Multi-Robot-Exploration

This project considers two distributed strategies for frontier-based multi-robot exploration: Non-Convex DisCoverage (NCD) and artificial potential fields (APF). We deploy simulated teams of N=4 agents to execute both algorithms to solve an exploration challenge in simulated environments. Our first algorithm is an adaptation of Non-convex DisCoverage (NCD), in which team of robots perform a local optimization to explore frontiers within their Voronoi cells. We adapt the original algorithm to use A* for path planning in non-convex regions. Our second algorithm extends traditional artificial potential field methods to include Voronoi frontier exploration, and A* for non-convex path planning. We compare the performance of NCD and APF in both uniform and an environment defined by a Gaussian density function. In both cases, artificial potential fields are able to fully explore the environment faster than Non-Convex DisCoverage, but we find that Non-Convex DisCoverage outperforms artificial potential fields at rapidly finding key features in the environment.