ToriLLE: Learning Environment for Hand-to-Hand Combat

TL;DR

ToriLLE is a new learning environment based on the game Toribash, designed for training and evaluating machine learning agents in hand-to-hand combat scenarios, supporting self-play and human comparison.

Contribution

This paper introduces ToriLLE, a novel environment for machine learning research using a fighting game, with detailed environment capabilities and experimental validation.

Findings

ToriLLE effectively supports training of ML agents in combat scenarios.

The environment enables evaluation against human players.

Experimental results demonstrate its applicability for reinforcement learning.

Abstract

We present Toribash Learning Environment (ToriLLE), a learning environment for machine learning agents based on the video game Toribash. Toribash is a MuJoCo-like environment of two humanoid character fighting each other hand-to-hand, controlled by changing actuation modes of the joints. Competitive nature of Toribash as well its focused domain provide a platform for evaluating self-play methods, and evaluating machine learning agents against human players. In this paper we describe the environment with ToriLLE's capabilities and limitations, and experimentally show its applicability as a learning environment. The source code of the environment and conducted experiments can be found at https://github.com/Miffyli/ToriLLE.