Training Unity Machine Learning Agents using reinforcement learning method

2019

Closed access Metadata (CSL-JSON)
DOI 10.1109/icecco48375.2019.9043194
Data confidence 70 medium
How this score is calculated

It is built from checkable properties of the record itself: whether the article has a DOI, how many independent sources confirmed its data, whether the required fields are filled in, and how recent the latest confirmation is. Disagreement between sources and a missing identifier lower it. A value of 80 or above means the record was published without further checking; 60 to 79 means it is published at reduced confidence, with a manual check advisable. A record verified by an editor by hand is never pushed below 90 by a recalculation. This scores our data about the article, not the article itself.

Abstract

This study aims to provide research of Machine Learning tools combined with Unity platform, to develop several Machine Learning Agents in various environments using multiple environment configurations and training scenarios. This work introduces a single competitive agent environment where the agent compete with simulated physics in a 3D world. Although the rules of the environment are generally basic, the trained agent learns a wide variety of interesting and complex solutions.