Optimization Engine

Fast and accurate embedded nonconvex optimization

Build high-performance optimizers for next-generation robotics, autonomous vehicles, and other cyber-physical systems without hand-writing solver infrastructure.

Get Started Read the Paper

Core languageRust

Primary usesMPC, MHE, Robotics

InterfacesPython, MATLAB, C/C++, ROS, TCP

Embeddable

All numerical routines are written in Rust, making OpEn a strong fit for embedded targets where speed, determinism, and memory safety matter.

Accurate

OpEn combines fast convergence with a practical problem formulation for nonconvex optimization, including augmented Lagrangian and penalty updates.

Fast

Benchmarks and applications show sub-millisecond performance in the right settings, enabling demanding control and estimation loops.

Easy code generation

Install OpEn in Python with pip, model your optimization problem with CasADi, and generate a solver that you can run through TCP, C/C++, ROS, or Rust.

The docs in Installation and Python Interface walk through the flow end to end.

Animated overview of OpEn code generation

Design and Deploy

Formulate your problem in Python or MATLAB, generate a Rust optimizer, and consume it over TCP, C/C++, ROS, or native Rust.

Benchmarks

OpEn is built for real optimization workflows, from reproducible academic experiments to embedded deployments and hardware-in-the-loop tests.

Model Predictive Control

Model Predictive Control (MPC) is a powerful optimization-based control methodology. It has become a standard tool in control engineering because it can handle nonlinear dynamics and state or input constraints, but it also requires solving an optimization problem in real time under tight runtime limits.

When the dynamics are nonlinear or the constraints are nonconvex, implementation becomes significantly harder. This is where OpEn helps: it offers extremely fast and robust numerical optimization methods tailored for embedded applications where both speed and memory usage matter.

Moving Horizon Estimation

Moving Horizon Estimation (MHE) is the bee's knees of nonlinear estimation: it is an optimization-based estimator for constrained nonlinear systems. MHE is backed by a strong theoretical bedrock that combines Bayesian estimation and dynamic programming; however, its applicability has been hampered by the associated computational burden and has limited its use to slow or linear dynamical systems.

OpEn can unlock the huge potential of MHE and facilitate its use in robotics, automotive, aerospace and other applications with high sampling frequencies.

Well Documented

The documentation covers installation, interfaces, optimal control tutorials, and end-to-end examples for robotics and autonomous systems.

Browse the Docs

Presentation at IFAC 2020

A short introduction to what OpEn does, how it works, and how to use it in practice.

Python OCP package

Design optimal control problems and MPC in Python

OpEn comes with a Python OCP module that facilitates the design of optimal control problems in an intuitive and straightforward way. You define the key ingredients of the problem, including stage and terminal costs, dynamics, and state or input constraints.

It is a practical starting point for building nonlinear optimal control and MPC formulations directly in Python before generating an embedded optimizer.

Explore the OCP docs See the OCP workflow

State trajectories from an optimal control problem built with the Python OCP package

New in opegen 0.11

ROS2 packages

OpEn can now generate ROS2 packages directly from a parametric optimizer. The generated package includes ROS2 messages, configuration files, a launch file, and a node that exposes the solver through topics.

This makes it easy to connect optimization-based controllers, estimators, and planning modules into a modern robotics stack without writing the ROS2 wrapper code by hand.

Learn more Legacy ROS1

Bot icons created by pbig - Flaticon

ros2_config = og.config.RosConfiguration() \
    .with_package_name("my_ros_pkg") \
    .with_node_name("open_node_ros2") \
    .with_rate(10.0)

build_config = og.config.BuildConfiguration() \
    .with_build_directory("my_optimizers") \
    .with_ros2(ros2_config)

Docker image

Run OpEn in a ready-made container

OpEn ships with a Docker image that gets you straight into a working environment with Jupyter, Python, and the tooling needed to explore examples without local setup friction.

It is a convenient way to try the Python interface, browse the notebooks, and experiment with the OCP workflows in a clean, reproducible environment.

Learn more Docker Hub