively get COS-action complexes for a specific task to find objects which are partially occluded or unobservable.

Task planning by hierarchical abstraction of world elements and behaviors(PhaWeB) is proposed for mobile robots task planning, where abstraction of world elements is used to reduce search space and abstraction of behaviors is used to efficiently find more detailed behaviors. To integrate abstraction of world elements with abstraction of behaviors, we construct hierarchical domain description based on joint of action hierarchy (JAH). In JAH, some actions in a level of action hierarchy are included in the adjacent level of action hierarchy. To show the validities of our proposed PhaWeB, experimental results are illustrated and will be compared with world elements abstraction based planning approach as well as behaviors abstraction based planning approach.

In order to provide improved convenience for a surgeon in spinal fusion surgery, a robot system should i) closely engage in surgeon's operation using an end effector, and ii) protect the surgeon from being exposed to harmful radiation due to repeated shootings of fluoroscope. This paper proposes a bilateral teleoperation system for spinal fusion, BiTESS-II, to accomplish the goals. We developed an end effector that can substitute the surgeon's manual operation and a novel closed-loop type slave robot that can exert strong reaction force to complete gimleting and screwing tasks. Master devices are used to control the position and orientation of the slave robot and to generate haptic information identical to that of the slave side. A novel force reflection method without force sensors allowed to design the end effector simple and light. BiTESS-II is among the first human guided teleoperation system for spinal fusion with an adequate end effector. The performance of the BiTESS-II was verified by experiments.

Abstract — Reinforcement learning (RL) has been used as a learning mechanism for a mobile robot to learn state-action relations without a priori knowledge of working environment. However, most RL methods usually suffer from slow conver-gence to learn optimum state-action sequence. In this paper, it is intended to improve a learning speed by compounding an existing Q-learning method with a shortest path finding algorithm. To integrate the shortest path algorithm with Q-learning method, a stochastic state-transition model is used to store a previous observed state, a previous action and a current state. Whenever a robot reaches a goal, a Stochastic Shortest Path(SSP) will be found from the stochastic state-transition model. State-action pairs on the SSP will be counted as more significant in the action selection. Using this learning method, the learning speed will be boosted when compared with classical RL methods. To show the validity of our proposed learning technology, several simulations and experimental results will be illustrated. I.

An ontology-based multi-layered robot knowledge framework (OMRKF) is proposed to implement robot intelligence to be useful in a robot environment. OMRKF consists of four classes of knowledge (KClass), axioms and two types of rules. Four KClasses including perception, model, activity and context class are organized in a hierarchy of three knowledge levels (KLevel) and three ontology layers (OLayer). The axioms specify the semantics of concepts and relational constraints between ontological elements in each OLayer. One type of rule is designed for relationships between concepts in the same KClasses but in different KLevels. These rules will be used in a way of unidirectional reasoning. And, the other types of rules are also designed for association between concepts in different KLevels and different KClasses to be used in a way of bi-directional reasoning. These features will let OMRKF enable a robot to integrate robot knowledge from levels of sensor data and primitive behaviors to levels of symbolic data and contextual information regardless of class of knowledge. To show the validities of our proposed OMRKF, several experimental results will be illustrated, where some queries can be possibly answered by using uni-directional rules as well as bi-directional rules even with partial and uncertain information.

This paper proposes restriction space projection (RSP) method to generate accurate direction of force reflection when a bilateral teleoperation system (BTS) is kinematically dissimilar. Through examples, it is shown that the previous force reflection methods are not applicable to the kinematically dissimilar BTSs. Two kinds of RSP methods using a novel concept, instantaneous restriction space (IRS), are implemented for impedance and admittance type of BTSs. Especially, new developed obstacle avoidance algorithm using the redundancy of the slave manipulator makes the RSP method applicable to all kinds of kinematically dissimilar BTSs. Experiments verify that the RSP method is powerful to describe the restriction space at the slave side without force sensors

Two 3-DOF parallel mechanisms, a translational mechanism and a spherical mechanism, are analyzed. Each of two mechanisms commonly possesses three UPS serial sub-chains, and an additional passive 3-DOF PPP-type serial sub-chain and a passive 3-DOF RRR-type serial sub-chain are incorporated for the translational device and the rotational device, respectively. Such additional passive serial sub-chains act as constraints to restrict the output motion of the mechanism either in a 3-DOF translational space or in a 3-DOF spherical space. The position solutions of the proposed mechanisms and their first-order kinematic models are derived. Then their workspaces and kinematic characteristics are examined via the kinematic isotropic index

A robot programming language (PLEASE) is to be proposed together with an ethology-based action-selection (EASE) architecture, where following features are supported; Hierarchical tree-structured behavior organization, layertailored arbiter programming, abstract expression of real world situation by logically connecting outputs of perception filters, concurrent behaviors.

This paper deals with the trajectory planning of a soccer ball driven by a kicking robot. The trajectory of a ball is function of its initial velocity, launch angle, and some aerodynamic effect. Specially, the initial velocity is created by an instant impulse given to the ball. The external impulse exerted to a ball by a kicking robot is function of robot geometry and dynamic parameters. Initially, we analyze the external impulses for several kicking postures and simulate trajectories for several different kicking conditions. Based on this observation, a trajectory-planning algorithm of a ball, in which the initial velocity and the launch angle of the ball are calculated for a desired trajectory of the ball, is proposed, and then an external impulse required for that motion is calculated and applied to the kicking simulation. The aerodynamic effect such as drag force and lift force is also incorporated into the dynamic simulation of the soccer ball. A kicking robot has been developed for experimental verification of the proposed analytical methodology. It was shown that experimental result is in good agreement with simulation result within 10% error bound.

The two most important abilities for a robot to survive in a given environment are selecting and learning the most appropriate actions in a given situation. Historically, they have also been the biggest problems in robotics. To help solve this problem, we propose a two-layered action selection mechanism (ASM) which designates an action pattern layer and a reactive behavior plan layer. In the reactive behavior plan layer, a task is selected by comparing behavior motivation values that, in an animal, correspond to external stimuli as well as internal states due to hormones. After a task is selected, its corresponding reactive behavior plan is executed as a set of sequential dynamic behavior motivations (DBMs), each of which is associated with an action pattern. In the action pattern layer, each action pattern can be functionally decomposed into primitive motor actions. Shortest path-based Q-learning (SPQL) is incorporated into both the reactive behavior plan and action pattern layers. In the reactive behavior plan layer, relationships between perceptions and action patterns are learned to satisfy a given motivation, as well as the relative priorities of these relationships. In the action pattern layer, the relations between sensory states and primitive motor actions can be learned. To establish the validity of our proposed ASM, experiments with our real designed robot was illustrated together with simulations.

Some of manufacturing tasks, such as sawing, often require continuous impulsive motion. In the case of sawing, such impulsive motions can be observed between the teeth of the saw and the object. The amount of the external impulse exerted on the object has been treated as an important control parameter. At the same time, the internal impulses experienced at the joints should be taken into account to avoid serious damage or injury at the joints of robot. The purpose of this paper is to improve the efficacy of sawing by using dual arms. For this, we suggest an external impulse model, and introduce a new concept of effective mass that accounts for the hardness of the object to corroborate the effectiveness of the proposed chip model employed in the derivation of the external impulse model. Closed-form internal impulse models are also proposed for both single and dual arms. Based on these models, the paper proposes a new measure for internal impulse. A normalized impulse ellipsoid reflecting the velocity direction is employed to visualize the impact geometry. Finally, we identify the optimal sawing region, wherein there is a maximum amount of external impulse and a minimal amount of the internal impulse. Simulation and experimentation demonstrate that the dual arm exhibits a better sawing performance than a single arm, in terms of external and internal impulses.

It is known that parallel type mechanisms have many singularities than serial type mechanisms. In haptic application, these singularities deteriorate the force reflecting performance. Moreover, different from general manipulators, haptic systems can't avoid the singular point, because they are operated by the user's random motion command. Although many singularity-free load distribution algorithms for kinematically redundant manipulators have been proposed, singularity-free load distribution algorithms for parallel haptic application have not been deeply discussed. In this work, various singularity-free load distribution algorithms that are appropriate to parallel haptic system are discussed. A new 6 DOF parallel mechanism equipped with four sub-chains is employed as a test device. To cope with singularity problem, four task-priority algorithms and redundant actuation algorithm are discussed and compared through simulation, and also experiments have been performed to show the effectiveness of those algorithms at the presence of singularity.

A novel action-selection-mechanism is proposed to deal with sequential behaviors, where associations between some of stimulus and behaviors would be learned by a shortest-path-finding-based reinforcement learning technique. To be specific, we define behavioral motivation as a primitive node for action selection, and then sequentially construct a network with behavioral motivations. The vertical path of the network represents a behavioral sequence. Here, such a tree for our proposed ASM can be newly generated and/or updated, whenever a new sequential behaviors is learned. To show the validity of our proposed ASM, some experimental results on a "pushing-box-into-a-goal (PBIG) task" of a mobile robot is illustrated.

A haptic rendering system is proposed for a puncture task of a virtual vertebra model. To build a mesh model from medical images, Delaunay triangulation is applied and physical models are based on elasticity theory. Also, a redundant actuated 6 DOF parallel type haptic device is designed to display large force and to resolve the singularity problem of parallel type mechanisms. Haptic feeling of puncture task and the performance of the proposed haptic device are tested by two puncture task experiments.

Parallel-type mechanisms provide many advantages over serial-type mechanisms. However, in case of using parallel structures as a haptic device, it is sometimes hard to ensure the performance of the force reflection due to many singular points existing in workspace. In this paper, we propose a redundantly actuated parallel 3-DOF haptic device, which is singularity-free in the workspace and also has an improved force reflection capability. Using two sets of actuators with different size and different force resolution, we propose several useful load distribution algorithms considering force resolution and torque limit. We confirm the performance of the force reflection capability throughout simulation and experimentation.

Some of manufacturing tasks such as sawing task often require continuous impulsive motion. In case of sawing task, such impulsive motions can be observed between the teeth of the saw and the object being sawn. The amount of the external impulse exerted on the object is treated as an important control parameter in the sawing task. The purpose of this work is to show the efficacy of sawing task by using a dual-arm. For this, a new concept of effective mass is introduced to model the impact phenomenon in the sawing task. Using this concept, the external impulse model of sawing tasks is suggested. The optimal sawing region is identified in which the amount of external impulse is maximized. It is demonstrated through simulation and experiment that the dual-arm exhibits better sawing performance over a single arm in aspects of external impulse.

A behavior-based control and learning architecture is proposed, where reinforcement learning is applied to learn proper associations between stimulus and response by using two types of memory called as short Term Memory and Long Term Memory. In particular, to cope with delayed-reward problem, a knowledge-propagation (KP) method is proposed, where well-designed or well-trained S-R(stimulus-response) associations for low-level sensors are utilized to learn new S-R associations for high-level sensors, in case that those S-R associations require same objective such as obstacle avoidance. To show the validity of our proposed KP method, comparative experiments are performed for the cases that:(1) only a delayed reward is used, (2) some of S-R pairs are preprogrammed, (3) immediate reward is possible, and (4) our KP method is applied.

Reinforcement learning (RL) has been widely used as a learning mechanism of an artificial life system. However, RL usually suffers from slow convergence to the optimum state-action sequence or a sequence of stimulus-response (SR) behaviors, and may not correctly work in non-Markov processes. In this paper, first, to cope with slow-convergence problem, if some state-action pairs are considered as disturbance for optimum sequence, then they are to be eliminated in long-term memory (LTM), where such disturbances are found by a shortest path-finding algorithm. This process is shown to let the system get an enhanced learning speed. Second, to partly solve non-Markov problem, if a stimulus is frequently met in a searching-process, then the stimulus is classified as a sequential percept for a non-Markov hidden state. And thus, a correct behavior for a non-Markov hidden state can be learned as in a Markov environment. To show the validity of our proposed learning technologies, several simulation results are illustrated.

In this paper, a modified two-step compliance control method for robot hands is proposed: resolved interfinger decoupling solver (RIFDS) and resolved interjoint decoupling solver (RIJDS). For this, we first investigate how many fingers are necessary to successfully implement stiffness characteristics in the operational space. RIFDS is then proposed to decompose the desired compliance characteristic specified in the operational space into the compliance characteristic in the fingertip space without interfinger coupling, and RIJDS is also proposed to decompose the compliance characteristic in the fingertip space without interjoint coupling. It is found in the process of RIFDS that some nondiagonal stiffness elements specified in the operational space cannot be planned arbitrarily, due to grasping geometry. Similar to independent finger control, RIJDS aims at independent joint control. This scheme facilitates the joint servo control. To show the effectiveness of the proposed compliance control method, some experimental results are illustrated for a compliant task by using two- and three-fingered hands, which consist of five-bar finger mechanisms. It is concluded that grasping geometry and finger structure are crucial to successfully performing multifingered hands operations.

Flexure hinge has been commonly used as a substitute for mechanical joints in the design of micropositioning mechanisms. However, inaccurate modeling of flexure hinges deteriorates the positioning accuracy. In this paper, a planar 3-DOF parallel-type micropositioning mechanism is designed with the intention of accurate flexure hinge modeling. For this, a preliminary kinematic analysis that includes inverse kinematics, internal kinematics, and analytic stiffness modeling referenced to the task coordinate is presented. First, the revolute type of 1-DOF flexure hinge is considered. The simulation result based on the finite element method, however, is not coincident to the analytic result. This is due to the minor axial elongation along the link direction that keeps the mechanism from precise positioning. To cope with this problem, a 2-DOF flexure hinge model that includes this additional motion degree as a prismatic joint is employed in part, and additional actuators are added to compensate for the motion of this new model. On the basis of this model, the positional accuracy is ensured. The effectiveness of this accurate model is shown through both simulation and experimentation. This paper emphasizes that the precise modeling of a flexure hinge is significant to guarantee the positional accuracy of parallel micromechanisms using flexure hinge.

It is known that parallel-type mechanisms have more singularities than serial-type mechanisms. In haptic application, these singularities deteriorate the system performance generating reflecting force. Moreover, different from general manipulators, haptic systems can't avoid the singular point, because they are operated by user random command. Although many singularity-free algorithms for serial mechanisms have been proposed and studied, singularity-free algorithms for parallel haptic application have not been deeply discussed. In this paper various singularity-free algorithms, which are appropriate to parallel haptic system, will be discussed. Furthermore, to overcome singularity problems, a new design including redundant actuation is proposed. The proposed mechanism has fairly large workspace and is relatively light by employing non-floating actuators.

Some of manufacturing tasks such as sawing task often requires continuous impulsive motion. In case of sawing task, such impulsive motions can be observed between the teeth of the saw and the object. The amount of the external impulse exerted on the object has been treated as an important control paramete. Also, the internal impulses experienced at the joints should be simultaneously taken into account to avoid serious damage or injury at the joints. The purpose of this work is to improve the efficacy of sawing task by using dual arm. For this, an external impulse model for sawing task is suggested. Also, the closed-form internal impulse models are proposed for both a single and dual arms. Based on these models, a new measure for internal impulse is proposed. A normalized impulse ellipsoid reflecting the velocity direction is employed to visualize the impact geometry. Finally, the optimal sawing region is identified in which the amount of external impulse is maximized and the amount of the internal impulse is minimized. It is demonstrated through simulation that the dual arm exhibits better sawing performance over a single arm in aspects of external and internal impulses.

A planar 3 DOF parallel-type micro-positioning mechanism is designed with the intention of accurate flexure hinge modeling. For this, a preliminary kinematic analysis that includes inverse kinematics, internal kinematics, and analytic stiffness modeling referenced to the task coordinate is presented. First, the revolute type of a 1 DOF flexure hinge is considered. The simulation result based on FEM, however, is not coincident to the analytic result. This is due to the minor axial elongation along the link direction that keeps the mechanism from precise positioning. To cope with this problem, a 2 DOF flexure hinge model that includes this additional motion degree as a prismatic joint is employed in part. On the basis of this model, the positional accuracy is ensured. The effectiveness of this accurate model is shown through both simulation and experimentation This work emphasizes that the precise modeling of a flexure hinge is significant to guarantee the positional accuracy of parallel micro-mechanisms using flexure hinge.

This paper presents a design method of transparency optimized feedback controller with robustness for a parallel 6-DOF haptic system. A disturbance observer is employed to eliminate coupling effects that exist in a mechanically coupled structure. A performance index for the transparency is defined by admittance matching, and by solving the H/sub 2/ optimal problem. The optimal solution that minimizes the performance index is obtained.

Based on the analysis of the stiffness relation between the operational space and the fingertip space of multifingered hands, this paper provides a guideline of task-based compliance planning for multifingered hands. 2D and 3D examples are illustrated to show the characteristics of the task-based stiffness matrix. It is shown that some of coupling stiffness elements cannot be planned arbitrarily due to grasping geometry. Through the analytical results, it is concluded that the operational stiffness matrix should be carefully specified by considering the location of compliance center and the grasp geometry of multifingered hands for successful grasping and manipulation tasks.

We provide a guideline for specifying compliance characteristics in the operational space of multifingered hands. From the analysis of the stiffness relation between the operational space and the fingertip space of multifingered hands, it is shown that some of the coupling stiffness elements cannot be planned arbitrarily. An independent finger-based compliance control method to achieve the given compliance characteristics is presented. Through the analytical results, it is concluded that the operational stiffness matrix should be carefully specified by considering the grasp geometry of multifingered hands for successful grasping and manipulation tasks, and the three-dimensional compliance control can also be achieved by a five-fingered human-like hand.

In this paper, a new 6-DOF parallel haptic master device is proposed. Many existing haptic devices require large power due to having floating actuator and also have small workspaces. The proposed new mechanism is relatively light, as it employs nonfloating actuators and has large workspace. Kinematic analysis and kinematic optimal design problem is performed for this mechanism. Dexterous workspace, global isotropic index, and global maximum force transmission ratio are considered as kinematic design indices. To deal with such multicriteria optimization problem, composite design index is employed. Actuator sizing for this mechanism is also carried out.

In this paper, we propose and develop a spatial 3-DOF tele-micromanipulator for precise position control of micro-objects. Typical feature of this device is a one-module flexure hinge that consists of a revolute joint and a spherical joint. Based on preliminary kinematic analysis and stiffness modeling of the system, optimal design and actuator sizing for the device are performed. Furthermore, FEM analysis and resonant frequency analysis are executed to validate the results of analytic design process. The designed device was successfully implemented to tele-micromanipulation system.

For a multi-fingered hand to grasp an object, there are numerous ways to grasp it stably, and thus an optimal grasp planning is necessary to find the optimal grasp point for achieving the objective of the given task. First, we define several grasp indices to evaluate the quality of each feasible grasp. Since the physical meanings of the defined grasp induces are different from each other, it is not easy to combine those indices to identify the optimal grasping. In this paper, we propose a new generalized grasping performance index to represent all of the grasp indices as one measure based on a non-dimensional technique. Through simulations, we show that the proposed optimal grasp planning is resemblant to the physical sense of human grasping.

This paper suggests an inverse optimal PID control design method to track trajectories in Lagrangian systems. The inverse optimal PID controller exists if and only if the Lagrangian system is extended disturbance input-to-state stable. First, we find the Lyapunov function and the control law that satisfy the extended disturbance input-to-state stability by using the characteristics of the Lagrangian system. The control law has a MID control form and satisfies the Hamilton-Jacobi-Isaacs equation. Hence, the H/sub /spl infin// inverse optimality of the closed-loop system dynamics is acquired through the PID controller if the conditions for the control law are satisfied. Also, simple coarse/fine performance tuning laws are suggested based on a performance limitation analysis of the inverse optimal PID controller. Selection conditions for gains are proposed as functions of the tuning variable. Experimental results for a typical Lagrangian system show that our analysis provides performance and H/sub /spl infin// optimality.

Underwater robotic vehicles (URV) have been employed for mobility, and robot manipulators attached to the underwater vehicle play the role of manipulation. URV motions are determined by inherent dynamic couplings between active and passive joints. Furthermore, the control problem becomes complex since there should be considered many hydrodynamic terms as well as intrinsic model uncertainties. To cope with these difficulties, we propose a disturbance observer-based robust control algorithm for underwater manipulators with passive joints. The proposed control algorithm is able to treat an underactuated system as a pseudo-active system in which passive joints are eliminated. Also, to realize a robust control method, a nonlinear feedback disturbance observer is applied to each active joint. A four-jointed underwater robotic system with one passive joint is considered as an illustrative example. Through simulation, it is shown that the proposed control algorithm has good position tracking performances even in the presence of several external disturbances and model uncertainties.

For an object grasped by a robot hand to work in the compliance control domain, we first analyze a necessary condition for successful stiffness modulation in the operational space. Next, we propose a compliance control method for robot hands which consist of two steps: the RIFDS (resolved inter-finger decoupling solver) decomposes the desired compliance characteristic specified in the operational space into the compliance characteristic in the fingertip space without inter-finger coupling; and the RIJDS (resolved inter-joint decoupling solver) decomposes the compliance characteristic in the fingertip space into the compliance characteristic in the joint space without inter-joint coupling. According to the analysis results, the finger structure should be biomimetic in the sense that either kinematic redundancy or force redundancy are required to implement the proposed compliance control scheme. Five-bar fingered robot hands are treated as illustrative examples to implement the proposed compliance control method. To show the effectiveness of the proposed compliance control method, simulations are performed for two-fingered and three-fingered robot hands.

This paper proposes a visual serving method based on QR decomposition and disturbance observer. The QR decomposition factorizes the image feature Jacobian into a unitary matrix and an upper triangular matrix. It is shown that several performance indices such as measurement sensitivity of visual features, sensitivity of the control to noise and controllability can be improved for any general image feature Jacobian by OR decomposition and disturbance observer. To show the validities of the proposed approach, visual servoing with stereo cameras is experimented on Samsung FARAMAN 6-axis industrial manipulator.

A transparency-optimized feedback controller with robustness is designed for a multi-axis haptic device, where a disturbance observer is employed to decouple the coupling effect existing in the multi-axis haptic device. A performance index for the transparency is defined from the viewpoint of admittance matching, and the optimal solution minimizing the performance index is obtained by solving an H/sub 2/ optimal problem. To implement a stabilizing haptic controller robust to parametric uncertainties of the haptic device, a robust stable condition using an H/sub /spl infin// norm derived from small gain theorem is proposed. The effectiveness of the proposed haptic control scheme has been experimentally verified for a virtual wall characterized by stiffness and damping properties.

This paper deals with stiffness analysis for effective peg-in/out-hole tasks using multifingered robot hand without inter-finger coupling. We first observe the fact that some coupling stiffness elements cannot be planned arbitrary. Then, we analyze the conditions of the specified stiffness matrix in the operational space to successfully and more effectively achieve the given peg-in/out-hole tasks. It is concluded that the location of compliance center on the peg and the coupling stiffness element between the translational and the rotational direction play important roles for successful peg-in/out-hole tasks. Simulation results are included to verify the feasibility of the analytic results.

It is well-known that bio-systems do not calculate inverse dynamics for trajectory planning, but they move by proper modulation of the system's impedances. Inspired by bio-systems, a biomimetic trajectory planning method is introduced. This scheme is based on employment of redundant actuation which prevails in bio-systems. We explain that in the generation of the biomimetic trajectory, intelligent structure of bio-systems plays an important role to completely modulate the dynamic behaviour of a system, such as natural frequencies and damping ratios. A five-bar mechanism equipped with two redundant actuators is employed to demonstrate the feasibility of the proposed biomimetic trajectory experimentally.

Springs have been employed in a wide range of mechanical systems. This work deals with the design of multi degree-of-freedom spring mechanism. The adaptable spring is desired for enhancing performances of various mechanical systems employing springs. We demonstrate that such adaptable springs can be realized by adopting anthropomorphic musculoskeletal structures of the human upper-extremity, which possesses highly nonlinear kinematic-coupling among redundant muscles existing in its structures. We propose multi-degree-of-freedom spring mechanisms resembling the musculoskeletal structure of the human upper-extremity, and verify the applicability of these mechanisms through simulation.

A new parallel gripping mechanism is proposed in this work. This device has a parallelogramic platform which can be flexibly folded. Therefore, this mechanism not only can be used to grasp an object having irregular shape or large volume, but also can be utilized as micro-positioning device after grasping the object. Forward position analysis and platform kinematics are investigated to deal with motion tracking and force control. Pneumatic rotator is employed for actuation and a compact sized, 4/3 way proportional pressure valve is also developed to deal with feedback-based dynamic control. The pressure valve also allows indirect force control by measuring the offset pressure occurring due to contact between the grasped object and the parallel platform. In experimental work, performances of the motion tracking and indirect force control are shown satisfactory.

A compliance control strategy for robot manipulators is proposed by employing a self-adjusting stiffness function. To be specific, each entry of the diagonal stiffness matrix corresponding to task coordinate in Cartesian space is adaptively adjusted during contact along the corresponding axis based on the contact force with its environment. It can also be used for both unconstrained and constrained motions without any switching mechanism which often causes undesirable instability and/or vibrational motion of the end-effector. The experimental results show the effectiveness of the proposed method by employing a two-link direct drive manipulator interacting with an unknown environment.

To improve the visual servoing performance, several strategies have been proposed by using redundant feature points, by using a point with different height, and by a weighted selection of image features. The performance of these visual servoing methods depends on the configuration between the camera and objects. And redundant feature points require much computational efforts. The paper proposes a visual servoing method based on a disturbance observer which compensates the effect of the off-diagonal component of the image feature Jacobian to be null. The performance indices such as measurement sensitivity of visual features, sensitivity of the control to noise, and controllability are improved when an image feature Jacobian is given as a block diagonal matrix. The proposed approach is applied to the PUMA560 and Samsung FARAMAN 6-axis industrial robot manipulators successfully.

We discuss an inherent problem occurring in the former operational-based compliance control scheme. In order to obviate this problem, an independent joint-based compliance control scheme is proposed. This control method requires that for the adjustment of the operational compliance matrix, a manipulator should have at least the same number of joints as the number of independent components of the desired compliance matrix specified in the operational space. Types of robot structures which are appropriate to independent joint-based compliance scheme are addressed. A five-bar finger with redundant actuators is treated as an illustrative example. Our experimental results show that the proposed compliance modulation method gave satisfactory performance. We also consider an additional stiffness effect due to the antagonistic conflict between the controlled-operational force and the equilibrium joint torque vector. It is also experimentally shown that this additional term notably improved the force tracking performance of the given mechanism.

A visual servoing method has been proposed based on a disturbance observer to eliminate the effect of the off-diagonal component of the image feature Jacobian, since performance indices such as measurement sensitivity of visual features, sensitivity of the control to noise, and controllability could be improved when an image feature Jacobian was given as a block diagonal matrix. In this paper, experimental results of disturbance observer-based visual servoing are discussed, where a Samsung FARAMAN-ASI 6-axis industrial robot manipulator is employed. Also, the feature saturator is proposed to stabilize the disturbance observer loop by saturating the differential changes of the image features.

In this paper design of a robust impedance control is proposed for kinematically redundant manipulators. To achieve this goal, we propose a new disturbance observer scheme which can handle the nonlinear dynamics of a manipulator. An extended task space formulation to describe the dynamics of redundant manipulator is employed. Using this extended task space formulation, a robust impedance control method is proposed based on the momentum feedback disturbance observer (MFDOB). The performance of the proposed extended impedance controller is verified through experiments with a planar three-link direct-drive manipulator.

Prototype-based methods are commonly used in cluster analysis and the results may be highly dependent on the prototype used. We propose a two-level fuzzy clustering method that involves adaptively expanding and merging convex polytopes, where the convex polytopes are considered as a "flexible" prototype. Therefore, the dependency on the use of a specified prototype can be eliminated. Also, the proposed method makes it possible to effectively represent an arbitrarily distributed data set without a priori knowledge of the number of clusters in the data set. In the first level of our proposed method, each cluster is represented by a convex polytope which is described by its set of vertices. Specifically, nonlinear membership functions are utilized to determine whether an input pattern creates a new cluster or whether an existing cluster should be modified. In the second level, the expandable clusters that are selected by an intercluster distance measure are merged to improve clustering efficiency and to reduce the order dependency of the incoming input patterns. Several experimental results are given to show the validity of our method.

Deals with a sawing task performed by two cooperating arms. The two-arm system under our hand consists of a four degree-of-freedom (DOF) SCARA robot and a five DOF PT200V robot. When the two arms are rigidly grasping a saw, the mobility of the system is three, which is not enough for sawing tasks. Therefore, we deliberately insert a passive joint at the end of the SCARA robot to increase the mobility up to four. A hybrid control method to regulate the force and velocity by the two arms is proposed in this work. The proposed scheme has three typical features; first, the two arms are treated as one arm in a kinematic viewpoint. Secondly, our approach is different from other acceleration-based approaches, in the sense that our hybrid control method is based on a Jacobian and an internal kinematics for a single closed-kinematic chain of the two arms to reflect the nature of the position-controlled industrial manipulator. Thirdly, the proposed scheme is not only able to operate the system even if a passive joint exists, but also is able to utilize the internal loads for useful applications such as pitch motion control. We experimentally show that the performance of the velocity and force response are satisfactory, and that one additional passive joint not only prevents the system from unwanted roll motion in the sawing task, but also allows an unwanted pitch motion to be notably reduced by an internal load control. To show the effectiveness of the proposed algorithms, we perform experimentation under several, different conditions for saw, such as three saw blades, two sawing speeds, and two vertical forces.

A five-bar finger mechanism driven by redundant actuators is given as an illustrative example. It is shown that judicial choice of the location of one redundant actuator greatly enhances the load handling capacity of the system, when compared to those of minimum actuation and more than two redundant actuators. Also, methodologies for stiffness and motion frequency modulations via redundant actuation are investigated. Internal load distribution associated with the stiffness and motion frequency modulations is further discussed. Specifically, the motion frequency of the system is modulated by employing inertial and spring-like impedance properties created by internal loading. The motion frequency as well as the amplitude of oscillation can be actively adjusted during the motion, and the equilibrium position about which the vibration occurs can also be arbitrarily changed during the motion. Furthermore, using the stiffness modulation capability, a point-to-point motion can be accomplished by a progressive movement of equilibrium posture, which is called a virtual trajectory. To show the effectiveness of the proposed algorithms, several simulation results are illustrated.

In this paper, a force estimation method is proposed for force control without force sensor. A disturbance observer is applied to each joint of an n degrees of freedom manipulator to obtain a simple equivalent robot dynamics being represented as an n independent double integrator system. To estimate the output of disturbance observer due to internal torque, the disturbance observer output estimator (DOOE) is designed, where uncertain parameters of the robot manipulator are adjusted by the gradient method to minimize the performance index which is defined as the quadratic form of the error signal between the output of disturbance observer and that of DOOE. When the external force is exerted, the external force is estimated by the difference between the output of disturbance observer and DOOE. A force controller is designed for force feedback control employing the estimated force signal. Several numerical examples and experimental results are illustrated for a 2-axis direct drive robot manipulator.

In order to develop a human hand mechanism, a 5-bar finger with redundant actuation is suggested. Optimal sets of actuator locations and link lengths for the cases of minimum actuator, one, two and three redundant actuators are obtained by employing composite design index which simultaneously consider several performance indices such as workspace, isotropic index, and force transmission ratio. Finally, several finger-configurations optimized for special performance indices are illustrated.

An approach to resolve the kinematic redundancy and to control the motion/force of redundant manipulators is presented. By defining a proper metric in joint space, minimal parametrization of motion and force controlled subspaces as well as the null motion component is realized. With this formulation, control of both motion/force and internal motion of redundant manipulator can be achieved via a new hybrid impedance control method with inertial decoupling of each space. Some numerical examples are given to demonstrate the performance of the proposed control method.

Typical biomechanical systems such as human body and mammals possess abundant muscles which are more than required for motion generation of such systems. The purpose of this work is to verify a biological phenomenon, the so called frequency modulation, in a mathematical manner. The frequency modulation represents simultaneous control of force and kinematic redundancies. The phenomenon of frequency modulation is explained through a human-like anthropomorphic robot. A load distribution method for frequency modulation via redundant actuation is also introduced. To show the effectiveness of the proposed algorithms, several simulation results are illustrated.

An impedance control approach based on extended task space formulation is addressed to control kinematically redundant manipulators. Defining a weighted inner product in joint space, a minimal parameterization of the null space can be achieved. Based on this formulation, we propose a control law called inertially decoupled impedance controller by expanding the conventional impedance control approach to control the motion of the end-effector as well as the internal motion. Experimental results are given to demonstrate the performance of the proposed control methods.

In this work, a parallel-type gripping and micro-positioning mechanism is proposed. This device has a parallelogrammatic platform which can be flexibly folded. Therefore, this mechanism not only can be used to grasp an object having irregular shape or large volume, but also can be utilized as a micro-positioning device after grasping the object. Configuration parameter for configuration control of the proposed gripping mechanism is defined. Grasping force and isotropic characteristic of the mechanism are considered as design indices for optimal design of the gripping mechanism. The proposed mechanism is expected to be used as an adaptable gripping and micro-positioning device by attaching it at the end-point of a robot.

Concerns a sawing task by two cooperating arms. To follow a line in a horizontal plane, three directional motions must be controlled, 2 translational and 1 rotational. A certain level of force must be controlled in the vertical direction, to keep contact with the object to be sawn. The two-arm system consists of a 4-DOF SCARA robot and a 5-DOF PT200V robot. When the two arms are rigidly grasping a saw, the mobility of the system is 3, which is not enough. Therefore, we deliberately insert a passive joint at the end of the SCARA robot to increase the mobility to 4. A hybrid control method to regulate force and position is proposed. The proposed scheme has three typical features: First, the two arms are treated as one in a kinematic viewpoint. Second, the hybrid control method is based on a Jacobian and an internal kinematics for a single closed-kinematic chain of the two arms to reflect the nature of the position-controlled manipulator. Third, the proposed scheme is not only able to operate the system even if a passive joint exists, but also is able to utilize the internal loads for useful applications such as pitch motion control. We experimentally show that the performance of the position and force response are satisfactory, and that one additional passive joint not only prevents the system from unwanted roll motion in the sawing task, but also allows an unwanted pitch motion to be notably reduced.

For a 5-bar finger mechanism with redundant actuators, it is shown that judicial choice of the location of one redundant actuator greatly enhances the load handling capacity of the system when compared to that of nonredundant systems, And, it is also shown that excessive redundant actuation enables the system to modulate the end-point stiffness or motion frequency by internal load distributions. Especially, the motion frequency modulation via internal loads in redundantly actuated closed-chain systems is believed to be fairly new. To show the effectiveness of the proposed algorithms, several simulation results are illustrated.

For continuous state space applications, a novel method of Q-learning is proposed, where the method incorporates a region-based reward assignment being used to solve a structural credit assignment problem and a convex clustering approach to find a region with the same reward attribution property. Our learning method can estimate a current Q-value of an arbitrarily given state by using effect functions, and has the ability to learn its actions similar to that of Q-learning. Thus, our method enables robots to move smoothly in a real environment. To show the validity of our method, the proposed Q-learning method is compared with conventional Q-learning method through a simple two dimensional free space navigation problem, and visual tracking simulation results involving a 2-DOF SCARA robot are also presented.

Design of a motion tracking controller is proposed for kinematically redundant manipulators. Using the weighted inner product in joint space, a minimal parametrization of the null space can be achieved. Combining this minimal set of null motion with kinematic relation, we obtain a new extended task space formulation. Based on this formulation, a trajectory tracking control law is proposed based on the momentum feedback disturbance observer (MFDOB). The performance of the proposed controller is verified through experiments with a planar three-link direct-drive manipulator.

A biologically inspired, novel concept for an actively adjustable frequency modulator is presented. The idea of frequency modulation stems from the human body. The human body is known to be able to modulate the spring-like impedance property in a feedforward fashion using antagonistically fighting muscles. Therefore, the motion frequency of the system can be also modulated by employing the information on the inertial property and the spring-like impedance property of the human body. Based on this concept, a methodology for the motion frequency modulation via redundant actuation is investigated. A five-bar finger mechanism driven by redundant actuators is given as an illustrative example. It is shown that the motion frequency as well as the amplitude of oscillation can be actively adjustable during the motion, and that the equilibrium position about which the vibration occurs can also be arbitrarily changed during the motion. Furthermore, using the frequency modulation capability, a point-to-point motion is accomplished by a progressive movement of equilibrium posture, which is termed a virtual trajectory. To show the effectiveness of the proposed algorithms, several simulation results are illustrated.

A visual servoing algorithm is proposed for a robot with a camera in hand, where fuzzy logics and fuzzy-neural networks are employed to represent and/or learn camera motion commands to track a moving object in terms of image features and their variations. Specifically, novel image features are suggested by employing a viewing model of perspective projection to estimate relative pitching and yawing angles between the object and the camera. And, owing to the uniqueness of the proposed image features, at most two input variables are shown to be sufficient for the design of fuzzy logics and/or fuzzy-neural networks. To compensate dynamic characteristics of the robot, desired feature trajectories for the learning of visually guided line-of-sight robot motion are obtained by measuring features by the camera in hand not in the entire workspace, but on a single linear path along which the robot moves under the control of a commercially provided function of linear motion. And then, control actions of the camera are approximately found by fuzzy neural networks to follow such desired feature trajectories. To show the validity of proposed algorithm, some experimental results are illustrated, where a four axis SCARA robot with a B/W CCD camera is used.

This paper presents a design experience of a supervisory control system for coordination of multiple robotic devices. To effectively program job commands, a Petri net-type graphical robot language(PGRL) is proposed, where some functions for coordination among tasks such as concurrency and synchronization, can be easily programmed. Each task of PGRL is described by employing formal model languages, which are composed of three modules, sensory, data handling, and action module. It is expected that by using our proposed PGRL and formal model languages, one can efficiently describe a job or task, and hence can easily operate a complex real-time concurrent system. The proposed control system has been implemented by using VME-based 32-bit microprocessor boards and a real-time multitasking operating system (VxWorks), and is shown to successfully work for robotic jobs.

The concept of dynamic tracking line is proposed as the feasible tracking region for a robot in a robot-conveyor system, which takes the variable conveyor speed into consideration. This paper presents an effective method to find the dynamic tracking line in a robotic workcell. The maximum permissible line-speed which is a quantitative measure of the robot capability for conveyor tracking, is defined on the basis of the relation between the end-effector speed and the bounds on the joint velocities, accelerations, and torques. This measure is derived in an analytic form, and some of its properties are established mathematically. The problem of finding the dynamic tracking line is then formulated as a root-solving problem for a single-variable equation, and solved by the use of a simple numerical technique.

A hierarchical path planning algorithm is proposed to plan a collision free path for mobile robots and robot manipulators in an environment with stationary obstacles as well as multi-robot manipulators in a common workspace. The proposed algorithm consists of two parts: tunnel finding and path planning. The tunnel finding algorithm constructs a free subspace in which start and goal configurations are included and a collision free path candidate can be found. Specifically, the tunnel is constructed by using big cells defined as a group of several basic cells. The path planning algorithm then plans a path in the tunnel. In addition to the capability of planning a collision free path in an environment with stationary obstacles, the proposed algorithm is capable of finding a collision-free path for multi-robot manipulators in a common workspace. To show the validity of the proposed algorithm, some numerical examples are illustrated for mobile robot and multiple robot manipulators in a common workspace.

A compliance control strategy is proposed by using a new type of self-controlled stiffness function, where no explicit wrist force/torque sensor is employed. Specifically, the stiffness gain is given as the exponential function of the error between virtually given desired trajectory and actual trajectory, where virtual desired trajectory is designed to be inside the surface of the object to be compliance-controlled. And, a motion speed scheduling technique is also proposed to avoid serious damages due to lack of environmental knowledge and to minimally maintain the task performances. It is experimentally shown that the proposed control method is useful for various manipulator motion types such as path following, initial contact with soft impact and force regulation during contact.

A hierarchical planning strategy for dextrous manipulation of multifingered hands with soft finger contact model is proposed. Dextrous manipulation planning can be divided into a high-level stage which specifies the position/orientation trajectories of the fingertips on the object and a low-level stage which determines the contact forces and joint trajectories for the fingers. In the low-level stage, various nonlinear optimization problems are formulated according to the contact modes and integrated into a manipulation planning algorithm to find contact forces and joint velocities at each time step. Montana's contact equations are used for the high-level planning. Quasistatic simulation results are presented and illustrated by employing a three-fingered hand manipulating a sphere to demonstrate the validity of the proposed low-level planning strategy.>

A hierarchical planning strategy for dextrous manipulation of multifingered hands with soft finger contact model is proposed. Dextrous manipulation planning can be divided into a high-level stage which specifies the position/orientation trajectories of the finger-tips on the object and a low-level stage which determines the contact forces and joint trajectories for the fingers. In the low-level stage, various nonlinear optimization problems are formulated according to the contact modes and integrated into a manipulation planning algorithm to find contact forces and joint velocities at each time step. Montana's contact equations (1988) are used for the high-level planning. A real-time compensation tactics to eliminate the trajectory errors of the object resulted from various uncertainties are also developed. Simulation results are presented and illustrated by employing a three-fingered hand manipulating a sphere to demonstrate the validity of the proposed strategy.>

In this paper, a fuzzy force control algorithm is suggested for commercialized industrial robots equipped with the position servo drives, where control rules of the proposed fuzzy controller are changed according to the magnitude of environmental stiffness in such a way that good force response is maintained regardless of changes of environmental stiffness. Specifically, some fuzzy control rules are designed for several representative environmental stiffness values, and then a control action for a given arbitrary environmental stiffness value is decided by a fuzzy interpolation method. To show the validity of the proposed fuzzy controller, several experimental results are illustrated, where a 5-axis articulated robot manipulator equipped with the wrist force/torque sensor system and our prototype dual robot controller are employed.>

It is shown that there exists a nonlinear mapping which transforms image features and their changes to the desired camera motion without measuring of the relative distance between the camera and the object. This nonlinear mapping can eliminate several difficulties occurring in computing the inverse of the feature Jacobian as in the usual feature-based visual feedback control methods. Instead of analytically deriving the closed form of this mapping, a fuzzy membership function (FMF) based neural network incorporating a fuzzy-neural interpolating network is proposed to approximate the nonlinear mapping, where the structure of the FMF network is similar to that of radial basis function neural network which is known to be very effective in the function approximation. Several FMF networks are trained to be capable of tracking a moving object in the whole workspace along the line of sight. For an effective implementation of the proposed FMF network, an image feature selection process is investigated, and the required fuzzy membership functions are designed. Finally, several numerical examples are presented to show the validity of the proposed visual servoing method.>

A generalized algorithm for the motion/force planning of the multifingered hand is proposed to generate finite displacements and changes in orientation of objects by considering sliding contacts as well as rolling contacts between the fingertip and the object at the contact point. Specifically, a nonlinear optimization problem is firstly formulated and solved to find joint velocities and contact forces to impart a desired motion to the object at each time step. Then, the relative velocity at the contact point is found by calculating the velocity of the fingertip and the object at the contact point. Finally, time derivatives of the surface variables and the contact angle of the fingertip and the object at the current time step is computed using the Montana's contact equation to find the contact parameters of the fingertip and the object at the next time step. To show the validity of the proposed algorithm, a numerical example is illustrated by using the robotic hand manipulating a circular cylinder with three fingers each of which has four joints.

A neural optimization network is proposed to control redundant robot manipulators in an environment with obstacles. The weightings of the network are adjusted by considering both the joint dexterity and the capability of collision avoidance of joint differential motion. The fuzzy rules are proposed to determine the capability of collision avoidance of each joint. To show the validities of the proposed method, computer simulation results are illustrated for a 3-d.o.f planar redundant robot.>

A force/position control method is proposed for robot manipulators equipped with a wrist force/torque sensor and position servo drives. Especially, fuzzy PI gain tuning rules are suggested to improve the transient force response. To show the validity of the proposed method, a numerical example employing a two-link manipulator is given.>

A visual servoing method is proposed for eye-in-hand robots. It employs a self-organizing fuzzy controller. There is defined a Jacobian that is not a function of a relative position of the object but is a function of the image features. Instead of obtaining an analytic form of the proposed Jacobian, a self-organizing fuzzy controller is then proposed to alleviate difficulties in real-time implementation. To show its validity, the proposed method is applied to a 2D visual servoing task.>

A novel method is developed to coordinate the motion of dual robot arms carrying a solid object, where the first robot (leader) grasps one end of the object rigidly and the second robot (follower) is allowed to change its grasping position at the other end of the object along the object surface while supporting the object. It is shown that this flexible grasping is equivalent to the addition of one more degree of freedom, giving the follower more maneuvering capabilities. In particular, motion commands for the follower are generated by using kinematic redundancy. To show the utility and power of the method, a system with two PUMA 560 robots carrying a beam is analyzed in detail.>

A method is developed to coordinate the motion of dual robot arms carrying a solid object, where the first robot (leader) grasps one end of the object rigidly and the second robot (follower) is allowed to change its grasping position at the other end of the object along the object surface while supporting the object. It is shown that this flexible grasping is equivalent to the addition of one more degree of freedom, giving the follower more maneuvering capabilities. Especially, motion commands for the follower are generated by using kinematic redundancy. To show the utility and power of the method, an example system with two PUMA 560 robots carrying a beam is analyzed.>

An iterative learning control algorithm for a class of linear periodic systems is proposed in which parameter estimation is performed in the domain of an iterative sequence of operations with the time frozen. A sufficient condition for the convergence of the proposed algorithm is given. It is also shown that the proposed learning controller can be applied to the continuous path control of robot manipulators. It is noted that the proposed algorithm works reasonably well only for the case of small perturbations with respect to a nominal trajectory.>