Hybrid algorithms of multi-agent control of mobile robots

Abstract

Multi-agent robot system (MARS) represents a group of autonomous robots, which corporately solve common task in real time in dynamic environment. The paper describes methods of multi-agent control of MARS and neurocontrol of mobile robots-agents. To solve the problem of MARS control in real time, it is proposed to combine methods of artificial intelligence (AI) and neurocontrol. Multi-agent control system has a hierarchical structure with parallel processing features. It includes strategic and tactical levels of control. At the strategic control level a number of problems is solved: common MARS task decomposition into series of local sub-tasks of robots-agents, optimal sub-task distribution between robots-agents, global environment modeling and avoidance of robot collisions. At the tactical control level problems such as control of robots-agents, local obstacle modeling, optimal path planning for robots-agents, programming of movement of agents along planned paths, adaptive control of robot actuators, are solved.