Evolutionary-based Coordination of Multi-Robot Systems with Dynamic Constraints

Large-scale manufacturing facilities and power plants comprise numerous subsystems that require consistent monitoring and inspection to ensure reliable and secure operations. The use of multiple robotic systems is an expanding, robust, and cost-effective solution for such operations. To enable automated infrastructure monitoring, there are many challenges, such as path planning and task allocation for multiple robots and dynamic constraints. We formulate such challenges as a multiple Traveling Salesman Problem (mTSP) and propose an efficient task assignment solution that uses an evolutionary algorithm considering priority, energy, and time constraints. Our solution incorporates: a digital twin of a monitored environment, A* and artificial potential field methods for path planning and navigation, and human-in-the- loop prioritization. We validate the proposed method with state- of-the-art simulations in Gazebo/ROS for unmanned ground vehicles (UGVs) in cluttered environments.