Implementation and Evaluation of Stanley Based Path-Following Controllers for Autonomous Vehicles with Conventional Steering

Kurzfassung

The miniAFM is an autonomous test vehicle on a 1:8 scale at DLR. The current path following controller, the geoPFC, was originally developed for an over-actuated vehicle with four independently steerable wheels. However, the miniAFM has conventional front-axle steering. This work implements three Stanley-based controllers and evaluates them against the geoPFC. The variants include the base Stanley, a predictive Stanley with a prediction horizon, and an enhanced Stanley with feedforward control via a kinematic reference model. Simulation tests across three scenarios show that the Predictive Stanley achieves the lowest lateral RMSE, followed by the base Stanley. Robustness tests with increased latency show the base Stanley to be the most robust choice without retuning, whilst the Predictive Stanley performs best after parameter tuning. On the real vehicle, the base Stanley outperforms the geoPFC in terms of lateral RMSE. The results confirm the Stanley-based controllers as promising alternatives to the geoPFC on the miniAFM. The enhanced Stanley requires a more detailed comparison by using a parameter optimization.