LLA-MPC: Fast Adaptive Control for Autonomous Racing

This repository contains the code for the paper "LLA-MPC: Fast Adaptive Control for Autonomous Racing" that was accepted in IROS 2025

BibTeX

If you find this repo useful, please consider citing our paper:

@article{alsunni2025llampc,
  title={LLA-MPC: Fast Adaptive Control for Autonomous Racing},
  author={AL-Sunni, Maitham F and Almubarak, Hassan and Horng, Katherine and Dolan, John M},
  journal={arXiv preprint arXiv:2505.19512},
  year={2025}
}

Summary

Video Presentation (Coming soon!)

Numerical Simulations

ETHZ Track

ETHZMobil Track

Experiment 1

ETHZ Track

LLA-MPC

APACRace

Oracle

ETHZMobil Track

LLA-MPC

APACRace

Oracle

Experiment 2

ETHZ Track

LLA-MPC

APACRace

Oracle

ETHZMobil Track

LLA-MPC

APACRace

Oracle

Experiment 3

ETHZ Track

LLA-MPC

APACRace

Oracle

ETHZMobil Track

LLA-MPC

APACRace

Oracle

CARLA Simulations

We perform an experiments on CARLA's Town #7:

The experiment we perform is identical to experiment #1 for the ETHZ and the ETHZMobil tracks, where we let the friction coefficient decay linearly. Here, we compare our method against a controller that does not account for the friction change to demonstrate LLA-MPC's adaptation ability.

With LLA-MPC:

With nominal model only (no adaptation):

How to run the Code

Installation

We recommend creating a new conda environment:

conda create --name llampc python=3.6
conda activate llampc

Then install LLA-MPC:

git clone https://github.com/DRIVE-LAB-CMU/LLA-MPC.git
cd LLA-MPC/
pip install -e .

LLA-MPC

Run LLA-MPC controller with the generated offline library of racelines as reference as follows:-

cd mpc
python run_nmpc_orca_llampc_rt.py

APACRace

Run APACRace controller with the generated offline library of racelines as reference as follows:-

cd mpc
python run_nmpc_orca_apacrace.py

Oracle

Run the Oracle:-

cd mpc
python run_nmpc_oracle.py