Cooperative Transportation by Multiple Robots with Machine Learning

This paper investigates the development of a physical multi-robot system, where a group of intelligent robots work cooperatively to transport an object to a goal location and orientation in an unknown dynamic environment. Multi-agent technology and machine learning are integrated into the same physical platform to provide innovative capabilities for carrying out the task. First, a new multi-agent architecture is developed. Second, two methods that facilitate optimized machine learning, Reinforcement learning (RL) and genetic algorithms (GA), are integrated into the decision-making agent in the architecture. An arbitrator is incorporated into a probabilistic switching scheme for selecting the optimal strategy in a given state of the task. Finally, the robot force/motion control and local modeling are integrated into the architecture to implement a physical multi-robot system. Simulation and experimental studies are carried out to demonstrate the feasibility of the system.