diffsqp

diffsqp is a batchable Sequential Quadratic Programming (SQP) solver built with PyTorch. It is designed to solve trajectory optimization and optimal control problems, natively supporting both forward and inverse dynamics formulations.

Features

• PyTorch-Native: Leverages PyTorch for tensor operations and GPU acceleration, allowing you to solve batches of optimization problems in parallel.
• Modular Dynamics & Costs: Includes built-in support for classical underactuated robotics systems like the Acrobot and CartPole. Easily specify LQR costs and underactuation constraints.
• Forward & Inverse Dynamics: Configure the solver to optimize over states and controls directly, or use inverse dynamics constraints depending on your problem setup.
• Visualization: Built-in animators for systems like CartPole and Acrobot to visualize optimized state trajectories.

Prerequisites

This project requires Python 3.13 or newer.

We recommend using uv, an extremely fast Python package and project manager written in Rust, to manage dependencies and virtual environments.

Installation

0. Install uv (if you haven't already):

   curl -LsSf [https://astral.sh/uv/install.sh](https://astral.sh/uv/install.sh) | sh

   (For Windows or alternative installation methods, refer to the uv documentation.)

1. Clone the repository:

    git clone https://github.com/hucebot/diffsqp
    cd diffsqp

2. Install dependencies and setup the environment: Because the project uses a pyproject.toml, you can use uv sync to automatically create a virtual environment and install all required dependencies (like torch, cvxpylayers, and matplotlib):

    uv sync

Prerequisites

The package includes an entry-point example script at examples/example.py. This script dynamically loads problem parameters, initializes the dynamics, builds the SQP problem, and animates the resulting trajectory.

You run the examples by passing a YAML configuration file to the script.

To run an example using uv run (which automatically executes the command inside the isolated virtual environment):

 # Example: Run the Acrobot solver with forward dynamics
 uv run python examples/example.py -c examples/configs/config_acrobot_forward.yaml

 # Example: Run the CartPole solver with inverse dynamics
 uv run python examples/example.py -c examples/configs/config_cartpole_inverse.yaml

The script will output the solver logs to the console and, upon completion, render a visualization of the optimized trajectory.

Project Structure

• src/diffsqp/: Core library containing the SQP solver, constraints, costs, dynamics definitions, and types.

• examples/: Demonstration scripts and YAML configuration files to run test problems.