Real-Time Capable Informative Path Planning for Autonomous Area Monitoring Using Unmanned Ground Vehicles

Abstract

Advancements in the field of robotics together with an increased need for surveillance around critical infrastructure have led to the possibility of utilizing autonomous agents for area monitoring. This paper presents a novel modification to an existing path planning algorithm for autonomous area monitoring, that is more optimized for real-time applications. The proposed Autonomous Surveillance Planner (ASP) discretizes an area into cells and assigns each cell a probability of there being an intruder present based on when the cell was last observed by the agent. The planner then chooses an optimal path through the area by minimizing a cost function describing the probability of not finding an intruder along a path. Since the minimization is computationally costly and scales exponentially with path length, only a short path is computed at a time. This is then repeated, using new information as it becomes available. The ASP was tested both in simulations and during field tests using unmanned ground vehicles with promising results, showcasing that it has potential to be used in a real-world application. ASP was fast enough to be used in real-time and was able to fully cover the intended area. The local planner used for low-level control was computationally costly and would benefit from more computation power or optimization.