Multi-Agent Path Planning for Level Set Estimation Using B-Splines and Differential Flatness
Grant Stagg, Cameron K. Peterson
Abstract
In this paper, we present a decentralized multi-agent path planning algorithm for level set estimation (LSE) and environmental monitoring missions. The planned paths are parameterized using B-splines and optimized using a novel objective function designed for LSE path planning that accounts for the exploration/exploitation trade-off while allowing the use of a gradient-based optimizer. We use the differential flatness property of the unicycle model to formulate constraints for our path optimization that ensure planned paths are kinematically feasible. We also employ a block coordinate ascent (BCA) algorithm that enables multi-agent coordination in exploring the environment. Finally, we present simulation and hardware results validating our approach.