Iterative Transportation by Mobile Robots in Unknown Environment
This research deals with an iterative transportation task (Fig.1), which is a basic and important task of multiple mobile robots. This task is demanded from industrial fields such as material transportation in factories. Considering about the real work space, it is supposed that the environment changes dynamically. In factory, for example, some changes in factory layout, robots or objects are assumed. The Motion planning of robots should autonomously realize efficient transportation coping with these changes.
Most of previous researches dealing with such dynamic environment have used reactive planning or single learning method. Those methods lead to too time-consuming convergence or difficulty of evaluation. This is because they treat the multiple problems underlying such complex problem simultaneously. By dividing a complex task into elemental problems and dealing with them separately, efficient task-execution can be realized. In this research, we divide the iterative transportation task into (1) acquisition of environmental information, (2) generation of appropriate transportation paths, and (3) acquisition of efficient robot formation. These problems are dealt with in three corresponding phases. In Environmental Exploration Phase, cooperative exploration is made using Learned Visibility Graph. In Path-Generation Phase, considering about the characteristic of iterative transportation, paths including 2-lane paths, in which two robots can cross over, are generated. And in Strategy-Making Phase, efficient robot formation is acquired by learning behaviors of each robot individually using Asymptotic Strategy-Making Method.The effectiveness of the proposed method is verified by a simulation. In this simulation, four robots transport objects in the environment including two unknown obstacles. Fig.2 shows the simulation environment and generated paths. Fig.3 indicates that robots change their formation based on the environmental changes. When a cost of handing over an bject is low, robots used the shortest paths and formed relay-type formation. When the cost is high, robots made loop-type formation to avoid handing over objects. This result shows that the proposed method realizes the motion planning which acquires efficient formation coping with dynamic environmental changes.
Keywords: multiple mobile robot system, motion planning, learning
References
Kousuke INOUE, Jun OTA, Tomokazu HIRANO, Daisuke KURABAYASHI and Tamio ARAI : “Iterative Transportation by Cooperative Mobile Robots in Unknown Environment,” Proceeding of International Symposium on Distributed Autonomous Robotic Systems (DARS’98), to be appeared, 1998.
Fig.1 Iterative Transportation Task Fig.2 Generated Paths Fig.3 Dynamic Group Formation