Flexible and Hybrid Action Selection Process for the Control of Highly Dynamic Multi-Robot Systems

1. Introduction

The domain of cooperative robotics constitutes an active research field and is currently linked to many key application areas with large importance. A variety of tasks is performed in a more reliable and flexible way by teams of robots cooperating among them and/or with humans. These tasks concern collaborative manipulation, cooperative search and transportation, ground, space and underwater exploration, surveillance and autonomous rescue operations. The scientific problems associated to multi-robot systems concern formation analysis and control, cooperative perception, multi-robot self-localization, multi-robot task coordination, architectures for cooperation and communication Cao et al. (1997), Ota (2006). Multi-robot cooperation offers advantages like acceleration of performance due to the parallel processing and coordination among robots and also robustness due to the presence and the modularity of multiple robots. Nevertheless, the coordination of multi-robot team in dynamic environments constitutes one of the fundamental problems. Coordination implies normally synchronization of robot actions and exchanges of information among the robots. The amount of synchronization and communication depends heavily on the tasks requirement, characteristics of the robot and its environment. More particularly, the control of simple and reactive group of robots in a distributed manner is a complex problem because global behaviors must emerge as a result of many local interactions Forrest (1991), (Arkin, 1998, p.105), Shen et al. (2004). The objective of our works is to control as minimalist as possible robotics entities Melhuish (2001), Adouane and Le Fort-Piat (2004a) and to use the local interactions between them to produce a form of advanced collective intelligence. More specially, the presented work concern the design and development of a control architecture exhibiting as well individual features adapted to highly dynamic multi-robot systems (a swarm of robots) than collective features favoring global goals. The architectures presented in this chapter are fully distributed and are based on reactive behaviors in order to perform complex cooperative tasks while using robots with limited sensorial and computational capabilities. The proposed process of coordination between behaviors is based on one hand, on a hierarchical coordination and on the other hand, on fusion mechanism. The goal is to select and combine appropriately elementary behaviors responses so that the robots will be able to perform complex tasks.