8.screen shot 2017 04 10 at 1.23.56 pm

NIPS 2017: Learning to Run

Reinforcement learning environments with musculoskeletal models

Days left


Welcome to Learning to Run, one of the 5 official challenges in the NIPS 2017 Competition Track. In this competition, you are tasked with developing a controller to enable a physiologically-based human model to navigate a complex obstacle course as quickly as possible. You are provided with a human musculoskeletal model and a physics-based simulation environment where you can synthesize physically and physiologically accurate motion. Potential obstacles include external obstacles like steps, or a slippery floor, along with internal obstacles like muscle weakness or motor noise. You are scored based on the distance you travel through the obstacle course in a set amount of time.

Amazon AWS has generously agreed to support participants of the challenge with $30,000 worth of cloud credits. Please read below for more information on how to obtain them.

Our objectives are to:

  • bring Deep Reinforcement Learning to solve problems in medicine,
  • promote open-source tools in RL research (the physics simulator, the RL environment, and the competition platform are all open-source),
  • encourage RL research in computationally complex environments, with stochasticity and highly-dimensional action spaces.

Follow the instructions in the Getting Started guide in the Dataset section of the challenge and visit our github repo to get started!

AWS credits

The top 100 performers as per the leaderboard on August 13th, 2017, 23:59:59 UTC, will receive $300 AWS cloud credits. In order to be eligible to receive these credits, your email account associated with crowdAI must be the same as the one you use at AWS, otherwise we won’t be able to credit your AWS account (no exceptions).

2 steps


stanford epfl berkley stanford mobilize


Amazon AWS


Stanford News



Your task is to build a function f which takes the current state observation (a 41 dimensional vector) and returns the muscle excitations action (18 dimensional vector) in a way that maximizes the reward. Your total reward is the position of the pelvis on the x axis after the last iteration minus a penalty for using ligament forces. Ligaments are tissues which prevent your joints from bending too much - overusing these tissues leads to injuries, so we want to avoid it. The penalty in the total reward is equal to the sum of forces generated by ligaments over the trial, divided by 1000. For details on evaluation please refer to the Getting Started guide in the Dataset section of the challenge.


In order to avoid overfitting to the training environment, the top 10 participants will be asked to resubmit their solutions in the second round of the challenge. Environments in the second round will have the same structure but they will be initialized with different seeds. The final ranking will be based on results from the second round.

Additional rules:

  • You are not allowed to use external datasets (e.g., kinematics of people walking)
  • Organizers reserve the right to modify challenge rules as required.


Prizes for top participants include:

  • Invitation to publish articles in the NIPS competition book.
  • Invitation to the 2nd Applied Machine Learning Days at EPFL in Switzerland on January 29 & 30, 2018, with travel and accommodation covered.
  • Invitation to give a research talk at Stanford, with travel and accommodation covered.
  • Reimbursement of travel and accommodation at NIPS 2017


Please refer to the Getting Started guide in the Dataset section of the challenge, for more details on how to access the challenge environments, and also for a basic tutorial on how to make your first submission.

Contact Us

We strongly encourage you to use the public channels mentioned above for communications between the participants and the organisers. In extreme cases, if there are any queries or comments that you would like to make using a private communication channel, then you can send us an email at :