The autonomous vehicle industry is fast developing with many companies vying to release the first fully autonomous passenger vehicle. However, the advancement of the field also drives forward related areas such as autonomous racing. This paper aims to investigate how to optimally control an autonomous race car such that its progression across a previously unknown race track is maximised. This paper is inspired by the Formula Student race. The paper investigates using a Model Predictive Controller, MPC, to convert current vehicle position to a set of control inputs through the use of a vehicle model. Two different models have been investigated, the first focusing purely on the kinematics of the system, whilst the second encodes tyre forces for a dynamic approach. Further, the planning horizon of the MPC has been investigated to see how changes to it affect the performance of the controller and thus obtain insight into the trade-off between model complexity and planning horizon. Finally the limitations of both models was investigated. A modular real time implementation has been developed with the help of The CT Toolbox, a fast, C++ based, MPC implementation control toolbox. Small track segments and a full track have been used to evaluate the performance between the two models. Results showed that the computational difference from a more complex model is not a major limiting factor for the update rate of the MPC. However, the performance of the two models was significantly different with the dynamic based MPC far outperforming the kinematic. Constraining lateral accelerations of the car is critical when designing the controller as failure to do so will allow the controller to plan unfeasible paths. The dynamic model, through encoding the tyre forces, accomplishes this successfully completing the lap. However, the kinematic model, which assumes zero slip, fails as the speed gets high with sharp corners. It was also found that too short planning horizons do not allow sufficient information to flow to the controller ...
Design Trade-offs in Optimisation Based Trajectory Planning for Autonomous Racing ; Optimal planering för en Autonom Racer-bil
2019-01-01
Hochschulschrift
Elektronische Ressource
Englisch
DDC: | 629 |
BASE | 2020
|BASE | 2022
|