Operating autonomous vehicles at the absolute limits of handling requires precise, real-time identification of highly non-linear tire dynamics. However, traditional online optimization methods suffer from "cold-start" initialization failures and struggle to model high-frequency transient dynamics. To address these bottlenecks, this paper proposes a novel vision-augmented, iterative system identification framework. First, a lightweight CNN (MobileNetV3) translates visual road textures into a continuous heuristic friction prior, providing a robust "warm-start" for parameter optimization. Next, a S4 model captures complex temporal dynamic residuals, circumventing the memory and latency limitations of traditional MLPs and RNNs. Finally, a derivative-free Nelder-Mead algorithm iteratively extracts physically interpretable Pacejka tire parameters via a hybrid virtual simulation. Co-simulation in CarSim demonstrates that the lightweight vision backbone reduces friction estimation error by 76.1 using 85 fewer FLOPs, accelerating cold-start convergence by 71.4. Furthermore, the S4-augmented framework improves parameter extraction accuracy and decreases lateral force RMSE by over 60 by effectively capturing complex vehicle dynamics, demonstrating superior performance compared to conventional neural architectures.
Figures & tables
Fig. 1 : Schematic representation of the dynamic single-track model.
Fig. 2 : Architecture of the vision-based friction estimation module. The system utilizes a MobileNetV3-Small backbone across four hierarchical stages for efficient feature extraction. The Inverted Residual Block topology incorporates a SE attention mechanism for texture recalibration. The Probabilistic Friction Mapper executes a classification-to-regression mapping, weighting the Softmax distribution p by the physical basis vector B to derive the continuous expected friction coefficient μ^ .
Fig. 3 : Architecture of the S4 model. The diagram illustrates the transition from HiPPO -initialized continuous state-space to a discrete convolutional representation, enabling parallelized computation of vehicle dynamic residuals.
Fig. 4 : Iterative framework for tire model identification.
Fig. 5 : Visualization of the data acquisition phase in the CarSim environment.
Metric
MobileNetV3 (ours)
ResNet-18 [ 9 ]
EfficientNet [ 19 ]
Params (M)
1.52
11.18
4.01
FLOPs (G)
0.12
3.65
0.83
RMSE
0.102
0.427
0.310
Latency (ms)
5.91
5.33
8.35
TABLE I : Quantitative comparison of candidate vision backbones.
Metric
No Vision
Vision
Init. Friction μ0
0.51
0.51
Vision Friction μ^
–
0.80
Fyf/Fzf RMSE
0.219
0.076
Fyr/Fzr RMSE
0.073
0.046
Iter.
7
2
TABLE II : Effect of Vision Prior on Identification Initialization
Fig. 6 : Comparison of identified tire force curves with and without the vision prior. The vision prior significantly improves the initial parameter accuracy.
Architecture
Fyf/Fzf RMSE
Fyr/Fzr RMSE
Time [s]
MLP
0.234
0.071
5.3
RNN
0.097
0.084
12.9
S4 (Ours)
0.051
0.033
12.2
TABLE III : Performance Metrics of Different Residual Architectures for Parameter Extraction
Fig. 7 : Comparison of identified Pacejka tire force curves using different neural architectures. The proposed S4 model achieves the highest identification accuracy and best alignment with the ground truth.
Accurate modeling of nonlinear vehicle dynamics is essential for high-speed autonomous racing, where controllers operate at the handling limits. Model-based methods are interpretable but rely on simplifying assumptions, while purely learned models capture nonlinearities yet often lack physical consistency, generalization, and adaptability to changing operating conditions. This paper presents DynSSM, a physics-aware state-space memory framework that combines learned temporal representations with a structured vehicle dynamics model. The proposed approach integrates state-space sequence modeling and recurrent encoders to capture long- and short-term dynamic behavior. It simultaneously adapts tire and vehicle-dynamics parameters within bounded ranges to preserve physical plausibility. A residual correction mechanism compensates for remaining unmodeled dynamics while preserving the underlying physics-based structure. DynSSM is evaluated on both simulated small-scale racing data and real-world full-scale autonomous Indy racecar data. When evaluated on an unseen real-world track, DynSSM reduces one-step prediction RMSE by up to 27.2% in longitudinal velocity, 73.9% in lateral velocity, and 88.8% in yaw rate compared with the state-of-the-art (\sota{}) baselines. Component-wise ablation studies demonstrate the importance of temporal memory, parameter adaptation, and residual correction for predictive performance and robustness. Further, closed-loop simulations using nonlinear model predictive control demonstrate that DynSSM remains feasible across the evaluated tracks and achieves up to a 13.2% reduction in one-lap completion time compared with \sota{} baselines. These results indicate that combining temporal memory, bounded physics-guided parameter adaptation, and residual correction provides an accurate, interpretable, and control-ready dynamics model for autonomous racing.
Musabbir Ahmed Arrafi, Malik Ali, Nicholas M. Stiffler +1
Real-time estimation of vehicle-tire-road friction is critical for allowing autonomous race cars to safely and effectively operate at their physical limits. Traditional approaches to measure tire grip often depend on costly, specialized sensors that require custom installation, limiting scalability and deployment. In this work, we introduce RAGE, a novel real-time estimator that simultaneously infers the vehicle velocity, slip angles of the tires and the lateral forces that act on them, using only standard sensors, such as IMUs and RADARs, which are commonly available on most of modern autonomous platforms. We validate our approach through both high-fidelity simulations and real-world experiments conducted on the EAV-24 autonomous race car, demonstrating the accuracy and effectiveness of our method in estimating the vehicle lateral dynamics.
In this work, we present RAGE-XY, an extended version of RAGE, a real-time estimation framework that simultaneously infers vehicle velocity, tire slip angles, and the forces acting on the vehicle using only standard onboard sensors such as IMUs and RADARs. Compared to the original formulation, the proposed method incorporates an online RADAR calibration module, improving the accuracy of lateral velocity estimation in the presence of sensor misalignment. Furthermore, we extend the underlying vehicle model from a single-track approximation to a tricycle model, enabling the estimation of rear longitudinal tire forces in addition to lateral dynamics. We validate the proposed approach through both high-fidelity simulations and real-world experiments conducted on the EAV-24 autonomous race car, demonstrating improved accuracy and robustness in estimating both lateral and longitudinal vehicle dynamics.