Optimal Race Strategy for Solar Team Twente Using Approximate Dynamic Programming

Every two years Solar Team Twente competes in the Bridgestone World Solar Challenge. Their goal is to finish in first place and to reach this goal, an optimal race strategy is needed. Both the uncertainty in the weather forecasts and the risk the Solar Team wants to take should be taken into account when obtaining the strategy. In order to achieve this, the race is modelled as a Markov Decision Process. The uncertainty in the weather forecasts is included by using Affine Kernel Dressing. Also, an exponential utility function is used to incorporate risk sensitivity into the model. The model is implemented using Approximate Dynamic Programming and tested for a simplified case.