Adaptive control of a multi-robot system for transportation of large-sized objects based on reinforcement learning

This paper describes models and algorithms for intelligent control of a group of autonomous mobile robots, which perform large-sized object transportation in a complex environment. The proposed models allow the multi-robot system to reach its target position while avoiding obstacles and maintaining object orientation with coordinated motion of several robots. We use neural based Q-learning to provide robots adaptability to unknown environments. The inputs of the learning subsystem are 2d-map data collected during system operation and target misalignments of multi-robot system. The primary output is a control decision with a maximum value of estimated efficiency. Experimental results presented in the paper fully confirm the reliability of the proposed approach.