A multi-robot self-deployment method based on Particle Swarm Optimization

In this paper, we present a distributed algorithm solution for multi-robot deployment problem. We hope that the robots can complete deployment task by simple rules and sensors, relying on the few precondition. So, there is not pre-described model of the environment which is precondition of task allocation. The robots do not communicate between each other in order to cut the communication cost. The robot can be designed easily because it has not the ability of locomotion and distinguishing robot from obstacle. We assume that the robots have the sense of out-diffusion and want to keep a certain distance with neighbors. They determine motion direction and speed by sensing whether they are close to or leave their neighbors. When they believe that they reach a certain distance with neighbors, they will stop moving in order to save energy. In our method, each robot computes the distance and the variable quantity of the distance with neighbors to gain motion direction and speed, only relying on sensing the signal-strength which neighbors broadcast. In order to have a fast convergence speed, we introduce the model of Particle Swarm Optimization (PSO) and use the distances with neighbors instead of global position to reduce this precondition. The robot will stop moving when its activity radius is smaller than a constant during a certain period. Simulation results on Player/Stage show our algorithm can get good performance, only relying on sensing the signal-strength which neighbors broadcast in different unknown environments.