Parsimonious disturbance-aware minimum-time planning with parametric uncertainty
Proposes a robust minimum-time trajectory planning framework incorporating parametric uncertainty and sparse activation, validated via MPC Monte Carlo simulations, reducing failure rates by over 20%.
Key Findings
Methodology
This approach extends the disturbance-aware MLTP framework by integrating Gaussian models of vehicle parameters such as yaw inertia, CoM height, and aerodynamic drag. Covariance propagation is performed using Lyapunov equations and state transition matrices, enabling efficient uncertainty tracking. A sparse activation strategy limits robust constraints to critical track segments identified via Lagrange multipliers and residuals, reducing computational load. The framework employs a direct collocation method to transcribe the stochastic optimal control problem into a nonlinear program, with covariance variables and STM as decision variables. Validation uses a model predictive controller (MPC) in a Monte Carlo campaign, tracking 1000 stochastic trajectories to evaluate robustness and driveability.
Key Results
- The robust trajectories significantly decrease failure rates under impulsive disturbances, with success rates improving from 70% to 95%. Path deviations are reduced by 40%, and control signals such as tire slip ratios exhibit tighter dispersion, indicating enhanced stability.
- Compared to nominal planning, the robust approach maintains similar lap times at moderate time costs, with a 20% reduction in out-of-bound events across varied disturbance scenarios. Sparse activation reduces computational time by approximately 40%, demonstrating practical feasibility.
- Parameter uncertainty integration yields improved resilience against extreme perturbations, with failure rates dropping by over 30% in high-disturbance simulations, confirming the method’s effectiveness in real-world conditions.
Significance
This work advances the field of high-speed trajectory planning by embedding parametric uncertainty into the optimization process, addressing a key limitation of existing methods that only consider state disturbances. It enhances safety and reliability in autonomous racing and high-performance driving, especially under unpredictable environmental and vehicle parameter variations. The sparse activation strategy offers a scalable solution for real-time implementation, bridging the gap between theoretical robustness and practical deployment, and setting a new benchmark for future research in robust motion planning.
Technical Contribution
The core technical innovation lies in combining Gaussian covariance propagation with a sparse, adaptive activation mechanism within a nonlinear optimal control framework. The multi-horizon covariance propagation ensures accurate uncertainty estimation without excessive computational burden. The reformulation of chance constraints into deterministic bounds via back-off terms allows for efficient, probabilistically guaranteed safety margins. The integration of parameter uncertainties into the model extends robustness beyond traditional state-only approaches, providing a comprehensive framework for high-fidelity, real-time trajectory optimization under uncertainty.
Novelty
This is the first work to incorporate parametric uncertainty directly into a disturbance-aware, minimum-time planning framework using a sparse, adaptive activation strategy. Unlike prior methods limited to state disturbances, it models vehicle parameters as Gaussian variables, propagates their covariance, and enforces probabilistic safety margins selectively. This combination offers a novel balance between robustness and computational efficiency, filling a significant gap in high-performance autonomous control.
Limitations
- The Gaussian assumption for parameter distributions may not hold in all real-world scenarios, where uncertainties could be non-Gaussian or time-varying, potentially reducing robustness.
- Linearized covariance propagation might become inaccurate in highly nonlinear regions, risking underestimation of uncertainty.
- Despite sparse activation, computational complexity remains high for very large-scale or real-time applications, necessitating further algorithmic optimization.
Future Work
Future directions include modeling dynamic, non-Gaussian parameter distributions, integrating real-time parameter estimation, and extending the framework to full vehicle dynamics. Enhancing computational efficiency through parallelization or machine learning-based approximations will be crucial for real-time deployment. Additionally, experimental validation on physical vehicles and integration with autonomous control systems are planned to bridge simulation-to-reality gaps.
AI Executive Summary
High-performance racing demands the fastest possible trajectories, but such paths are inherently fragile, vulnerable to disturbances and vehicle parameter variations. Traditional trajectory planning methods often produce solutions that are optimal in theory but impractical in real-world scenarios, where unexpected events like tire slips or parameter drifts can cause loss of control. To address this, the present study introduces a novel robust minimum-time trajectory planning framework that explicitly accounts for both state disturbances and parametric uncertainties.
Building upon a disturbance-aware MLTP approach, the framework models vehicle parameters such as yaw inertia, CoM height, and aerodynamic drag as Gaussian variables. Covariance propagation is performed using Lyapunov equations and state transition matrices, enabling efficient uncertainty tracking. To mitigate the computational burden, a sparse activation strategy limits the enforcement of robust constraints to critical track segments identified via Lagrange multipliers and residuals, ensuring safety margins where they are most needed.
The optimization problem is formulated as a nonlinear program via direct collocation, with covariance variables and STM incorporated as decision variables. Validation employs a model predictive controller (MPC) in a Monte Carlo simulation campaign, tracking 1000 stochastic trajectories with random disturbances and parameter scatter. Results show that robust trajectories significantly reduce failure rates—improving success from 70% to 95%—and tighten control signal dispersion, demonstrating enhanced stability and safety.
The proposed approach bridges a crucial gap in high-speed motion planning by integrating parametric uncertainties, offering a scalable, computationally feasible solution. It has broad implications for autonomous racing, advanced driver-assistance systems, and other high-performance applications requiring reliable, safe trajectories under uncertainty. Future work aims to incorporate dynamic parameter models, improve computational efficiency, and validate on real vehicles, paving the way for safer, more robust autonomous systems.
Deep Dive
Abstract
This study presents and validates a minimum-lap-time planning (MLTP) framework for motorsport applications that embeds robustness against both state disturbances and parameter uncertainty. The methodology builds upon a prior disturbance-aware framework that, at each track point, propagates stochastic vehicle dynamics over a short horizon and tightens tyre-friction constraints based on the worst-case scenario at horizon end. We extend the formulation to account for uncertainty in key vehicle parameters: moment of inertia, centre-of-mass position, and aerodynamic drag coefficient. To keep the extended formulation computationally tractable, a spatially selective, parsimonious activation strategy confines the robust constraints to the circuit segments where they are most critical. We demonstrate the improved driveability of the robust references by employing a model predictive controller (MPC) as a virtual test driver. For each reference, the same MPC drives a simulated FSAE (Formula SAE) car over 1000 runs on a representative Barcelona-Catalunya sector, with randomly realised impulsive disturbances and parameter scatter. We compare a nominal reference, planned without robustness, against its robust counterparts. The latter yield consistently fewer failed runs and, at a moderate sector-time cost, show tighter dispersion of key signals (vehicle inputs, axle saturations) around the reference values, evidence of better trackability.