locuslab/mpc.pytorch

Health v2 · maintenance only

Flags: no_releases

Adoption not part of the score

1078 stars · 164 forks observed · 2026-08-28

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

mpc.pytorch is a fast and differentiable model predictive control (MPC) solver implemented as a PyTorch library. It solves receding-horizon optimal control problems using the box-DDP heuristic and supports gradients flowing through the solver for learning-based applications.

Use cases

• solve model predictive control problems inside PyTorch models
• train neural networks end-to-end through an MPC solver
• implement differentiable MPC for model-based reinforcement learning
• control simulated robots with receding-horizon optimization
• reproduce research from the differentiable MPC and OptNet papers

When to choose

• you need MPC as a differentiable layer in PyTorch
• you are doing model-based RL research with control priors
• you want a GPU-friendly, batched MPC solver in Python

When to avoid

• you need a production-grade, actively developed MPC solver
• you work outside PyTorch or need non-Python integration
• you need robust support for hard nonlinear constraints beyond box-DDP

Facets

library · maturity maintenance

machine-learning simulation math machine-learning robotics reinforcement-learning python model-predictive-control differentiable-optimization pytorch optimal-control box-ddp algorithms

Member repositories

For agents

markdown · JSON · MCP: product_card(name="locuslab/mpc.pytorch")