A distributed approach to load balance for multi-robot task allocation

In this paper we develop a distributed scheme for multiple robots engaged in task allocation. Vision-based control is extensively used in robotics, however large data volume may delay the response of robots and lead to the mismatch of actuators and sensors. We build a cascade information retrieval model and the signal vector based sensing delivery mapping to maintain a tight coupling between sensing and acting. And we propose a novel mathematical model which overcomes the problem of the existing model which has no extendibility for multi-robot task allocation. The new Markov model is composed of a set of rate equations and uses the theory of eigen-values in order to augment the number of task types and get the analytical solution of the model. In the physical experiments the results verify that the multi-robot system can split itself according to the proportions of various tasks in the environment, and show the effectiveness of distributed method for desired load balance.