Inverse Dynamics Trajectory Optimization

Implements the contact-implicit trajectory optimization algorithm described in

Inverse Dynamics Trajectory Optimization for Contact-Implicit Model Predictive Control by Vince Kurtz, Alejandro Castro, Aykut Özgün Önol, and Hai Lin. https://arxiv.org/abs/2309.01813.

Dependencies

This software is built on Drake. You do not need a separate Drake installation, but all the requirements for building Drake from source apply. Most notably, that includes Bazel and a C++17 compiler.

The easiest way to install these dependencies is with Drake's install_prereqs.sh script:

git clone https://github.com/RobotLocomotion/drake.git
cd drake
sudo ./setup/ubuntu/install_prereqs.sh

For Mac OS, replace the last line with ./setup/mac/install_prereqs.sh.

Installation

Install the dependencies (see above).

Clone this repository:

git clone https://github.com/ToyotaResearchInstitute/idto
cd idto

Compile the package:

bazel build //...

Python Bindings

A limited subset of solver functionality is available in python via pybind11. To use the python bindings, first install the dependencies and compile with bazel, as described above.

Then update the python path:

export PYTHONPATH=${PYTHONPATH}:"/path/to/idto/bazel-bin"

This line can be added to .bashrc if you want to permanently update the path.

Some examples of using these bindings for open-loop trajectory optimization and MPC can be found in the python_examples folder.

Examples

The examples folder contains various examples, including those described in our paper. Run them with, e.g.,

bazel run //examples/spinner:spinner

A link will appear (e.g., http://localhost:7000), which you can use to open the Meshcat visualizer in a web browser.

Most of the examples (e.g., spinner) run a simulation with contact-implicit model predictive control. Some others (e.g., kuka) perform a single open-loop trajectory optimization.

Problem definitions, solver parameters, whether to run MPC, etc. are set in YAML config files, e.g., spinner.yaml. Here are some common options:

mpc : {true, false} choose whether or not to run MPC. If this is set to true, Meshcat will show and record a simulation where IDTO is used as an MPC controller.
num_threads : N sets the number of threads used for parallel derivative computations.
play_target_trajectory : {true, false} whether to play an animation of the target trajectory over Meshcat.
play_initial_guess : {true, false} whether to play an animation of the initial guess over Meshcat.
play_optimal_trajectory : {true, false} whether to play an animation of the optimal trajectory over Meshcat. This is not a simulation: the generated trajectory may or may not be dynamically feasible.

NOTE If Meshcat plays multiple things, only the last one will be recorded for playback via the dropdown menu. For example, if play_target_trajectory, play_optimal_trajectory, and mpc are all set to true, Meshcat will first play the target trajectory, followed by the open-loop solution, followed by a simulation with MPC. Only the simulation will be saved for playback.

Other Tips and Tricks

Use an existing Drake installation

By default, Bazel pulls in a copy of Drake as an external and compiles it. If you have an existing local checkout at /home/user/stuff/drake that you would like to use instead, set the environment variable

export IDTO_LOCAL_DRAKE_PATH=/home/user/stuff/drake

before building.

Run unit tests

Run a (fairly minimal) set of unit tests and lint checks with

bazel test //...

Contributing

We welcome your contributions, whether they are bug fixes, solver improvements, or new features! To make a contribution:

Open a new pull request
Make sure all the unit tests and lint checks pass
Obtain a review from a code owner (e.g., @vincekurtz, @amcastro-tri, or @aykut-tri)
Once the review is approved, we'll merge it into the main branch!

Since this is research code, we will not review to the same production-quality standards as Drake. Nonetheless, new contributions should be clean and well-documented, and unit tests are encouraged. The standard of review should be improving the health of the codebase rather than perfection.

ToyotaResearchInstitute / idto