Multi-agent system

With the rapid development of the ability of a single agent, cooperation between multiple agents becomes possible. Compared with the single-robot system, collaborative multi-robot intelligence dramatically improves the ability. However, the constrained computing and communication resources in the actual scene become the major obstacle to the application of collaborative intelligence.

In order to solve the above problems, our group considers the real-world non-ideal factors in multi-robot systems. In terms of perception, we study the map representation method with high data efficiency, and thus we reduce the communication resources consumption. In terms of decision-making, we study the distributed path-planning method under different map representations, and thus we enable the multi-robot exploration when the communication is unstable and the node drops. In terms of system, we study the multi-robot middleware to achieve automatic scheduling of communication and computing, and thus we can build the remotable multi-robot test field for fast verification.