mpcc
A Model Predictive Contouring Controller (MPCC) for autonomous racing, enabling high-speed path following and obstacle avoidance.
What is mpcc?
MPCC is a Model Predictive Contouring Controller developed for autonomous racing, enabling vehicles to follow reference paths at high speeds while minimizing errors and avoiding obstacles. It formulates path following as an optimization problem that maximizes progress along the path, penalizes contouring and lag errors, and enforces track constraints. The controller uses nonlinear vehicle dynamics and efficient solvers to achieve real-time performance in simulation and experimental platforms.
Researchers and engineers working on autonomous vehicle control, particularly in racing or high-performance path-following applications. It is also suitable for academic institutions and labs focusing on model predictive control, robotics, and optimization-based control systems.
MPCC provides a specialized, open-source implementation of a contouring controller tailored for racing, with support for obstacle avoidance and realistic vehicle dynamics. Its dual C++ and Matlab implementations offer flexibility for simulation and deployment, backed by research from ETH Zurich and proven in real-world autonomous racing systems like the AMZ Driverless car.
Overview
Model Predictive Contouring Controller (MPCC) for Autonomous Racing
Use Cases
Best For
- Simulating autonomous racing controllers in research environments
- Implementing high-speed path following for miniature or full-scale race cars
- Studying optimization-based control with nonlinear vehicle dynamics
- Developing obstacle avoidance algorithms for autonomous vehicles
- Experimenting with model predictive control in real-time applications
- Teaching advanced control concepts in academic courses or labs
Not Ideal For
- Commercial projects needing simple, out-of-the-box controllers without deep optimization knowledge
- Real-time systems with ultra-low latency requirements beyond the C++ implementation's optimizations
- Applications requiring extensive documentation or active community support beyond academic papers
- Teams looking for plug-and-play obstacle avoidance in both C++ and Matlab versions
Pros & Cons
Pros
Incorporates a nonlinear bicycle model with Magic Formula tire models, accurately simulating high-speed racing behavior as detailed in the formulation section.
Provides both C++ and Matlab code, allowing for research simulation in Matlab and deployment-ready implementations in C++, based on the README's separate instructions.
Supports efficient solvers like hpipm, Yalmip, CVX, and QuadProg, enabling real-time performance by approximating the nonlinear problem as a quadratic program.
Backed by ETH Zurich research and used in real-world systems like AMZ Driverless, with multiple cited papers validating its effectiveness in autonomous racing.
Cons
Obstacle avoidance is only available in the Matlab implementation, not in C++, limiting deployment options for full autonomous systems as noted in the README.
Requires installation of specific solvers and tools, with separate instructions for C++ and Matlab, making it challenging for users unfamiliar with optimization software.
Focused primarily on autonomous racing with a specific vehicle model, so adapting it to other autonomous driving scenarios requires significant modification and expertise.
Frequently Asked Questions
Related Projects
pinocchioA fast and flexible implementation of Rigid Body Dynamics algorithms and their analytical derivatives
EGO-PlannerEGO-Planner is an ESDF-free gradient-based local planner designed for quadrotor navigation. It significantly reduces computation time compared to state-of-the-art methods by avoiding the computationally expensive Euclidean Signed Distance Field (ESDF) construction, enabling real-time performance with total planning times around 1ms. ## Key Features - **ESDF-Free Planning** — Eliminates the need to compute Euclidean Signed Distance Fields, drastically reducing computational overhead. - **Lightweight Gradient-Based Optimization** — Uses a gradient-based approach for efficient local trajectory generation. - **GPU/CPU Versatility** — Offers both GPU and CPU versions of its local sensing module for depth image generation or pointcloud processing. - **Fast Computation** — Achieves planning times of approximately 1ms, suitable for real-time drone control. - **Simulation-Ready** — Includes a lightweight quadrotor simulator and supports integration with sensors like Intel RealSense for hardware testing. ## Philosophy EGO-Planner prioritizes computational efficiency and real-time performance by removing the ESDF construction bottleneck, making advanced local planning accessible for resource-constrained aerial robotics applications.
casADiCasADi is a symbolic framework for numeric optimization implementing automatic differentiation in forward and reverse modes on sparse matrix-valued computational graphs. It supports self-contained C-code generation and interfaces state-of-the-art codes such as SUNDIALS, IPOPT etc. It can be used from C++, Python or Matlab/Octave.
omplThe Open Motion Planning Library (OMPL)
Found a gem we're missing?
Open-Awesome is built by the community, for the community. Submit a project, suggest an awesome list, or help improve the catalog on GitHub.