The Bradley-Terry (BT) model is widely practiced in reward modeling for reinforcement learning with human feedback (RLHF). Despite its effectiveness, reward models (RMs) trained with BT model loss as one-way classifiers are prone to over-optimization, losing generalizability to unseen inputs. In this paper, we study the cause of over-optimization and its downstream effects on the RLHF procedure, highlighting the importance of robustness in RMs. First, we show that the excessive dispersion of hidden state norms is the main source of over-optimization. Correspondingly, we propose batch-wise sum-to-zero regularization (BSR) that enforces reward sum for each batch to be zero-centered, constraining the rewards with abnormally large magnitudes. We assess the impact of BSR in improving robustness in RMs through four scenarios of over-optimization, where BSR consistently manifests better robustness on unseen inputs. Then, we compare the plain BT model and BSR on RLHF training and empirically show that robust RMs better align the policy to the gold preference model. Finally, we apply BSR to high-quality data and models, which surpasses state-of-the-art RMs in the 8B scale by adding more than 5\% in complex preference prediction tasks. By conducting RLOO training with 8B RMs, AlpacaEval 2.0, with reducing generation length by 40\% while adding a 7\% increase in win rate, further highlights that robustness in RMs induces robustness in RLHF training.

Large language models are trained with tokenizers that map text to a fixed vocabulary, yet the resulting token distribution is highly imbalanced: a few words dominate the stream while most occur rarely. Recent practice favours ever-larger vocabularies, but it is unclear whether the benefit comes from better word segmentation or from amplifying this frequency skew. To this end, we perform a controlled study that scales the vocabulary of a constant-size Transformer from 24K to 196K symbols while holding data, compute and optimisation unchanged. Above 24K every common word is already a single token, so further growth only increases imbalance. Word-level loss decomposition shows that larger vocabularies reduce cross-entropy almost exclusively by lowering uncertainty on the ~$2,500$ most frequent words, even though loss on the rare tail rises. Same frequent words cover roughly $80\%$ of tokens in downstream benchmarks, this training advantage transfers intact. We further show that enlarging model parameters with a fixed tokenizer yields the same frequent-word benefit, revealing a shared mechanism behind vocabulary and model scaling. Our results recast “bigger vocabularies help” as “sharper frequency imbalance helps,” offering a simple, principled knob for tokenizer–model co-design and clarifying the loss dynamics that govern language-model scaling in pre-training.

Stable pre-training is essential for achieving better-performing language models.However, tracking pre-training stability by calculating gradient variance at every step is impractical due to the significant computational costs.We explore Token Embedding Variability (TEV) as a simple and efficient proxy for assessing pre-training stability in language models with pre-layer normalization, given that shallower layers are more prone to gradient explosion (section 2.2).Moreover, we propose Multihead Low-Rank Attention (MLRA) as an architecture to alleviate such instability by limiting the exponential growth of output embedding variance, thereby preventing the gradient explosion (section 3.2).Empirical results on GPT-2 with MLRA demonstrate increased stability and lower perplexity, particularly in deeper models.

This paper presents a method for LiDAR sensor-based traversability analysis for autonomous mobile robots in urban environments. Although urban environments are structured environments, a typical terrain comprises hazardous regions for mobile robots. Therefore, a reliable method for detecting traversable regions is required to prevent robots from getting stuck in the middle of the road. Conventional approaches require considerable efforts to obtain a model for traversability analysis for a specific robot or environment. In particular, learning-based methods require explicit training data. This paper introduces a method for traversability mapping based on a self-training algorithm to eliminate the hand labeling process. A neural network was applied to the underlying classifier of the self-training algorithm. With our approach, the model can be learned with even weakly labeled data obtained from robot-specific parameters and the robot’s footprint. In practical experiments, the self-trained model performed better performance than the existing supervised learning method. Moreover, as the fraction of unlabeled data increased, the performance also increased. Therefore, the demonstrations in the urban environments indicate the effectiveness of the proposed method for traversability mapping.

The article proposes a new heuristic for task allocation and routing of heterogeneous robots. Specifically, we consider a path planning problem where there are two (structurally) heterogeneous robots that start from distinctive depots and a set of targets to visit. The objective is to find a tour for each robot in a manner that enables each target location to be visited at least once by one of the robots while minimizing the maximum travel cost. A solution for Multiple Depot Heterogeneous Traveling Salesman Problem (MDHTSP) with min-max objective is in great demand with many potential applications, because it can significantly reduce the job completion duration. However, there are still no reliable algorithms that can run in short amount of time. As an initial idea of solving min-max MDHTSP, we present a heuristic based on a primal-dual technique that solves for a case involving two robots while focusing on task allocation. Based on computational results of the implementation, we show that the proposed algorithm produces a good quality of feasible solution within a relatively short computation time.

Tree-structured neural network architectures for sentence encoding draw inspiration from the approach to semantic composition generally seen in formal linguistics, and have shown empirical improvements over comparable sequence models by doing so.Moreover, adding multiplicative interaction terms to the composition functions in these models can yield significant further improvements.However, existing compositional approaches that adopt such a powerful composition function scale poorly, with parameter counts exploding as model dimension or vocabulary size grows.We introduce the Lifted Matrix-Space model, which uses a global transformation to map vector word embeddings to matrices, which can then be composed via an operation based on matrix-matrix multiplication.Its composition function effectively transmits a larger number of activations across layers with relatively few model parameters.We evaluate our model on the Stanford NLI corpus, the Multi-Genre NLI corpus, and the Stanford Sentiment Treebank and find that it consistently outperforms TreeLSTM (Tai et al., 2015), the previous best known composition function for treestructured models.

In this study, we propose a novel Light Detection and Ranging (LiDAR) sensor-based localization method for localization of a mobile robot. In localization using the LiDAR sensor, localization errors occur when real range measurements differ from reference distances computed from a map. This study focuses on three factors that cause differences between real range measurements and reference distances. The first factor corresponds to optical characteristics of the LiDAR sensor for objects such as glass walls and mirrors. The second factor corresponds to occlusions by dynamic obstacles. The third factor corresponds to static changes in the environment. In practical applications, three factors often simultaneously occur. Although there have been many previous works, robust localization to overcome these three difficulties is still a challenging problem. This study proposes a novel robust localization scheme that exploits only reliable range measurements. A LiDAR sensor-based localization scheme can be successfully executed by utilizing only a few reliable range measurements. Therefore, the computation of reliability plays a significant role. The computation of reliability is divided into two steps. The first step considers characteristics of optical sensors. The second step mainly deals with the effects of obstacles. The observation likelihood model exploits computed reliability for pose estimation. The proposed scheme was successfully verified through various experiments under challenging situations.

Human-following in a cluttered environment is one of the challenging issues for mobile service robot applications. Since a laser range finder (LRF) is commonly installed for autonomous navigation, it is advantageous to adopt an LRF for detection and tracking humans. In this paper, we aim at the reliable human tracking performances in a dynamic cluttered environment. The key idea is to develop a hierarchical sample-based joint probabilistic data association filter (HSJPDAF) by focusing on the leg positions as well as human positions. The proposed HSJPDAF consists of two levels in order to consider the interdependence between targets. Possible locations of multiple human targets can be simultaneously estimated on the basis of Bayesian filtering. Comparison with the general technique was carried out to verify the performance of HSJPDAF in both artificial indoor and real-world environments. Owing to the hierarchical framework, the proposed method shows the improved robustness by reducing the target loss rate significantly in a dynamic cluttered environment.

Urban road environments that have pavement and curb are characterized as semi-structured road environments. In semi-structured road environments, the curb provides useful information for robot navigation. In this paper, we present a practical localization method for outdoor mobile robots using the curb features in semi-structured road environments. The curb features are especially useful in urban environment, where the GPS failures take place frequently. A curb extraction is conducted on the basis of the Kernel Fisher Discriminant Analysis (KFDA) to minimize false detection. We adopt the Extended Kalman Filter (EKF) to combine the curb information with odometry and Differential Global Positioning System (DGPS). The uncertainty models for the sensors are quantitatively analyzed to provide a practical solution.

Human tracking has become an important part in the mobile robotics community to achieve human-friendly navigation. When it comes to extracting humans from sensor data when there are many people in its environment, matching between targets and measurements can be a significant problem. In this paper, a single laser range finder is adopted as a sensor, and we implement JPDAFs, one of the most eminent methods for matching, to track humans. The performance is tested through experiments on actual human walking.

Odometry using wheel encoders provides fundamental pose estimates for wheeled mobile robots. Systematic errors of odometry can be reduced by the calibration of kinematic parameters. The UMBmark method is one of the widely used calibration schemes for two wheel differential mobile robot. In this paper, the accurate calibration scheme of kinematic parameters is proposed by extending the conventional UMBmark. The contributions of this paper can be summarized as two issues. The first contribution is to present new calibration scheme that reduce the approximation error of calibration equations in UMBmark method. The new equations were derived from the robot's final orientation errors in test track. The second contribution is to propose the consideration of the coupled effects between wheel diameter errors and wheelbase errors in test track for calibration experiments. The calibration performance can be significantly improved by applying new calibration equations were derived to overcome the limitations of the conventional scheme. The numerical simulations and experimental results show that the odometry accuracy can be improved by the proposed calibration schemes.

There have been two major streams for the motion control of mobile robots. The first is the model-based deliberate control and the second is the sensor-based reactive control. Since two schemes have complementary advantages and disadvantages, one cannot completely replace the other. There are a variety of environmental conditions which affect the navigation performances. The main idea of this paper is to design discrete navigation behaviors and to integrate behaviors by an appropriate selection framework. In this paper, we propose a behavior selection framework using the GSPN (Generalized Stochastic Petri Nets). We have designed two navigation behaviors which show completely different performances. In order to define behavior selection criteria, two kinds of navigation statuses are defined to monitor navigation performances of the robot. The proposed navigation strategy is simulated using the open source simulator Player/Stage to investigate the performances in a variety of conditions. Through the simulations, it was made clear that different behaviors show remarkably different performances. Moreover, the average navigation time of the proposed behavior selection framework is significantly decreased than that of any single navigation scheme DWA or tracking.

There are two different motions of a trailer system that consists of a car and n passive off-hooked trailers. When a car “pulls” n passive trailers, a trailer system moves forward. A trailer system moves backward when a car “pushes” n passive trailers. Backing up a trailer system is difficult because it is an open loop unstable problem. In this paper, we investigate the backward motion control problem of a car with n passive trailers. We have shown that n passive trailers can be successfully controlled by an omni-directional mobile robot in our prior works. Unlike an omni-directional robot, a car-like mobile robot has nonholonomic constraints and limitations of the steering angle. For these reasons, the backward motion control problem of a car-like mobile robot with n passive trailers is not trivial. In spite of difficulties, backing up a trailer system is useful for parking control. In this study, we proposed a mechanical alteration which is connecting n passive trailers to the front bumper of a car to improve the backward motion control performance. By adopting the new design, a car pushes n passive trailers by its forward motion. A practical trailer-pushing control algorithm was also proposed. Stability analysis of the controller under kinematic modeling error was presented. Theoretical verification and experimental results proved that the control strategy of pushing n passive trailers by forward motion of a car can be successfully implemented.

For outdoor navigation, it is necessary to find the relevant features of outdoor road environments and detect drivable region for robot's motion. This paper presents a methodology for extracting the drivable road region by detecting the prominent road features and obstacles through a single laser range finder. The prominent features of roads are curbs and the road surface. The laser range finder is mounted on the mobile robot, looks down the road with a small tilt angle, and obtains two-dimensional range data. The proposed method is computationally more efficient in comparison with vision-based techniques and applicable for various road conditions in target environment. Experimental results confirm the reliability of the algorithm.

Recently, automatic parking assist systems have become commercially available in some cars. In order to improve the reliability and accuracy of parking control, pose estimation problem needs to be solved. Odometry is widely used for pose estimation of a mobile robot. However, most previous odometry calibration methods have focused on two wheeled mobile robots. In this paper, we consider systematic error sources of the car-like mobile robot(CLMR), and we suggest a useful calibration method for systematic errors. Finally, our calibration method is verified by experiments using a miniature car.

Automatic parking assist system is one of the key technologies in future automobiles. A car-like robot is difficult to control due to the nonholonomic constraints. In this paper, a path planning algorithm based on the slice projection technique, optimized particularly for car-parking is proposed. Collision-free region of a nonholonomic motion can be easily computed by using the slice projection technique. Sufficient collision-free, nonholonomic feasible paths can be directly obtained by computing reachable regions by using the slice projection technique for nonholonomic motions from the initial pose. The proposed planning scheme provide not a single solution, but a candidate solution set. Therefore, the parking path can be easily optimized with respect to performance criteria such as safety, maneuvering, and so on. The simulation results clearly show that the proposed scheme provides various practical solutions.

A mobile service robot should have the ability to carry out multiple practical tasks. However, most mobile mobile service robots have been designed to carry out a specific task. Hence it has been difficult to design a navigation framework for multiple navigation tasks. In this paper, we will design and implement a navigation framework for multiple tasks using the GSPN (Generalized Stochastic Petri-Nets). The GSPN-based navigation framework has several advantages, such as performance analysis and dynamic system analysis. Furthermore, such a design can process system events in realtime. Also, the typical problem of a hybrid control system, namely the synchronization and scheduling between reactive layer and high-level planner can be solved using the GSPNpsilas event monitoring ability using the timed transition and immediate transition. Finally, the idle time and activation time of the tasks can be predicted under the GSPN-based navigation framework. In this paper, firstly, the transportation task and patrol task are implemented. Then, a navigation framework is designed using the GPSN. The proposed GSPN-based navigation framework is verified in the human co-existing environments by experiments. The results showed that the proposed navigation scheme can be used practically in real environments.

In this paper, we present one approach to achieve safe navigation in indoor dynamic environment. So far, there have been various useful collision avoidance algorithms and path planning schemes. However, those algorithms have a fundamental limitation that the robot can avoid only "visible" obstacles. In real environment, it is not possible to detect all the dynamic obstacles around the robot. There exist a lot of "invisible" regions due to the limitation of field of view. In order to avoid possible collisions, it is desirable to consider visibility information. Then a robot can reduce the speed or modify a path. This paper proposes a safe navigation scheme to reduce the risk of collision due to unexpected dynamic obstacles. The robot's motion is controlled according to a hybrid control scheme. The possibility of collision is dually reflected to a path planning and a speed control. The proposed scheme clearly indicates the structural procedure how to model and to exploit the risk of navigation. The proposed scheme is experimentally tested in a real office building. The presented result shows that the robot moves along the safe path to obtain sufficient field of view, while appropriate speed control is carried out.

Most present localization algorithms are either range or vision-based. In many environments, only one type of sensor cannot often ensure successful localization; furthermore, using low-priced range sensors instead of expensive, but accurate, laser scanners often lead to poor performance. This paper proposes an MCL-based localization method that robustly estimates the robot pose with fusion of the range information from a low-cost IR scanner and the SIFT based visual information gathered using a mono camera. With sensor fusion, the rough pose estimation from range-based sensors is compensated by the vision-based sensors and slow object recognition can be overcome by the frequent update of the range information. In order to synchronize the two sensors with different bandwidths, the encoder information gathered during object recognition is exploited. This paper also suggests a method for evaluating localization performance that is based on the normalized probability of a vision sensor model. Various experiments show that the proposed algorithm can estimate the robot pose reasonably well and can accurately evaluate the localization performance.

Appropriate design and control of behaviors of mobile robots are important for their successful autonomous navigation in a real dynamic environment. This paper proposes a formal selection framework of multiple navigation behaviors for a service robot. In the presented approach, modeling, analysis, and performance evaluation are carried out based on generalized stochastic Petri nets (GSPNs). By adopting a probabilistic approach, the proposed framework helps the robot to select the most desirable navigation behavior in run time according to environmental conditions. Moreover, after mission completion, the robot evaluates its prior navigation performance from accumulated data, and automatically uses the results to improve its future operations. Also, GSPNs have several advantages over direct use of other modeling formalisms such as finite state automata (FSA) or Markov processes (MPs). We conduct experiments on real guidance tasks with visitors by implementing the framework in the guide robotJinnyat the National Science Museum of Korea. The results show that the proposed strategy is useful for a robot's selection of an appropriate navigation behavior in a dynamic environment.

Monte Carlo localization is known to be one of the most reliable methods for pose estimation of a mobile robot. Many studies have been conducted to improve performance of MCL. Although MCL is capable of estimating the robot pose when the initial pose of a robot is not given, it takes much time for convergence because a large number of random samples are required, especially for the large-scale environment. For practical implementation of MCL, therefore, it is desirable to reduce the number of samples without affecting the localization performance. This paper presents a novel approach to reduce the number of samples used in the particle filter for efficient implementation of MCL. To this end, the topological information is extracted in real time through the thinning algorithm from the range data of a laser scanner. The topological map is first created from the given grid map of the environment. The robot scans the local environment and generates a local topological map. The robot then navigates along this local topological edge, which coincides with the global topological map obtained off-line from the given global grid map. By constraining the robot's motion on this local edge, random samples are drawn only around the neighborhood of the topological edge rather than over the entire free space. Hence the sample size required for MCL can be drastically reduced, thereby reducing computational time for the MCL process. A series of experiments based on this proposed MCL/TI show that the number of samples can be reduced considerably, and thus the time required for pose estimation can be substantially decreased

This paper proposes a selection framework of multiple navigation primitives for a service robot using Generalized Stochastic Petri Nets (GSPN’s). By adopting probabilistic approach, our framework helps the robot to select the most desirable navigation primitive in run time through the performance estimation according to environmental conditions. Moreover, after a mission, the robot evaluates prior navigation performance from accumulated data, and uses the results for the improvement of future operations. Modeling, analysis, and performance evaluation are conducted on firm mathematical foundation. Also, GSPN’s have several advantages over classic automata or direct use of Markov Process. We conducted simulations of the model derived from our experience of practical installations. The results showed that the framework is useful for primitive selection and performance analysis.

A passive multiple trailer system provides various practical advantages in order to develop multi-functional service robots. So far, we have proposed kinematic design of passive trailers, and backward motion control schemes in our prior work. In this paper, we propose two significant considerations based on practical experiences on backward control of multi-trailer system. The first issue is an analysis on the effect of joint errors. Joint errors include joint angle sensing errors and backlashes at the joints. Due to the joint error, the trajectory shows an offset distance from a desired reference path. A second issue is a slip detection based on a simplified dynamic model of the trailer system. The trailer system is designed based on the kinematic constraints, which imply rolling without slipping at wheels. We discuss how a practical trailer system can be controlled without violating kinematic constraints. Finally, we show the analogies and relationships between the off-hooked trailer and the previously proposed three point trailer.

This paper proposes a formal selection framework of multiple navigation behaviors for a service robot. In our approach, modeling, analysis, and performance evaluation are carried out based on the generalized stochastic Petri nets (GSPN). By adopting probabilistic approach, our framework helps the robot to select the most desirable navigation behavior in run time according to environmental conditions. Moreover, after a mission, the robot evaluates prior navigation performance from accumulated data, and uses the results for the improvement of future operations. Also, GSPN has several advantages over classic automata or direct use of Markov process. The basic ideas of the framework were introduced in our previous work (Gunhee Kim et al., 2005). Thus, this paper focuses on experimental verification by implementing the framework into the guide robot Jinny. We conduct the experiments about real guidance tasks with visitors in the National Science Museum of Korea. The results show that the proposed strategy is useful to select an appropriate navigation behavior in a dynamic space.

Due to dynamic changes of an environment and various kinds of uncertainties in a real world, mobile robot localization is difficult to be solved by a single continuous algorithm. In order to achieve a practical localization solution generally, this paper proposes a strategy to deal with various uncertainties using explicit discretization of robot's status. Discrete status of localization is designed with three criteria as follows: (i) polygonal environment and non-polygonal environment; (ii) static environment and dynamic environment; and (iii) global positioning problem and local tracking problem are defined. An appropriate strategy is adopted according to the robot's status. The feasibility of the proposed method is demonstrated by simulation results.

This paper describes an integrated navigation strategy for the autonomous service robot in large-scale indoor environments. It includes architecture of navigation system, the development of crucial navigation algorithms like map, path planning, and localization, and planning scheme such as error/fault handling. Major advantages of proposed navigation are as follows: 1) A range sensor based generalized scheme of navigation without modification of the environment. 2) Intelligent navigation-related components. 3) Framework supporting the selection of multiple behaviors and error/fault handling schemes. A experimental result shows the feasibility of proposed navigation system. The result of this research has been successfully applied to our three service robots in a variety of task domains including a delivery, a patrol, a guide, and a floor cleaning task.

This paper describes the implementation of multi-functional service robots using the Tripodal schematic control architecture. Our strategy has two major advantages. First, the proposed architecture supports Petri net based formal description of tasks and error/fault handling schemes. Second, it provides intuitive and straightforward guidelines system for integration issues. We show reusability and scalability of the proposed architecture by giving two examples of our experience. First, we explain how to add a newly developed cleaning function to our robot system. Second, we introduce the implementation process of a newly developed guide robot Jinny. Most of the modules developed for former robots are used directly in the Jinny system. Experimental results clearly showed that the developed strategy is efficient and easy-to-use.

Although a great deal of localization methods have been proposed, when it comes to human coexisting real world there are still many unsolved problems. It is because real world contains various kinds of uncertainties. For reliable navigation in such a world, this paper proposes a new localization synthesis integrated localization. The integrated localization is a dependable active localization approach, which is the structural synthesis of navigation modules. Due to a dynamic change of an environment, robot navigation cannot be solved by a single algorithm. In this paper, various situations are classified into different status, which is modeled as discrete events. Then, developed algorithms are synthesized in a structured way. The discrete event control structure enables efficient combination of position estimation algorithms and synthesis of navigation modules. Furthermore, the scheme provides structural framework for dead lock avoidance. The proposed technique is applied to KIST public service robots and shown to be useful in real experiments.

A mobile service robot can achieve reconfigurability by exploiting passive trailer systems. Reconfigurability provides significant practical advantages in order to deal with various service tasks. However, a motion control problem of a passive multiple trailer system is difficult, mainly because a kinematic model is highly nonlinear. It is shown how a robot with n passive trailers can be controlled in backward direction. Once a desired trajectory of a last trailer is computed, then the control input of a pushing robot is obtained by the proposed control scheme. A kinematic design of a trailer system is proposed in our prior work. It is shown that the high performance of trajectory tracking is also valid for the backward motion control problem. Experimental verifications were carried out with the PSR-2 (public service robot) with three passive trailers. Experimental result showed that the backward motion control could be successfully carried out using the proposed control scheme.

KIST service robot is composed of a mobile platform, a 6 DOF robotic manipulator, and a multi-fingered robotic hand. We discuss motion control and coordination strategy in order to deal with uncertainty problem in practical applications. A door opening is our target task. Since the environment is not prepared for a service robot, it is essential to deal with various uncertainties due to a robot positioning error, sensing error, as well as manipulation errors. In this paper, practical parameter estimation schemes are proposed from the viewpoint of coordinative motion control of a hand, a manipulator and a mobile robot. Analysis of physical properties of each component provides a methodology of appropriate role assignment for each component. In order to carrying out compliance control, an external force is computed using fingertip force information of the three fingered robot hand, instead of using force torque sensor at the wrist. Presented experimental result clearly shows the effectiveness of the proposed scheme.

This paper explains a new tour-guide robot Jinny. The Jinny is developed by focusing on human robot interaction and autonomous navigation. In order to achieve reliable and safe navigation performance, an integrated navigation strategy is established based on the analysis of a robot's states and the decision making process of robot behaviors. According to the condition of environments, the robot can select its motion algorithm among four types of navigation strategy. Also, we emphasized the manageability of a robot's knowledge base for human friendly interactions. The robot's knowledge base can be extended or modified intuitively enough to be managed by non-experts. In order to show the feasibility and effectiveness of our system, we also present experimental results of the navigation system and some experiences on practical installations.

This paper describes a control architecture design and a system integration strategy for the autonomous service robot PSR (Public Service Robot). The PSR is under development at the KIST (Korea Institute of Science and Technology) for service tasks in public spaces such as office buildings and hospitals. The proposed control architecture is designed by tripodal frameworks, which are layered functionality diagram, class diagram, and configuration diagram. The tripodal schematic design clearly points out the way of integrating various hardware and software components. The developed strategy is implemented on the PSR and successfully tested.

In this paper, a reliable position estimation method of the indoor service robot is proposed. The service robot PSR1 is a wheeled mobile manipulator which navigates in office buildings. Our localization method is a map-matching scheme using scanned range data, without using any artificial landmark. The proposed algorithm can provide solutions for both a global localization problem and a local position tracking. A probabilistic position estimation scheme is designed based on MCL (Monte Carlo localization). Two measure functions are developed for computing positional probabilities. The robot automatically decides whether it uses geometric pattern matching (i.e. walls, pillars) by Hough transform. The proposed scheme shows reliable performance in both polygonal environments and non-polygonal environments even there exist many obstacles. Experimental results demonstrate the validity and feasibility of the proposed localization algorithm for the service robot to navigate in an office building, using the natural environmental characteristics.

In this paper, an autonomous map building method and an intelligent position estimation method for an indoor service robot are presented. Map building is composed of three processes: (1) environmental information gathering, (2) scan registration, and (3) grid map building. A grid map of the environment can be successfully generated using the proposed strategy. Previously [Dongheui Lee et al., 2003] we proposed a localization method which is a map-matching scheme using scanned range data, without using any artificial landmarks. In this paper, an extended localization method called smart localization is presented. Smart localization includes the Petri net based discrete event control concept as well as the position estimation scheme using map matching. A mobile robot is able to act intelligently even if various real world problems arise when using discrete event control. For example, when the robot is unable to compute its position, discrete event based error handling logics are activated according to the predetermined behavioral configuration. Experimental results demonstrate the validity and feasibility of the proposed algorithm for a service robot to navigate in an office building.

Map building and localization are essential to an autonomous mobile robot system. A global topological map is commonly constructed from a global grid map using a Voronoi diagram. In this hierarchical method, however, the advantage of a topological map such as compactness may not be fully utilized. In this paper sensor data are used to build a local grid map from which a local topological map is created using a thinning method. The thinning method is an alternative of the Voronoi diagram. The local topological map is then updated to the global one. Localization based on the topological map is usually difficult, but additional nodes created by the thinning method can improve localization performance. Various experiments with a laser scanner demonstrate validity and feasibility of the proposed algorithms for map building and localization.

It is shown how a passive multiple trailer system is designed and controlled for practical applications. The main focus is reducing the trajectory tracking error and the scope of the paper is limited to the passive systems due to the cost and simplicity of the trailer mechanism. Three kinds of passive trailers are introduced, and then the analysis of path tracking performance is carried out, in order to show that the proposed off-hooked trailer is advantageous in a practical sense. Stability, issues are discussed with constant reference velocity. Numerical simulation results and experimental results are presented.

Nonholonomic constraints are exploited to design a controllable n-joint manipulator with only two inputs. Gears subject to nonholonomic constraints are designed to transmit velocities from the inputs to the passive joints. The system possesses a triangular structure for which a conversion into chained form is presented. The nonholonomic manipulator can, therefore, be controlled with an open loop or a closed loop using existing controllers for chained form. Mechanical design is established, and experimental results proved the usefulness of design of the nonholonomic manipulator and applied control schemes. While previous publications have assumed that the nonholonomic systems are given and have developed theory for these systems, this paper points out a new direction where the nonholonomic theory is used to design controllable and stabilizable systems.

A wheeled mobile robot with trailers has been studied as a class of nonholonomic systems. It is proved that a system of a tractor and trailers with an appropriate connecting mechanism can be stabilized to desired positions via nonholonomic motion control. Trailers, on the other hand, have been developed and widely used in the industry. The main focus of industrial design is set on reducing tracking error from a reference trajectory. This paper attempts to bridge over the gap between these two approaches. We develop a design theory of trailer systems with passive steering. The designed systems show a good performance in practical path following, and accept the chained form transformation and nonlinear control strategies for nonholonomic systems.

A wheeled mobile robot with trailers has been studied as a class of nonholonomic systems. It is proved that a system of a tractor and trailers with an appropriate connecting mechanism can be stabilized to desired positions via nonholonomic motion control. Trailers, on the other hand, have been developed and widely used in the industry. The main focus of industrial design is set on reducing tracking error from a reference trajectory. This paper attempts to bridge over the gap between these two approaches. We develop a design theory of trailer systems with passive steering. The designed systems show a good performance in practical path following, and accept the chained form transformation and nonlinear control strategies for nonholonomic systems.

For robot teleoperation, a lot of research has been done to control the slave robot from a remote site using a masterarm. The concept of a distributed controller architecture for the masterarm is introduced in the paper. To reduce possible non-uniform time delay and to have higher position command update rate, the host controller with distributed satellite controllers is developed. Each satellite controller measures the corresponding joint angle, while the host controller performs initial calibration, inverse kinematics calculation, and position commands generation. The host controller and the satellite controllers are networked via a SPI (serial peripheral interface) protocol. For force feedback, the host controller solves the joint torque relations based on the kinematics of the masterarm and the satellite controller controls each joint's torque. This distributed controller greatly reduces calculation and control loads and makes the electrical wiring very simple. In order to measure finger movement of the operator, the master hand is developed using infrared light. A similar slave satellite controller for this master hand is interfaced to the host controller via a SPI network. For teleoperation of the KIST humanoid robot which has two arms and two hands, the masterarm controller with one host controller and 16 satellite controllers is implemented. Force feedback with less than 2 msec update rate is achieved. This masterarm/hand is also integrated with a graphic model instead of using the slave robot. The experimental results show the performance of the distributed controller.

Exploiting unique features of nonholonomic systems, innovative and advantageous mechanisms can be designed. We have proposed the chained form manipulator which is a controllable n-joint manipulator with only two actuators. The chained form manipulator is designed not only to satisfy chained form convertibility, but also to achieve control simplicity. Design requirements and mechanical design of the chained form manipulator were proposed in our prior work. For the experimental verifications, we fabricated a prototype. So far, various control strategies for nonholonomic systems have been proposed. In this paper, an efficient motion planning scheme for the chained form is presented to approximate any holonomic path with the feasible nonholonomic path. Furthermore, a new concept of motion planning is proposed through the analysis of the initial-condition sensitivity. Combining these two approaches, the motion planning scheme is constructed towards practical applications. Presented experimental results show the usefulness of the design and the applied control scheme.

Exploiting the unique features of nonholonomic systems, we have proposed and fabricated the nonholonomic manipulator which is a controllable n joint manipulator with only two actuators. In order to create its nonholonomic constraint, a special type of velocity transmission, called the nonholonomic gear, was used. There are many possible alternatives of designing underactuated manipulators using the nonholonomic gear. The nonholonomic manipulator was designed focusing on the mechanical simplicity. In this paper, we establish the design procedure of underactuated manipulators whose kinematic model can be converted to the chained form, giving priority to the practical control aspects. As an example that has control simplicity, the chained form manipulator was designed according to the proposed design concept.

We proposed a nonholonomic manipulator which is a controllable n joints manipulator with only two inputs, exploiting a special kind of velocity transmission called a nonholonomic gear. Since the nonholonomic manipulator was theoretically designed from the viewpoint of kinematic constraints and nonlinear control, the mechanical implementation and prototyping are extremely important in practice. In this paper, the principle of mechanical design of a nonholonomic manipulator is established, and the experimental results are shown using a prototype nonholonomic manipulator.

Nonholonomic systems are typically controllable in a configuration space of higher dimension than the input space. Here, it is shown how nonholonomic constraints can be exploited to design a controllable n-joint manipulator with only two inputs. Gears subject to nonholonomic constraints are designed to transmit velocities from the inputs to the unactuated joints. The designed nonholonomic manipulator is shown to be completely controllable in the whole configuration space. The system is designed with a triangular structure for which a conversion into chained form is presented. The nonholonomic manipulator can, therefore, be controlled using existing controllers for chained form.>