Introduction to the Special Issue on Robotics and Computer Vision
Departamento de Ciência da Computação
Universidade Federal de Minas Gerais
31270-010 Belo Horizonte, MG - Brazil
Instituto de Automação
Centro Tecnológico para Informática
13089-500 Campinas, SP - Brazil
1. Robot Systems and Computer Science
Why should the Journal of the Brazilian Computer Society publish a special issue on Robotics and Computer Vision? Possibly one of the best answers to this question is implicit in one of the definitions of Robotics, as suggested by Michael Brady:" Robotics is the intelligent connection of perception to action" [2].
The fields of Robotics and Computer Vision address the development of the foundational capabilities required by semi-autonomous and autonomous robotic systems executing complex tasks in the real world. These capabilities include: sensor data processing andinterpretation; fusion of multi-origin, multi-sensor data; building spatial and temporal models of the external world and reasoning about these models; planning, executing and monitoring specific sensor, deliberation and motor actions; navigation in known and unknown environments, which encompasses trajectory planning and execution, and self-localization; self-monitoring, self-diagnosis, and error recovery; mission and task level planning and plan execution; human-machine interfaces; and appropriate robot control architectures, among others [4, 5].
Due to its multifaceted requirements, building substantially autonomous robot systems provides both a stringent testbed for new concepts and approaches, and a context for focussing and integrating new technologies in the various areas mentioned above. Since robot systems need to cope with complex, dynamically changing, and often unpredictable real-world environments, the component technologies are required to exhibit levels of performance that go far beyond the competence usually displayed by intelligent hardware or software agents that perform well in toy domains or in highly specialized applications.
Robot systems draw technologies from a variety of fields, and particularly from the fields of Computer Science, Electrical and Electronics Engineering, and Mechanical Engineering. For the computer scientist, the complexity of the interaction between a robotic agent and the real world poses two types of challenges: firstly, the design, implementation and deployment of efficient and robust robot systems provides an application and testing ground for a vast array of technologies developed within the field of Computer Science. Secondly, Robotics and Computer Vision pose new challenges for computer scientists in such diverse areas as real-time systems, geometric modelling and reasoning, decision-making under uncertainty, estimation and information theory, computer graphics and virtual reality, man-machine interfaces, and many others.
The purpose of this special issue is, therefore, two-fold: firstly, to introduce those members of the Computer Science community unfamiliar with recent development in Robotics and Computer Vision to some of the issues and challenges that occur in building significantly autonomous robotic systems; and secondly, to convey to the readers of the Journal of the Brazilian Computer Society a representative cross-section of current research work in Robotics and Computer Vision.
2. Towards Significantly Autonomous Robotic Systems
Robots have long been a subject of fascination, both to scientists and to the popular mind. Efforts to develop mechanisms with some degree of mobility and autonomy can be traced back to antiquity, and became particularly evident in the Renaissance. It was, however, the dual developments of feedback control and automated computation over the last century that have provided the basis for modern Robotics.
Much of the early work in Robotics, performed during the 1950s and 1960s, focussed on robot manipulators and their industrial use. Their early successes have ensured that research in manipulation, grasping and related sensing technologies has continued and expanded up to the present day. In contrast, although mobile robot vehicles started to be developed during the late 1960s, their role during the 1970s and early 1980s was by and large restricted to research platforms. Over the last fifteen years, however, mobile robot systems have matured and many new areas of application of robot technologies have emerged. Some present-day applications include the deployment of robots in space, underwater and in the atmosphere; the use of robots in hazardous environments, with particular emphasis on mining, deep-sea operations and environmental clean-up; the application of robots in medicine and surgery, as well as in the care of elderly, handicapped or hospitalized individuals; and initial experiments in the deployment of robots in home and business applications.
What the key research areas in Robotics and Computer Vision? One possible characterization decomposes robots in the following conceptual subsystems: Perception, Deliberation, Navigation, Actuation, and Control [1, 3]. Robot Perception covers the areas of sensor processing, sensor fusion, building of spatial and temporal models of the world, and active vision, which deals with explicit planning of sensor data acquisition strategies. Deliberation deals with planning and decision-making, bringing together the mission to be executed by the robot, the internal world model the system has, and the robot's own capabilities; it typically has to take into account conditions of uncertainty and missing or conflicting information. Navigation covers trajectory planning and execution in known, partially known or unknown environments, which can additionally be static or dynamic in nature, as well as robot self-localization or position estimation. Actuation covers the problems that are posed by the execution, monitoring and low-level control of manipulation or locomotion actions. Finally, Control addresses issues related to the design of flexible, efficient and robust robot control architectures.
3. Contents of the Special Issue
The papers in this special issue represent a broad spectrum of research in the areas of Robotics and Computer Vision, and touch on many of the themes outlined above. Trajectory planning has been a long-standing area of research in Robotics. While general solutions to the movement of polygonal robots in polygonal environments have been developed, the area of motion planning continues to be very active. The paper by Fraichard addresses what is still an open area of research, namely motion planning in the presence of moving obstacles, while the paper by Kubota, Arakawa and Fukuda discusses an alternative to exact methods by introducing concepts from computer learning to reduce the computational effort faced in trajectory planning.
Grasping deals with object acquisition and fine manipulation by the end effector of a robot manipulator. As such, it is on the one hand closely related to trajectory planning, but poses on the other hand significant problems of its own, due to the fact that issues such as geometric and positional uncertainty, surface friction, interaction between the end effector and the object being grasped, and others have to be taken into account. The paper by Rosa and Okada presents a novel approach to object acquisition using a scrollic gripper mechanism.
The papers by Rillo, Barros and Bianchi, and by Waldmann and Bispo are representative of the significant interest that approaches to active and purposive vision have spurred in recent years. Active vision sees sensor data acquisition as a process that should be explicitly planned and controlled by the robot, while purposive vision focusses on the interconnection between specific tasks to be executed by a robotic agent and the sensory requirements of that task.
Control architectures for significantly autonomous robot systems have to face a number of challenges: they have to integrate processes that operate at radically different speeds, going from high-speed, low-level control loops to slow, high-level planning and decision-making modules; they have to ensure that the system is able to interact in real-time with their environment; and they have to provide robust mechanisms for internal system diagnosis, execution monitoring, and error recovery. Medeiros provides a tutorial overview of different types of control architectures that have been proposed for autonomous mobile robots, and discusses their relative merits.The paper by Xavier and Schneebeli also addresses the control architecture issue, and presents a concurrent object-oriented approach to one particular class of approaches to robot system control, namely, behavior-based architectures.
The paper by Elfes et al presents a novel application of robotic and computer vision technologies to the development of a robotic aerial vehicle for environmental research and monitoring applications. This is an integrative project where many of the Robotic and Computer Vision research themes mentioned above are brought together.
Acknowledgments
The editors would like to express their thanks to all those individuals who have made this special issue possible. We are particularly indebted to Prof. Cláudia Bauzer Medeiros, Editor-in-Chief of the JBCS, for her enthusiastic support and patience; to Dr. Marcel Bergerman (IA/CTI) for his crucial involvement and help in the review process; to Ms. Sueli A. Silva (IA/CTI), for her administrative and logistics support; to the reviewers of the submitted papers, whose dedicated work was essential to guarantee the quality of the issue; and, above all, to the authors of this special issue.
- [1] R. Bajcsy and M. F. M. Campos, Active and Exploratory Perception, Journal of Computer Vision, Graphics and Image Processing - Image Understanding, 56(1):31-40, 1992.
- [2] M. Brady, Artificial Intelligence and Robotics, Artificial Intelligence, 26:79-121, 1985.
- [3] A. Elfes, Robot Navigation: Integrating Perception, Environmental Constraints and Task Execution Within a Probabilistic Framework. In Reasoning With Uncertainty in Robotics, Dorst, L., van Lambalgen, M., Voorbraak, F. (eds.), Lecture Notes in Artificial Intelligence, vol. 1093, Springer-Verlag, Berlin, 1996.
- [4] S. S. Iyengar and A. Elfes (eds.), A Tutorial on Autonomous Mobile Robots. Vol. 1: Perception, Mapping and Navigation. Vol. 2: Planning, Architectures, and Control. IEEE Press, Los Alamitos, CA, 1991.
- [5] P. J. McKerrow, Introduction to Robotics. Addison-Wesley, New York, 1991.
Publication Dates
-
Publication in this collection
08 Oct 1998 -
Date of issue
Apr 1998
