Affiliations: Perception and Robotics Laboratory, Department of Electrical and Computer Engineering, École Polytechnique, Montréal, Quebec, Canada, H3C 3A7, Tel.: +1 514 340-5783; Fax: +1 514 340-5802; E-mail: [email protected], [email protected], [email protected]
Correspondence:
[*]
Correspondence to: Prof. Paul Cohen.
Abstract: Several autonomous robots performing a common task not achievable by a single robot must coordinate their actions in order to succeed. A centralized control strategy may ensure this coordination, but it is often impractical due to strong restrictions on required feedback structure, inter-robot communications and tolerable time-delays. Decentralized control is shown to be an advantageous alternative, as it makes no use of explicit communication and is based only on local information about the robot position and task evolution. A two stage control structure is proposed. At a higher level, the task's specification is given along with information about the desired changes in the task space. At alower level, the assigned task evolution is ensured by an appropriate control strategy. As an application of the proposed technique, we analyze a system of two car-like robots moving a beam to a desired location. Nonholonomy of the robots, friction phenomena, and sensory feedback are all taken into account. The robustness of the proposed control structure is demonstrated by its ability to handle the obstacle avoidance problem: in order to avoid an obstacle, it is sufficient to have it detected by only one of the robots. Simulation results and our experimental setup are presented.
