leggedrobotics/ocs2

Health v2 · maintenance only

Adoption not part of the score

1501 stars · 329 forks observed · 2026-08-28

What it is AI-extracted, prompt v1, taxonomy v1, 2026-08-30, confidence not recorded

OCS2 is a C++ toolbox for formulating and solving nonlinear optimal control problems, with an emphasis on real-time Model Predictive Control (MPC) for robotics. It provides multiple solvers (SLQ, iLQR, SQP, SLP, IPM), switched-system support, automatic differentiation, URDF-based robot modeling tools, and ROS interfaces.

Use cases

• implement real-time model predictive control for a legged robot
• solve nonlinear optimal control problems with switched-system dynamics
• build MPC controllers from a URDF robot model with self-collision avoidance
• generate derivatives of system dynamics and costs via automatic differentiation
• deploy MPC on robots with limited onboard computation using ROS
• run cartpole, quadrotor, or mobile manipulator MPC examples

When to choose

• you need real-time nonlinear MPC for robotic platforms
• your problem involves switched systems with mode schedules and jump maps
• you want URDF-based modeling, Pinocchio integration, and ROS deployment tooling
• you need a choice of solvers (SLQ, iLQR, SQP, SLP, IPM) with constraint handling

When to avoid

• you need a lightweight general-purpose optimizer without robotics focus
• your project does not use C++ or ROS and you only need simple scripting
• you require a mature Python-first control library instead

Facets

library · maturity active

simulation machine-learning robotics math developer-tools robotics simulation autonomous-vehicles cpp python optimal-control mpc model-predictive-control switched-systems slq ilqr sqp automatic-differentiation urdf pinocchio legged-robots algorithms linux ros

Member repositories

For agents

markdown · JSON · MCP: product_card(name="leggedrobotics/ocs2")