Intelligent Control Navigation Emerging on Multiple Mobile Robots Applying Social Wound Treatment

Abstract

In robotics, learning new tasks is a complex solving problem. This learning depends on the environment, the robot configuration, the difficulty of the problem task, even the prior knowledge. Reinforcement learning has been widely employed for learning from scratch and policy search; however, it is very time-consuming. Multi-robots, as collaborative learners, have been proposed to improve the speed of learning in robotics. In this paper, we propose a collaborative intelligent control navigation strategy in robots, including a social wound treatment approach, such that robots can jointly learn how to avoid obstacles and move freely around the environment. This collective learning about social treatment aims to detect unexpected or inefficient behaviors of the robots, allowing them to redirect the right tasks with more agility, as observed in some animals. Experimental results over a multiple homogeneous robot system simulation validated our proposal.