Decentralized cooperation protocol for autonomous robotic agents

This paper proposes a decentralized cooperation protocol for an object-sorting task in a distributed robotic system. The multiple agents are based on a homogeneous agent architecture that consists of search, motion, and communication modules coordinated through a global state. Advantages of the system architecture are simplicity, intra-agent distributed control, no explicit inter-agent control and flexibility. The protocol encourages the agents to help each other in order to facilitate overall task achievement. Simulation results showed that 1) The protocol is stable and reliable under a spectrum of workload. 2) The execution time increases linearly with the number of objects. 3) Increasing the number of agents can significantly decrease the waiting time and improve the performance.