Path planning for a Mars Rover: A prognostic based offline decision-making approach

Abstract

A rover mission consists of visiting waypoints to take scientific samples. Due to telecommunication unreliability, the rover must have an autonomous decision-making system. The path-planning problem can be separated into two stages online and offline to reduce computational cost. In this thesis, an offline prognostic decision-making (PDM) system is described. The PDM problem was formulated from an optimization point of view that, unlike previous approaches, uses a genetic algorithm (GA) to find feasible solutions. The optimization problem is mathematically developed and accounts for the battery state of charge, the number of waypoints visited, and the terrain profile. The GA decides the number of waypoints to visit, in which order the rover will visit them, and when it needs to recharge batteries. The PDM system was implemented and tested through simulations under different terrain maps and a study of the parameters sensitivity was performed. Results showed that the system can efficiently find feasible solutions in different scenarios, prevents energy consumption overload and plans battery recharge to satisfy the SoC policy. An analysis of the impact of different sources of uncertainty in the model is performed, which demonstrates that the approach taken gives a degree of slack to operate.