Counting 2,899 Big Data & Machine Learning Frameworks, Toolsets, and Examples...
Suggestion? Feedback? Tweet @stkim1

Last Commit
Jun. 23, 2018
Oct. 15, 2016

NIPS2018: AI for prosthetics

This repository contains software required for participation in the NIPS 2018 Challenge: AI for prosthetics. See more details about the challenge here. See full documentation of our reinforcement learning environment here. In this document we will give very basic steps to get you set up for the challenge!

In this competition, you are tasked with developing a controller to enable a physiologically-based human model with a prosthetic leg to walk and run. You are provided with a human musculoskeletal model, a physics-based simulation environment where you can synthesize physically and physiologically accurate motion, and datasets of normal gait kinematics. You are scored based on how well your agent adapts to requested velocity vector changing in real time.

AI for prosthetics

To model physics and biomechanics we use OpenSim - a biomechanical physics environment for musculoskeletal simulations.

What's new compared to NIPS 2017: Learning to run?

We took into account comments from the last challenge and there are several changes:

  • You can use experimental data (to greatly speed up learning process)
  • We released the 3rd dimensions (the model can fall sideways)
  • We added a prosthetic leg -- the goal is to solve a medical challenge on modeling how walking will change after getting a prosthesis. Your work can speed up design, prototying, or tuning prosthetics!

You haven't heard of NIPS 2017: Learning to run? Watch this video!

HUMAN environment

Getting started

Anaconda is required to run our simulations. Anaconda will create a virtual environment with all the necessary libraries, to avoid conflicts with libraries in your operating system. You can get anaconda from here In the following instructions we assume that Anaconda is successfully installed.

We support Windows, Linux, and Mac OSX (all in 64-bit). To install our simulator, you first need to create a conda environment with the OpenSim package.

On Windows, open a command prompt and type:

conda create -n opensim-rl -c kidzik opensim python=3.6.1
activate opensim-rl

On Linux/OSX, run:

conda create -n opensim-rl -c kidzik opensim python=3.6.1
source activate opensim-rl

These commands will create a virtual environment on your computer with the necessary simulation libraries installed. Next, you need to install our python reinforcement learning environment. Type (on all platforms):

conda install -c conda-forge lapack git
pip install git+

If the command python -c "import opensim" runs smoothly, you are done! Otherwise, please refer to our FAQ section.

Note that source activate opensim-rl activates the anaconda virtual environment. You need to type it every time you open a new terminal.

Basic usage

To execute 200 iterations of the simulation enter the python interpreter and run the following:

from osim.env import ProstheticsEnv

env = ProstheticsEnv(visualize=True)
observation = env.reset()
for i in range(200):
    observation, reward, done, info = env.step(env.action_space.sample())

Random walk

The function env.action_space.sample() returns a random vector for muscle activations, so, in this example, muscles are activated randomly (red indicates an active muscle and blue an inactive muscle). Clearly with this technique we won't go too far.

Your goal is to construct a controller, i.e. a function from the state space (current positions, velocities and accelerations of joints) to action space (muscle excitations), that will enable to model to travel as far as possible in a fixed amount of time. Suppose you trained a neural network mapping observations (the current state of the model) to actions (muscle excitations), i.e. you have a function action = my_controller(observation), then

# ...
total_reward = 0.0
for i in range(200):
    # make a step given by the controller and record the state and the reward
    observation, reward, done, info = env.step(my_controller(observation))
    total_reward += reward
    if done:

# Your reward is
print("Total reward %f" % total_reward)


Your task is to build a function f which takes the current state observation (a dictionary describing the current state) and returns the muscle excitations action (19-dimensional vector) maximizing the total reward. The trial ends either if the pelvis of the model falls below 0.6 meters or if you reach 1000 iterations (corresponding to 10 seconds in the virtual environment).

Round 1

The objective is to run at a constant speed of 3 meters per second. The total reward is 9 * s - p * p where s is the number of steps before reaching one of the stop criteria and p is the absolute difference between horizonal velocity and 3.

Round 2

In the second round the task is also to follow a requested velocity vector. However, in this round the vector will change in time and it will be a random process. We will provide the distribution of this process in mid-July.

You can test your model on your local machine. For submission, you will need to interact with the remote environment: crowdAI sends you the current observation and you need to send back the action you take in the given state.

Training your models

We suggest you start from reviewing solutions from the last year.

You can find many tutorials, frameworks and lessons on-line. We particularly recommend:

Tutorials & Courses on Reinforcement Learning:

Frameworks and implementations of algorithms:

OpenSim and Biomechanics:

This list is by no means exhaustive. If you find some resources particularly well-fit for this tutorial, please let us know!


Assuming your controller is trained and is represented as a function my_controller(observation) returning an action you can submit it to crowdAI through interaction with an environment there:

import opensim as osim
from osim.http.client import Client
from osim.env import ProstheticsEnv

# Settings
remote_base = ""
crowdai_token = "[YOUR_CROWD_AI_TOKEN_HERE]"

client = Client(remote_base)

# Create environment
observation = client.env_create(crowdai_token)

# my_controller = ... (for example the one trained in keras_rl)

while True:
    [observation, reward, done, info] = client.env_step(my_controller(observation), True)
    if done:
        observation = client.env_reset()
        if not observation:


In the place of [YOUR_CROWD_AI_TOKEN_HERE] put your token from the profile page from website.


In order to avoid overfitting to the training environment, the top participants will be asked to resubmit their solutions in the second round of the challenge. The final ranking will be based on the results from the second round.

Additional rules:

  • Organizers reserve the right to modify challenge rules as required.

Details of the environment

In order to create an environment, use:

    from osim.env import ProstheticsEnv

    env = ProstheticsEnv(visualize = True)


  • visualize - turn the visualizer on and off

reset(project = True)

Restart the enivironment.

The function returns:

  • observation - a vector (if project = True) or a dictionary describing the state of muscles, joints, and bodies in the biomechanical system.

step(action, project = True)

Make one iteration of the simulation.

  • action - a list of continuous values in [0,1] corresponding to excitation of muscles. The length of the vector is expected to be: 22 and 18 for 3D and 2D models without the prosthesis; 19 and 15 with a prosthesis.

The function returns:

  • observation - a vector (if project = True) or a dictionary describing the state of muscles, joints, and bodies in the biomechanical system.

  • reward - reward gained in the last iteration.

  • done - indicates if the move was the last step of the environment. This happens if either 10000 iterations were reached or the pelvis height is below 0.6 meters.

  • info - for compatibility with OpenAI, currently not used.

change_model(model='3D', prosthetic=True, difficulty=0, seed=None)

Change model parameters. Your solution will be graded in the environment with difficulty = 2, prosthetic = True and model = 3D, yet it might be easier to train a simplified model first (where model = 2D, difficulty = 0, prosthetic = False` is the simplest).

  • model - 3D model can move in all directions, 2D one dimension is fixed, i.e. the model cannot fall to the left or right.

  • prosthetic - if True the right leg of the model is a prosthesis.

  • difficulty - For the 3D model: 0 - go forward at 3 meters per second, (other values not used for now)

  • seed - starting seed for the random number generator. If the seed is None, generation from the previous seed is continued.

This function does not return any value. reset() must be run after changing the model.

Physics and biomechanics of the model

The model is implemented in OpenSim[1], which relies on the Simbody physics engine. Note that, given recent successes in model-free reinforcement learning, expertise in biomechanics is not required to successfully compete in this challenge.

To summarize briefly, the agent is a musculoskeletal model that include body segments for each leg, a pelvis segment, and a single segment to represent the upper half of the body (trunk, head, arms). The segments are connected with joints (e.g., knee and hip) and the motion of these joints is controlled by the excitation of muscles. The muscles in the model have complex paths (e.g., muscles can cross more than one joint and there are redundant muscles). The muscle actuators themselves are also highly nonlinear. For example, there is a first order differential equation that relates electrical signal the nervous system sends to a muscle (the excitation) to the activation of a muscle (which describes how much force a muscle will actually generate given the muscle's current force-generating capacity). Given the musculoskeletal structure of bones, joint, and muscles, at each step of the simulation (corresponding to 0.01 seconds), the engine:

  • computes activations of muscles from the excitations vector provided to the step() function,
  • actuates muscles according to these activations,
  • computes torques generated due to muscle activations,
  • computes forces caused by contacting the ground,
  • computes velocities and positions of joints and bodies,
  • generates a new state based on forces, velocities, and positions of joints.

In each action, the following 22 muscles are actuated (11 per leg):

  • hamstrings,
  • biceps femoris,
  • gluteus maximus,
  • iliopsoas,
  • rectus femoris,
  • vastus,
  • gastrocnemius,
  • soleus,
  • tibialis anterior. The action vector corresponds to these muscles in the same order (11 muscles of the right leg first, then 11 muscles of the left leg).

The observation contains a dictionary describing the state.

For more details on the simulation framework, please refer to [1]. For more specific information about the muscles model we use, please refer to [2] or to OpenSim documentation.

[1] Delp, Scott L., et al. "OpenSim: open-source software to create and analyze dynamic simulations of movement." IEEE transactions on biomedical engineering 54.11 (2007): 1940-1950.

[2] Thelen, D.G. "Adjustment of muscle mechanics model parameters to simulate dynamic contractions in older adults." ASME Journal of Biomechanical Engineering 125 (2003): 70–77.

Frequently Asked Questions

I'm getting 'version GLIBCXX_3.4.21 not defined in file with link time reference' error

If you are getting this error:

ImportError: /opensim-rl/lib/python2.7/site-packages/opensim/
  symbol _ZTVNSt7__cxx1119basic_istringstreamIcSt11char_traitsIcESaIcEEE, version
  GLIBCXX_3.4.21 not defined in file with link time reference

Try conda install libgcc.

Can I use languages other than python?

Yes, you just need to set up your own python grader and interact with it Find more details here OpenAI http client

Some libraries are missing. What is required to run the environment?

Most of the libraries by default exist in major distributions of operating systems or are automatically downloaded by the conda environment. Yet, sometimes things are still missing. The minimal set of dependencies under Linux can be installed with

sudo apt install libquadmath0 libglu1-mesa libglu1-mesa-dev libsm6 libxi-dev libxmu-dev liblapack-dev

Please, try to find equivalent libraries for your OS and let us know -- we will put them here.

Why there are no energy constraints?

Please refer to the issue

I have some memory leaks, what can I do?

Please refer to and to

How to visualize observations when running simulations on the server?

Please refer to

I still have more questions, how can I contact you?

For questions related to the challenge please use the challenge forum. For issues and problems related to installation process or to the implementation of the simulation environment feel free to create an issue on GitHub.


If you use osim-rl in your research, you can cite it as follows:

  author      = "Kidzi\'nski, {\L}ukasz and Mohanty, Sharada P and Ong, Carmichael and Hicks, Jennifer and Francis, Sean and Levine, Sergey and Salath\'e, Marcel and Delp, Scott",
  title       = "Learning to Run challenge: Synthesizing physiologically accurate motion using deep reinforcement learning",
  editor      = "Escalera, Sergio and Weimer, Markus",
  booktitle   = "NIPS 2017 Competition Book",
  publisher   = "Springer",
  address     = "Springer",
  year        = 2018

If you use the top solutions or other solution reports from the Learning to Run challenge, you can cite them as:

  author      = "Ja\'skowski, Wojciech and Lykkeb{\o}, Odd Rune and Toklu, Nihat Engin and Trifterer, Florian and Buk, Zden\v{e}k and Koutn\'{i}k, Jan and Gomez, Faustino",
  title       = "{Reinforcement Learning to Run... Fast}",
  editor      = "Escalera, Sergio and Weimer, Markus",
  booktitle   = "NIPS 2017 Competition Book",
  publisher   = "Springer",
  address     = "Springer",
  year        = 2018

  author      = "Kidzi\'nski, {\L}ukasz and Mohanty, Sharada P and Ong, Carmichael and Huang, Zhewei and Zhou, Shuchang and Pechenko, Anton and Stelmaszczyk, Adam and Jarosik, Piotr and Pavlov, Mikhail and Kolesnikov, Sergey and Plis, Sergey and Chen, Zhibo and Zhang, Zhizheng and Chen, Jiale and Shi, Jun and Zheng, Zhuobin and Yuan, Chun and Lin, Zhihui and Michalewski, Henryk and Miłoś, Piotr and Osiński, Błażej and Melnik andrew and Schilling, Malte and Ritter, Helge and Carroll, Sean and Hicks, Jennifer and Levine, Sergey and Salathé, Marcel and Delp, Scott",
  title       = "Learning to run challenge solutions: Adapting reinforcement learning methods for neuromusculoskeletal environments",
  editor      = "Escalera, Sergio and Weimer, Markus",
  booktitle   = "NIPS 2017 Competition Book",
  publisher   = "Springer",
  address     = "Springer",
  year        = 2018


This challenge would not be possible without:

Latest Releases
Final NIPS 2017 L2R challenge release
 Dec. 20 2017
Learning how to walk competition
 Apr. 11 2017
 Feb. 21 2017