Rsearch On Model And Algorithm Of Fast Information Sharing Based Multi-Robot Replanning

Due to enhanced dynamics of multi-robot systems, the environment is of high uncertainty and cannot be predicted by the robots. Thus the robots cannot rely on configured plans and multi-robot replanning problem has become a big challenge. Current researches still have some flaws in dynamic environment with complex tasks and heterogeneous robots. In this thesis, an information sharing based multi-robot replanning approach is proposed and this problem is solved by providing more valuable information for replanning decision.The first problem to prevent heterogeneous robots from sharing information is inconsistent information representation. A uniform information category and representation techniques are proposed, according to the key information in multi-robot decision models. The heterogeneous information is represented in three parts: the state of robot itself, the environment state and action of robot. And the information is semantically represented in OWL.Secondly, based on the semantic information, information is shared according to the register-based and push-based heuristics, so as to provide more valuable information for replanning decision. In register-based information sharing, heuristics are proposed to estimate information requirement for the robot, according to its role and location. In push-based information sharing, the information importance is evaluated according to the change of semantic graph structure. Register-based approach could share information more accurately but with more communication cost while push-based approach has a better balance. Push-based approach could also catch hidden information requirement, which is even not known by the robot itself.A multi-robot replanning model is built to find out how information sharing contributes to replanning. According to the analysis, information sharing is proved to be helpful to reduce computational complexity and improve accuracy.Finally, a simulation for multi-UAV path planning is built to demonstrate the approach. By incorporating semantic information representation and information sharing techniques in multi-UAV path planning, the information sharing based replanning approach is proved to be effective.