Experimental Testing of Adaptive Route Planning for Autonomous Rovers

To date, route planning for planetary rovers has been performed at the ground station. However, with improvements in autonomous navigation enabling rovers to make their way between targets independently, future rover missions could incorporate onboard route planning. This would enable targets that are identified while the rover is moving around to be added to the plan, if possible. In this work the NEAT algorithm is applied to this route planning problem, which enables a rover to create its entire route plan autonomously and modify it should new targets be identified or in response to unexpected events. Resources, such as battery charge and memory, are also managed in combination with the science planning. In a previous work this algorithm was successfully applied to the problem of planning imaging operations for an Earth observation satellite. NEAT (Near-optimal Evolutionary Autonomous Task-manager) is a generic planning algorithm that recursively optimises future plans, taking into account circumstances and new requests, whilst enforcing temporal and resource constraints. A model autonomous rover has been constructed for experimental testing of this autonomous route planning. The rover is able to navigate between 16 targets in a test area and is operated autonomously using the NEAT planning system. In test scenarios targets are identified as sites of interest for the rover to visit at random times so that NEAT must continually evolve the rover’s plans while it is moving between targets. Results will be presented showing efficient planning and adaptation in this real-time experimental environment.