Data-Driven Predictive Control for Autonomous Vehicle Racing under Varying Friction
Authors: Julius Beerwerth, Julius Tischbein, Bassam Alrifaee
Organizations: Department of Aerospace Engineering, University of the Bundeswehr Munich, 85577 Neubiberg, Germany · Department of Computer Science, RWTH Aachen University, 52062 Aachen, Germany
Data-driven predictive control (DPC) predicts future outputs by combining recorded input-output trajectory data such that the combined trajectory approximately matches a measured initial trajectory and tracks a provided reference. This paper investigates to what extent and under which conditions DPC using data pooled across operating conditions retains the prediction accuracy of DPC using data from the current operating condition. In autonomous vehicle racing, the tire-road friction varies between racetrack sections and tire sets. If the initial trajectory and the recorded data originate from different frictions, prediction accuracy of DPC degrades. Ensuring the frictions match requires knowing the friction during recording and estimating the current friction explicitly. Instead, we pool data recorded at different frictions. We compare the prediction accuracy of DPC using data recorded under the current friction (matched), a different friction (mismatched) and data pooled from both frictions. We evaluate in simulation and small-scale vehicle experiments in the Cyber-Physical Mobility Lab (CPM Lab). Our results show that pooled data keep the multi-step root-mean-square prediction error within 1.5 times that of matched data, whereas mismatched data increase it by a factor of 2.2 to 3.4. To assess how this prediction accuracy carries over to control, we evaluate the tracking error from the raceline in closed loop. In simulation, DPC with pooled data tracks comparably to nonlinear model predictive control (13.7 cm versus 13.3 cm) and close to DPC with matched data in the CPM Lab, differing by 0.1 cm on the low-friction tires (1.6 cm versus 1.5 cm) and by 0.2 cm on the high-friction tires (0.9 cm versus 1.1 cm). DPC thus provides partial adaptation to varying friction, without a dedicated adaptation method.
Figures & tables
Fig. 1: Distribution of the N -step RMS prediction error for (1) pooled DPC, (2) matched DPC, and (3) mismatched DPC. Boxes show quartiles and median, whiskers the 10th to 90th percentiles, and the dashed line the median of matched DPC. Colors indicate the evaluation friction.
Tracking error [ cm ]
Prediction error [ mm ]
zone μ
zone μ
0.5
0.8
1.1
lap
0.5
0.8
1.1
lap
Pooled DPC
12.1
15.4
14.0
13.7
2.1
2.6
1.8
2.0
Matched DPC
14.5
16.0
13.0
14.4
1.5
2.6
0.7
1.5
Mismatched DPC
11.6
14.8
19.2
15.0
4.6
2.6
3.7
3.8
NMPC
17.4
12.1
9.7
13.3
–
–
–
–
TABLE I: Median closed-loop tracking and one-step prediction errors over all steps aggregated for each zone.
Fig. 2: Top: tracking error over racetrack progress for pooled DPC and NMPC. The curves show the median over 30 evaluated laps, and the shaded regions indicate the corresponding minimum and maximum errors. Middle and bottom: differences in tracking and prediction errors between pooled DPC or mismatched DPC and matched DPC. The background color indicates the track friction profile, with blue denoting low friction ( μ=0.5 ), gray intermediate friction ( μ=0.8 ), and red high friction ( μ=1.1 ).
Fig. 3: True friction (inner ring) and the friction implied by the pooled prediction (outer ring) aggregated over 10m sections of the racetrack. The thin outline marks sections on which the predictions from the data recorded at each friction differ by less than 2cm .
Low-friction tires
High-friction tires
tracking
pred.
lap
tracking
pred.
lap
[ cm ]
[ mm ]
[ s ]
[ cm ]
[ mm ]
[ s ]
Pooled DPC
1.6
1.72
9.62
0.9
1.52
10.17
Matched DPC
1.5
1.69
9.51
1.1
1.56
9.84
Mismatched DPC
3.0
2.08
9.30
0.7
1.47
10.35
TABLE II: Median tracking and one-step prediction errors over three runs, and mean lap time in small-scale vehicle experiments in the CPM Lab.
Fig. 4: Difference in tracking error between pooled DPC or mismatched DPC and matched DPC in small-scale vehicle experiments in the CPM Lab.
This paper presents Adaptive Differentiable Model Predictive Contouring Control (AD-MPCC), a framework for autonomous racing that integrates differentiable MPCC with online parameter estimation to handle varying road-surface conditions. For online parameter estimation, we leverage a parameterized Pacejka Magic Formula together with a regularized moving-horizon estimation scheme with exponentially decaying weights to capture road interactions and update parameters in real time. Furthermore, we propose a differentiable MPCC (Diff-MPCC) framework that enables optimal adjustment of objective weights based on predefined long-horizon performance costs. To implement Diff-MPCC for online objective weight adaptation, we propose a Pacejka-informed machine learning model that is trained in a supervised manner using data generated by Diff-MPCC to tune the objective weights. Simulation results demonstrate that AD-MPCC reliably ensures safety and achieves faster lap times compared to baseline controllers in both single-surface and multiple-surface scenarios.
Nam T. Nguyen, Binh Nguyen, Ahmad Amine +3
Department of Electrical and Computer Engineering, University of Central Florida, Orlando, FL 32816, USA. · Department of Electrical and Systems Engineering, University of Pennsylvania, Philadelphia, PA 19104, USA. · CTI Lab4EV, School of Electrical and Electronic Engineering, Hanoi University of Science and Technology, Hanoi 100000, Vietnam.
Autonomous race cars, such as in Formula Student Driverless, operate close to their physical handling limits. The resulting highly nonlinear vehicle behavior increases the path tracking complexity, especially on narrow tracks. Model Predictive Control (MPC) is commonly used to address this issue, a method whose performance is closely tied to the accuracy of the underlying prediction model. This paper presents a novel, real-time capable prediction model for autonomous race cars that adjusts to changing conditions by combining information from past runs and the current driving situation. Our model is divided into three consecutive submodels: a nominal Kinematic Bicycle Model, an offline Bayesian Linear Regression (BLR) model, and an online Sparse Gaussian Process Regression (SGPR) model. The proposed approach enables efficient integration of all available data without significantly increasing computational cost, ensuring high prediction accuracy and a quantitative uncertainty assessment right from the start of the run. Compared to existing approaches, an improvement in prediction accuracy of up to 57% was achieved. Further, we successfully demonstrated the practical applicability of the model within an MPC-based path tracking controller on a real Formula Student race car.
Sebastian Baader, Tamara Bergerhoff, Pascal Meißner +1
Center for Artificial Intelligence and Robotics (CAIRO); TUAS Würzburg-Schweinfurt, Germany
Friction uncertainty is a significant challenge for autonomous vehicle control, undermining feasibility and performance. This paper proposes a novel two-layer control framework to achieve vehicle motion control under large friction uncertainty by combining optimization-based planning with indirect adaptive robust control (IARC). The upper layer solves a constrained optimal control problem to generate a feasible trajectory that satisfies all the constraints in a time-optimal manner along the desired path. The lower layer, IARC, adapts the unknown parameters related to friction uncertainty and combines model compensation with robust feedback to ensure transient and steady-state tracking performance. An estimation monitor is designed to prevent poor learning when data indicate a model mismatch, while a smooth parameter reset reduces the estimated tire capacity and limits abrupt tightening of constraints. The updated parameters are fed back into the model and constraints in the planner, enabling trajectory replanning as friction conditions change. Simulation results indicate that the proposed method achieves accurate tracking of the desired trajectory and online friction adaptation under large friction uncertainty while ensuring time-optimal vehicle motion. The simulation results demos are available at https://github.com/TedZhou0603/CO-IARC.
Yupeng Zhou, I-Chia Chang, Ziran Wang +1
College of Engineering, Purdue University, West Lafayette, IN 47906, USA