Coordination between autonomous robots

In a multirobot system, each robot can choose between two basic behaviors: obstacle avoidance and conflict resolution behaviors. A local supervisor (LS) is used to switch between the two behaviors. Coordination between robots is necessary to avoid collision: each robot communicates to the nearest robot one angular datum used for concerted trajectory modification. In our implementation, the parameters for the LS and the conflict resolution algorithm are encoded into an integer string, and a micro-GA ( t~GA) is used for optimization. © 1997 Elsevier Science Inc. K E Y W O R D S : multirobot system, fixed-obstacle avoidance, mobile-robot avoidance, cooperation, fuzzy logic, lxGA, hierarchical control, knowledge exploitation. 1. P R E S E N T A T I O N In many industrial situations, mobile robots are used to t ranspor t products f rom one place to another , for security or cleaning tasks. These tasks can be achieved by n o n a u t o n o m o u s robots, with trajectories fixed once and material ized by a bur ied cable or a floor painted strip. This system is very efficient but suffers f rom a total lack of flexibility: • every change in trajectory requires expensive t ransformations, • unexpected obstacles on a t rajectory can cause collisions. Progress in e m b e d d e d computers makes possible the design of aut o n o m o u s robots, capable o f moving in known or unknown envi ronments comprising both fixed and mobile obstacles (o ther a u t o n o m o u s robots or human-dr iven vehicles). Such an a u t o n o m o u s robot, in a mul t i robot enviAddress correspondence to Prof. P.-Y. Glorennec, INSA, Dept. Informatique, 20, ave. des Buttes de Coesmes, 35043 Rennes Cedex, France. Received September 1, 1996; accepted December 1, 1996. International Journal of Approximate Reasoning 1997; 17:433 446 © 1997 Elsevier Science Inc. All rights reserved. 0888-613X/97/$17.00 655 Avenue of the Americas, New York, NY 10010 PII S0888-613X(97)00004-2 434 Pierre-Ives Glorennec ronment has to carry out different tasks: • avoiding fixed obstacles situated on the robot 's trajectory and whose position is not known, • reaching a goal, • cooperating with other robots for concerted modifications of high-risk trajectories, • abiding by an avoidance protocol, compatible with the corresponding human behavior (some sort of common human-robot "highway code"). To these basic tasks, we can add two constraints for a realistic hardware implementation: • minimum information transfer between the robots, • real-time learning methods. For some problems previously mentioned, different solutions exist, e.g. for fixed-obstacle avoidance [1, 10, 11, 14, 4], cooperat ion between robots [18, 6, 19], or on-line learning [20, 7]. in this paper we extend some previous results o fuzzy-logic-based reactive navigation for autonomous robots, and we consider the management of a fleet of robots. The leading strand is the exploitation of knowledge at every stage of the design process. This paper is organized as follows. In Section 2, we present our motivations and choices. In Section 3, we present the fixed and mobile-obstacle avoidance strategies and the local supervisor. A micro-GA (/.~GA) is used in Section 4 for optimization of the system parameters . Finally, we present some results and we conclude. 2. MOTIVATIONS AND CHOICES