Learning Based MPC for Autonomous Driving Using a Low Dimensional Residual Model

TL;DR

Autonomous driving MPC is highly sensitive to inaccuracies in vehicle dynamics. The paper presents a learning-based MPC that decomposes the nominal single-track dynamics into invariable and variable parts and augments it with a low-dimensional Gaussian Process residual, using features $\mathbf{z}=[\alpha_f,\alpha_r, T]^T$ and a Squared Exponential kernel to learn deviations. Three physical constraints define a valid feature region, ensuring safe and feasible residual corrections integrated into a Model Predictive Contouring Control (MPCC) framework for racing scenarios. Validation through Carsim simulations and real-vehicle experiments shows notable improvements in state prediction accuracy and substantial reductions in lap times, demonstrating robust performance across operating conditions.

Abstract

In this paper, a learning based Model Predictive Control (MPC) using a low dimensional residual model is proposed for autonomous driving. One of the critical challenge in autonomous driving is the complexity of vehicle dynamics, which impedes the formulation of accurate vehicle model. Inaccurate vehicle model can significantly impact the performance of MPC controller. To address this issue, this paper decomposes the nominal vehicle model into invariable and variable elements. The accuracy of invariable component is ensured by calibration, while the deviations in the variable elements are learned by a low-dimensional residual model. The features of residual model are selected as the physical variables most correlated with nominal model errors. Physical constraints among these features are formulated to explicitly define the valid region within the feature space. The formulated model and constraints are incorporated into the MPC framework and validated through both simulation and real vehicle experiments. The results indicate that the proposed method significantly enhances the model accuracy and controller performance.

Figures (11)

• Figure 1: The learning based MPC framework.
• Figure 2: Vehicle dynamics model.
• Figure 3: Handling diagram of vehicle.
• Figure 4: Valid region within the feature space.
• Figure 5: Racing trajectory of lap 1 and lap 6.