Long-Horizon Informative Path Planning with Obstacles and Time Constraints

Abstract

We apply non-myopic informative path planning in a simulated river plume case study with several constraints on our agent. A cost valley philosophy is proposed to guide the agent through the field. The purpose of this path planner is to reveal the river plume front with the long-horizon while safely returning home in time. Among others, we employ RRT*, a variant of RRT (rapidly-exploring random trees), as the path planner to determine the least-cost path between locations. The distance budget from start to end destination, the obstacle constraint, and directional change are penalties, whereas the reduced variance of the field and an excursion set are the two rewards. The cost valley is then computed by superimposing those five fields. The simulation results demonstrate the efficiency of such a strategy. They show that the suggested approach balances exploitation and exploration while bearing in mind the go-home constraint.