Learning to Run challenge solutions: Adapting reinforcement learning methods for neuromusculoskeletal environments
Machine Learning
2018-04-03 v1 Artificial Intelligence
Machine Learning
Abstract
In the NIPS 2017 Learning to Run challenge, participants were tasked with building a controller for a musculoskeletal model to make it run as fast as possible through an obstacle course. Top participants were invited to describe their algorithms. In this work, we present eight solutions that used deep reinforcement learning approaches, based on algorithms such as Deep Deterministic Policy Gradient, Proximal Policy Optimization, and Trust Region Policy Optimization. Many solutions use similar relaxations and heuristics, such as reward shaping, frame skipping, discretization of the action space, symmetry, and policy blending. However, each of the eight teams implemented different modifications of the known algorithms.
Cite
@article{arxiv.1804.00361,
title = {Learning to Run challenge solutions: Adapting reinforcement learning methods for neuromusculoskeletal environments},
author = {Łukasz Kidziński and Sharada Prasanna Mohanty and Carmichael Ong and Zhewei Huang and Shuchang Zhou and Anton Pechenko and Adam Stelmaszczyk and Piotr Jarosik and Mikhail Pavlov and Sergey Kolesnikov and Sergey Plis and Zhibo Chen and Zhizheng Zhang and Jiale Chen and Jun Shi and Zhuobin Zheng and Chun Yuan and Zhihui Lin and Henryk Michalewski and Piotr Miłoś and Błażej Osiński and Andrew Melnik and Malte Schilling and Helge Ritter and Sean Carroll and Jennifer Hicks and Sergey Levine and Marcel Salathé and Scott Delp},
journal= {arXiv preprint arXiv:1804.00361},
year = {2018}
}
Comments
27 pages, 17 figures