# Artificial Intelligence for Prosthetics - challenge solutions

## TL;DR

This paper reviews the NeurIPS 2018 challenge on AI-controlled prosthetics, highlighting diverse deep reinforcement learning solutions that incorporate various heuristics and modifications for improved control.

## Contribution

It presents thirteen novel deep reinforcement learning approaches for prosthetic control, demonstrating different modifications and heuristics used by participants.

## Key findings

- Multiple solutions used reward shaping and discretization.
- Teams divided tasks into subtasks and incorporated expert knowledge.
- Solutions achieved effective control in the prosthetic simulation.

## Abstract

In the NeurIPS 2018 Artificial Intelligence for Prosthetics challenge, participants were tasked with building a controller for a musculoskeletal model with a goal of matching a given time-varying velocity vector. Top participants were invited to describe their algorithms. In this work, we describe the challenge and present thirteen solutions that used deep reinforcement learning approaches. Many solutions use similar relaxations and heuristics, such as reward shaping, frame skipping, discretization of the action space, symmetry, and policy blending. However, each team implemented different modifications of the known algorithms by, for example, dividing the task into subtasks, learning low-level control, or by incorporating expert knowledge and using imitation learning.