Adaptive approach to regulate task distribution in swarm robotic systems

Abstract

Abstract The task allocation problem has been a challenging issue in swarm robotic systems. In this paper, we handle dynamic task allocation in the object foraging task, where multiple robots are supposed to collect various objects in parallel during a given time span. We propose a decentralized strategy of the response threshold model without any communication among robots. Here, we introduce a task selection probability function for each robot to balance the task demands with robots working on tasks. The method can produce varying tendencies to tasks by changing the response thresholds given in the task selection probability function. With this property, each robot can select its own task among all available tasks, by regulating the response threshold values. This ultimately promotes a desired task distribution in a group level and reduces the number of task changes. Our approach suggests that the response threshold of robot is updated based on what objects have been observed recently and what tasks have been done by its neighboring robots observed in the local surrounding area. We provide a convergence analysis that the system can stochastically converge to the equilibrium of the desired task distribution. The method is effective even when only local information about the environment is given to individual robots. The suggested method is tested with a simulation of multiple robots taking the foraging task, and a dynamic task allocation process is demonstrated.