Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

LLA-MPC: Fast Adaptive Control for Autonomous Racing

Maitham F. AL-Sunni, Hassan Almubarak, Katherine Horng, John M. Dolan

TL;DR

LLA-MPC introduces a learning-free, real-time adaptive MPC for autonomous racing by employing a large bank of parametric, physics-based models and a look-back mechanism to select the best-fitting model from recent history. The selected model is then used in a look-ahead MPC to plan trajectories, while real-time friction estimation from tire parameters informs adaptive speed profiles. The approach demonstrates rapid adaptation to both gradual and abrupt changes in tire–surface interactions, outperforming state-of-the-art on-track methods in simulation and high-fidelity CARLA scenarios, with favorable computation overhead. The framework offers practical benefits for high-speed racing across multi-surface tracks, eliminating the need for extensive pre-race data collection and enabling robust, safe, and fast operation in dynamic environments.

Abstract

We present Look-Back and Look-Ahead Adaptive Model Predictive Control (LLA-MPC), a real-time adaptive control framework for autonomous racing that addresses the challenge of rapidly changing tire-surface interactions. Unlike existing approaches requiring substantial data collection or offline training, LLA-MPC employs a model bank for immediate adaptation without a learning period. It integrates two key mechanisms: a look-back window that evaluates recent vehicle behavior to select the most accurate model and a look-ahead horizon that optimizes trajectory planning based on the identified dynamics. The selected model and estimated friction coefficient are then incorporated into a trajectory planner to optimize reference paths in real-time. Experiments across diverse racing scenarios demonstrate that LLA-MPC outperforms state-of-the-art methods in adaptation speed and handling, even during sudden friction transitions. Its learning-free, computationally efficient design enables rapid adaptation, making it ideal for high-speed autonomous racing in multi-surface environments.

LLA-MPC: Fast Adaptive Control for Autonomous Racing

TL;DR

LLA-MPC introduces a learning-free, real-time adaptive MPC for autonomous racing by employing a large bank of parametric, physics-based models and a look-back mechanism to select the best-fitting model from recent history. The selected model is then used in a look-ahead MPC to plan trajectories, while real-time friction estimation from tire parameters informs adaptive speed profiles. The approach demonstrates rapid adaptation to both gradual and abrupt changes in tire–surface interactions, outperforming state-of-the-art on-track methods in simulation and high-fidelity CARLA scenarios, with favorable computation overhead. The framework offers practical benefits for high-speed racing across multi-surface tracks, eliminating the need for extensive pre-race data collection and enabling robust, safe, and fast operation in dynamic environments.

Abstract

We present Look-Back and Look-Ahead Adaptive Model Predictive Control (LLA-MPC), a real-time adaptive control framework for autonomous racing that addresses the challenge of rapidly changing tire-surface interactions. Unlike existing approaches requiring substantial data collection or offline training, LLA-MPC employs a model bank for immediate adaptation without a learning period. It integrates two key mechanisms: a look-back window that evaluates recent vehicle behavior to select the most accurate model and a look-ahead horizon that optimizes trajectory planning based on the identified dynamics. The selected model and estimated friction coefficient are then incorporated into a trajectory planner to optimize reference paths in real-time. Experiments across diverse racing scenarios demonstrate that LLA-MPC outperforms state-of-the-art methods in adaptation speed and handling, even during sudden friction transitions. Its learning-free, computationally efficient design enables rapid adaptation, making it ideal for high-speed autonomous racing in multi-surface environments.

Paper Structure

This paper contains 17 sections, 12 equations, 7 figures, 3 tables, 1 algorithm.

Table of Contents

  1. Introduction
  2. Look-Back and Look-Ahead Adaptive MPC
  3. Problem Setting
  4. Methodology
  5. Look-Back
  6. Look-Ahead
  7. Application: Autonomous Racing
  8. Vehicle Model and Parameters
  9. Racing Line Generation and Adaptive Planning
  10. Results
  11. Numerical Simulations
  12. Gradual friction decay (Experiment 1)
  13. Sudden Friction Drop (Experiment 2)
  14. Early Sudden Friction Drop (Experiment 3)
  15. CARLA Simulations
  16. ...and 2 more sections

Figures (7)

  • Figure 1: Illustration of the Look-Back and Look-Ahead Adaptive MPC (LLA-MPC) concept: The look-back window (blue) leverages past data for adaptation, while the look-ahead horizon (gold) optimizes future trajectory planning.
  • Figure 2: Impact of model bank size ($N$) on LLA-MPC performance. A larger $N$ reduces the MPC cost, improving adaptation and control.
  • Figure 3: Impact of size of the look-back window ($W$) when $N=20000$. A balanced $W$ reduces the MPC cost.
  • Figure 4: Comparison of friction coefficient ($\mu$) estimations over time for the ETHZ track.
  • Figure 5: (Experiment 1) Comparison of racing trajectories on the ETHZMobil track under progressive friction decay.
  • ...and 2 more figures

Theorems & Definitions (2)

  • Remark 1
  • Remark 2