Academic literature on the topic 'Mobile robotics, planning, POMDPs, decision-making'

Olfaction is a valuable source of information about the environment that has not been sufficiently exploited in mobile robotics yet. Certainly, odor information can contribute to other sensing modalities, e.g., vision, to accomplish high-level robot activities, such as task planning or execution in human environments. This paper organizes and puts together the developments and experiences on combining olfaction and vision into robotics applications, as the result of our five-years long project IRO: Improvement of the sensory and autonomous capability of Robots through Olfaction. Particularly, it investigates mechanisms to exploit odor information (usually coming in the form of the type of volatile and its concentration) in problems such as object recognition and scene–activity understanding. A distinctive aspect of this research is the special attention paid to the role of semantics within the robot perception and decision-making processes. The obtained results have improved the robot capabilities in terms of efficiency, autonomy, and usefulness, as reported in our publications.

As people attempt to build biomimetic robots and realize automation processes through artificial intelligence, computational intelligence plays a very important role in robotics and automation. This special issue contains several important papers that address various aspects of computational intelligence in robotics and automation. While acknowledging its limited coverage, this special issue offers a range of interesting contributions such as intelligent trajectory planning for flying and land mobile robots, fuzzy decision making, control of rigid and teleoperated robots, modeling of human sensations, and intelligent sensor fusion techniques. Let us scan through these contributions of this special issue. The first paper, "Planar Spline Trajectory Following for an Autonomous Helicopter," by Harbick et al., proposes a technique for planar trajectory following for an autonomous aerial robot. A trajectory is modeled as a planar spline. A behavior-based control system stabilizes the robot and enforces trajectory following of an autonomous helicopter with a reasonable trajectory tracking error on the order of the size of the helicopter (1.8m). In the second paper, "A Biologically Inspired Approach to Collision-Free Path Planning and Tracking Control of a Mobile Robot," by Yang et al., a novel biologically inspired neural network approach is proposed for dynamic collision-free path planning and stable tracking control of a nonholonomic mobile robot in a non-stationary environment, based on shunting equations derived from Hodgkin and Huxley's biological membrane equation. The third paper, "Composite Fuzzy Measure and Its Application to Decision Making," by Kaino and Kaoru, builds a composite fuzzy measure from fuzzy measures defined on fuzzy measurable spaces using composite fuzzy weights by the authors, with a successful application to an automobile factory capital investment decision making problem. In "Intelligent Control of a Miniature Climbing Robot," by Xiao et al., a fuzzy logic based intelligent optimal control system for a miniature climbing robot to achieve precision motion control, minimized power consumption, and versatile behaviors is presented with validation via experimental studies. The fifth paper, "Incorporating Motivation in a Hybrid Robot Architecture," by Stoytchev and Arkin, describes a hybrid mobile robot architecture capable of deliberative planning, reactive control, and motivational drives, which addresses three main challenges for robots living in human-inhabited environments: operating in dynamic and unpredictable environment, dealing with high-level human commands, and engaging human users. Experimental results for a fax delivery mission in a normal office environment are included. In the next paper, "Intelligent Scaling Control for Internet-based Teleoperation," by Liu et al., an adaptive scaling control scheme, with a neural network based time-delay prediction algorithm trained using the maximum entropy principle, is proposed with successful experimental studies on an Internet mobile robot platform. The next paper, "Feature Extraction of Robot Sensor Data Using Factor Analysis for Behavior Learning," by Fung and Liu, discusses important knowledge extraction of sensor data for robot behavior learning using a new approach based on the inter-correlation of sensor data via factor analysis and construction of logical perceptual space by hypothetical latent factors. Experimental results are included to demonstrate the process of logical perceptual space extraction from ultrasonic range data for robot behavior learning. "Trajectory Planning of Mobile Robots Using DNA Computing," by Kiguchi et al., presents an optimal trajectory planning method for mobile robots using Watson-Crick pairing to find the shortest trajectory in the robot working area with the DNA sequences representing the locations of the obstacles removed during the process. The proposed algorithm is especially suitable for computing on a DNA molecular computer. In the ninth paper, "Computational Intelligence for Modeling Human Sensations in Virtual Environments," by Lee and Xu, cascade neural networks with node-decoupled extended Kalman filter training for modeling human sensations in virtual environments are proposed, with a stochastic similarity measure based on hidden Markov models to calculate the relative similarity between model-generated sensations and actual human sensations. A new input selection technique, based on independent component analysis capable of reducing the data size and selecting the stimulus information, is developed and reported. The next paper, "Intelligent Sensor Fusion in Robotic Prosthetic Eye System," by Gu et al., is concerned with the design, sensing and control of a robotic prosthetic eye that moves horizontally in synchronization with the movement of the natural eye. It discusses issues on sensor failure detection and recovery and sensor data fusion techniques using statistical methods and artificial neural network based methods. Simulation and experimental results are included to demonstrate the effectiveness of the results. The final contribution in our collection is a paper by Sun et al., entitled "A Position Control of Direct-Drive Robot Manipulators with PMAC Motors Using Enhanced Fuzzy PD Control." It presents a simple and easy-to-implement position control scheme for direct-drive robot manipulators based on enhanced fuzzy PD control, incorporating two nonlinear tracking differentiators into a conventional PD controller. Experiments on a single-link manipulator directly driven by a permanent magnet AC (PMAC) motor demonstrate the validity of the proposed approach. The Guest Editors would like to thank the contributors and reviewers of this special issue for their time and effort in making this special issue possible. They would also like to express their sincere appreciation to the JACIII editorial board, especially Profs. Kaoru and Fukuda, Editors-in-Chief and Kenta Uchino, Managing Editor, for the opportunity and help they provided for us to put together this special issue.

This special issue is devoted to robot motion planning. The main scope of this issue covers research work on mobile robotics. Motion planning is necessary when the robot determines its own actions. For the last decade, the paradigm of motion planning in mobile robotics has shifted from off-line motion planning to on-line motion planning and from planning in a static environment to planning in a time-varying environment. Recent progress of computational power has enabled this paradigm shift, since on-line motion planning and planning in time-varying environments require repeated computation based on sensory information which is always renewed. The guest editors organized this special issue in order to highlight those two new paradigms. We present two survey papers: One is a survey of on-line motion planning for a sensor-based navigation of a mobile robot, and the other is a survey of motion planning for mobile robots in a time-varying environment. Along with the survey papers, distinguished technical papers are provided in this special issue. Concerning path planning, a paper describing a case study on motion planning with teaching is included (Ogata et al). Motion planning based on Fuzzy logic is one approach, and three papers from Maeda, Ishikawa et al. and Nagata et al. also belong to this category. To offer a case study on reactive motion decision making, one paper by Ando et al. is included. A recently emerging subject is related to motion planning for cooperation of multiple mobile robots or a single robot among multiple moving obstacles. Three papers from Yoshioka et al., Ota et al., and Fujimura discuss problems on motion planning for cooperation of multiple mobile robots. One paper from Tsubouchi et al. discussed the motion planning of a single robot among multiple moving obstacles. Motion planning to select an appropriate corner cube as a landmark is addressed in the paper from Hashimoto et al. The guest editors hope that this special issue will provide the readers with a lock at some current issues and new perspectives on robot motion planning.

In many areas, such as mobile robots, video games and driverless vehicles, path planning has always attracted researchers’ attention. In the field of mobile robotics, the path planning problem is to plan one or more viable paths to the target location from the starting position within a given obstacle space. Evolutionary algorithms can effectively solve this problem. The non-dominated sorting genetic algorithm (NSGA-II) is currently recognized as one of the evolutionary algorithms with robust optimization capabilities and has solved various optimization problems. In this paper, NSGA-II is adopted to solve multi-objective path planning problems. Three objectives are introduced. Besides the usual selection, crossover and mutation operators, some practical operators are applied. Moreover, the parameters involved in the algorithm are studied. Additionally, another evolutionary algorithm and quality metrics are employed for examination. Comparison results demonstrate that non-dominated solutions obtained by the algorithm have good characteristics. Subsequently, the path corresponding to the knee point of non-dominated solutions is shown. The path is shorter, safer and smoother. This path can be adopted in the later decision-making process. Finally, the above research shows that the revised algorithm can effectively solve the multi-objective path planning problem in static environments.

PhD
This thesis is concerned with planning and acting under uncertainty in partially-observable continuous domains. In particular, it focusses on the problem of mobile robot navigation given a known map. The dominant paradigm for robot localisation is to use Bayesian estimation to maintain a probability distribution over possible robot poses. In contrast, control algorithms often base their decisions on the assumption that a single state, such as the mode of this distribution, is correct. In scenarios involving significant uncertainty, this can lead to serious control errors. It is generally agreed that the reliability of navigation in uncertain environments would be greatly improved by the ability to consider the entire distribution when acting, rather than the single most likely state. The framework adopted in this thesis for modelling navigation problems mathematically is the Partially Observable Markov Decision Process (POMDP). An exact solution to a POMDP problem provides the optimal balance between reward-seeking behaviour and information-seeking behaviour, in the presence of sensor and actuation noise. Unfortunately, previous exact and approximate solution methods have had difficulty scaling to real applications. The contribution of this thesis is the formulation of an approach to planning in the space of continuous parameterised approximations to probability distributions. Theoretical and practical results are presented which show that, when compared with similar methods from the literature, this approach is capable of scaling to larger and more realistic problems. In order to apply the solution algorithm to real-world problems, a number of novel improvements are proposed. Specifically, Monte Carlo methods are employed to estimate distributions over future parameterised beliefs, improving planning accuracy without a loss of efficiency. Conditional independence assumptions are exploited to simplify the problem, reducing computational requirements. Scalability is further increased by focussing computation on likely beliefs, using metric indexing structures for efficient function approximation. Local online planning is incorporated to assist global offline planning, allowing the precision of the latter to be decreased without adversely affecting solution quality. Finally, the algorithm is implemented and demonstrated during real-time control of a mobile robot in a challenging navigation task. We argue that this task is substantially more challenging and realistic than previous problems to which POMDP solution methods have been applied. Results show that POMDP planning, which considers the evolution of the entire probability distribution over robot poses, produces significantly more robust behaviour when compared with a heuristic planner which considers only the most likely states and outcomes.

Dans un contexte industriel aéronautique où les problématiques de sécurité constituent un facteur différentiateur clé, l’objectif de cette thèse est de répondre à la problématique ambitieuse de la réduction des accidents de type opérationnel. Les travaux de recherche s’inscrivent dans le domaine des systèmes d’alarmes pour l’évitement de collision qui ne font pas une analyse approfondie des solutions d’évitement par rapport à la situation de danger. En effet, les situations d’urgence en vol ne bénéficient pas à ce jour d’une représentation et d’un guide des solutions associées formels. Bien que certains systèmes d’assistance existent et qu’une partie de la connaissance associée aux situations d’urgence ait pu être identifiée, la génération dynamique d’une séquence de manœuvres sous fortes contraintes de temps et dans un environnement non connu à l’avance représente une voie d’exploration nouvelle. Afin de répondre à cette question et de rendre objective la notion de danger, les travaux de recherche présentés dans cette thèse mettent en confrontation la capacité d’évolution d’un aéronef dans son environnement immédiat avec une enveloppe physique devenant contraignante. Afin de mesurer ce danger, les travaux de recherche ont conduit à construire un module de trajectoires capable d’explorer l’espace en 3D. Cela a permis de tirer des enseignements en terme de flexibilité des manœuvres d’évitement possibles à l’approche du sol. De plus l’elicitation des connaissances des pilotes et des experts d’Airbus Helicopters (ancien Eurocopter) mis en situation d’urgence dans le cas d’accidents reconstitués en simulation a conduit à un ensemble de paramètres pour l’utilisation de la méthode multicritère PROMETHEE II dans le processus de prise de décision relatif au choix de la meilleure trajectoire d’évitement et par conséquent à la génération d’alarmes anti-collision
In the aeronautics industrial context, the issues related to the safety constitute a highly differentiating factor. This PhD thesis addresses the challenge of operational type accident reduction. The research works are positioned and considered within the context of existing alerting equipments for collision avoidance, who don’t report a thorough analysis of the avoidance manoeuvres with respect to a possible threat. Indeed, in-flight emergency situations are various and do not all have a formal representation of escape procedures to fall back on. Much of operational accident scenarios are related to human mistakes. Even if systems providing assistance already exist, the dynamic generation of a sequence of manoeuvres under high constraints in an unknown environment remain a news research axis, and a key development perspective. In order to address this problematic and make the notion of danger objective, the research works presented in this thesis confront the capabilities of evolution of an aircraft in its immediate environment with possible physical constraints. For that purpose, the study has conducted to generate a module for trajectory generation in the 3D space frame, capable of partitioning and exploring the space ahead and around the aircraft. This has allowed to draw conclusions in terms of flexibility of escape manoeuvres on approach to the terrain. Besides, the elicitation of the Airbus Helicopters (former Eurocopter) experts knowledge put in emergency situations, for reconstituted accident scenarios in simulation, have permitted to derive a certain number of criteria and rules for parametrising the multicriteria method PROMETHEE II in the process for the relative decision-making of the best avoidance trajectory solution. This has given clues for the generation of new alerting rules to prevent the collisions

Developments in sensor technology and sensory input processing algorithms have enabled the use of mobile robots in real-world domains. As they are increasingly deployed to interact with humans in our homes and offices, robots need the ability to operate autonomously based on sensory cues and high-level feedback from non-expert human participants. Towards this objective, this chapter describes an integrated framework that jointly addresses the learning, adaptation, and interaction challenges associated with robust human-robot interaction in real-world application domains. The novel probabilistic framework consists of: (a) a bootstrap learning algorithm that enables a robot to learn layered graphical models of environmental objects and adapt to unforeseen dynamic changes; (b) a hierarchical planning algorithm based on partially observable Markov decision processes (POMDPs) that enables the robot to reliably and efficiently tailor learning, sensing, and processing to the task at hand; and (c) an augmented reinforcement learning algorithm that enables the robot to acquire limited high-level feedback from non-expert human participants, and merge human feedback with the information extracted from sensory cues. Instances of these algorithms are implemented and fully evaluated on mobile robots and in simulated domains using vision as the primary source of information in conjunction with range data and simplistic verbal inputs. Furthermore, a strategy is outlined to integrate these components to achieve robust human-robot interaction in real-world application domains.