Welcome to the project page of “A Hierarchical Approach for Strategic Motion Planning in Autonomous Racing.”

This work is related to a European Control Conference 2023 paper by Rudolf Reiter, Jasper Hoffmann, Joschka Boedecker, and Moritz Diehl. Moreover, this work was accepted to be presented as a poster at the International Conference on Machine Learning 2024.

An approach is presented for safe trajectory planning, where a strategic task related to autonomous racing is learned within a simulation environment. The following videos show evaluations of the trained HILEPP-II policy in different scenarios related to autonomous racing tasks. Opponent vehicles are simulated with obstacle-avoiding MPC formulations for autonomous racing, which do not explicitly consider strategic driving (PPP)

Scenario 1: Blocking

The ego vehicle starts first and has to block three stronger opponents. “Stronger” refers to parameters such as vehicle mass, acceleration limits, and available braking force.

Scenario 2: Overtake

The ego vehicle starts last and has to overtake three weaker opponents. “Weaker” refers to parameters such as vehicle mass, acceleration limits, and available braking force.

Scenario 3: Mixed

The ego vehicle starts between a stronger following vehicle and a weaker leading vehicle. It has to safely overtake and block the following opponent to achieve a good rank for as long as possible.