Establishing and maintaining wireless communication coverage among multiple mobile robots using a radial basis network controller trained via reinforcement learning

For a wirelessly-connected multi-robot system operating in a realistic environment, the wireless communication condition among mobile robots is generally unstable and fluctuating due to the signal loss, attenuation, multi-path fading and shadowing. This paper presents a decentralized control strategy, using the technique of reinforcement learning artificial neural network, to learn and approach a desired wireless communication coverage in a realistic environment for a team of collaborative mobile robots. A reinforcement learning neural network, based on the radial-basis function, is designed for each robot to learn the control law of maintaining the wireless link quality in a target environment and applied to the multi-robot deployment process to form communication coverage. The learning process of a robot is carried out through consecutive interactions between the controller and environment to establish the relationship between the wireless link quality and robot motion decision. In several environments simulated with the probabilistic log-distance path loss model, the simulation results show that the proposed reinforcement learning neural network based control approach leads to a desired and reliable multi-robot wireless communication coverage.