Behavioral Model Architectures: A New Way Of Doing Real-Time Planning In Intelligent Robots

摘要

Traditional hierarchical robot control systems, although well suited for manufacturing applications, appear to be inefficient for innovative applications, such as mobile robots. The research we present aims to the development of a new architecture, designed to overcome actual limitations. The control system was named BARCS (Behavioral Architecture Robot Control System). It is composed of several modules, that exchange information through a blackboard. The original point is that the functions of the modules were selected according to a behavioral rather than a functional decomposition model. Therefore, the system includes, among other, purpose, strategy, movement, sensor handling and safety modules. Both the hardware structure and the logical decomposition allow a great freedom in the design of each module and of the connections between modules, that have to be as flexible and efficient as possible. In order to obtain an "intelligent" behavior, a mixture of traditional programming, artificial intelligence techniques and fuzzy logic are used, according to the needs of each moddle. The approach is particularly interesting because the robot can be quite easily "specialized", i.e. it can be given behaviors and problem solving strategies that suit some applications better than other. Another interesting aspect of the proposed architecture is that sensor information handling and fusion can be dynamically tailored to the robot's situation, thus eliminating all time-consuming useless processing.