Tom0Brien / tinympc

A lightweight implementation of MPC and NMPC in C++ using Eigen3

tinympc is a lightweight C++ library that implements general linear and non-linear model predictive control (MPC) algorithms using Eigen3, osqp-eigen and NLopt.

Features

• Uses efficient sparse matrix computations via Eigen3.
• Allows users to define custom system dynamics, constraints, and cost functions.
• Robust error handling and solver feedback.
• Warm start capability for faster real-time optimization.

Usage

After initializing the MPC object with system matrices, constraints, and desired weights, the following functions are available:

Function	Description
compute_control()	Solves the MPC optimization problem and computes the optimal control input for the current state.
get_control()	Get computed control input.
propagate_system()	Propagates the system dynamics forward using the optimal control input obtained in the previous step.
get_states()	Get current states.

Dependencies

• Eigen3: Required for matrix computations.
• OsqpEigen: Interface for the OSQP solver used to solve the quadratic program.
• NLopt. Non-linear optimization solver.

Installation and Setup

git clone https://github.com/Tom0Brien/tinympc.git && cd tinympc
mkdir build && cd build
cmake ..
make
sudo make install # copies files in the include folder to /usr/local/include*

Examples

See examples folder for a number of examples.

Below is a simple example for a 2-state system that illustrates how to use the tinympc::MPC class:

---

Example: Using the tinympc::MPC class for a 2-state system

1. Include the necessary headers:

#include <tinympc/mpc.hpp>

2. Define your system parameters:

// Dimensions
const int n = 2;   // Number of states
const int m = 1;   // Number of inputs
const int N = 50;  // Prediction horizon

// Define system matrices for the 2-state system
double Ts = 0.1;
Eigen::Matrix<double, n, n> A;
A << 1, Ts, Ts, 1;

Eigen::Matrix<double, n, m> B;
B << 0, Ts;

// Define cost matrices
Eigen::Matrix<double, n, n> Q = 10 * Eigen::Matrix<double, n, n>::Identity();
Eigen::Matrix<double, m, m> R = Eigen::Matrix<double, m, m>::Identity();

// Define state and input bounds
Eigen::Matrix<double, n, 1> x_min, x_max;
x_min << -10, -10;
x_max << 10, 10;
Eigen::Matrix<double, m, 1> u_min, u_max;
u_min << -6;
u_max << 6;

// Define initial state and reference
Eigen::Matrix<double, n, 1> x0, x_ref;
x0 << 0.1, 0.1;
x_ref << 1, 0;

3. Instantiate the MPC class:

tinympc::MPC<double, 2, 1, N> mpc(A, B, Q, R, x_min, x_max, u_min, u_max, x0, x_ref);

4. Compute and apply control for several steps:

for (int i = 0; i < 20; i++) {  // Simulate for 20 steps
    mpc.compute_control();
    mpc.propagate_system();
}

Contact

For any inquiries, bug reports, or feature requests, please open an issue on our GitHub repository.