The Arcade Learning Environment: An Evaluation Platform for General Agents

About

In this article we introduce the Arcade Learning Environment (ALE): both a challenge problem and a platform and methodology for evaluating the development of general, domain-independent AI technology. ALE provides an interface to hundreds of Atari 2600 game environments, each one different, interesting, and designed to be a challenge for human players. ALE presents significant research challenges for reinforcement learning, model learning, model-based planning, imitation learning, transfer learning, and intrinsic motivation. Most importantly, it provides a rigorous testbed for evaluating and comparing approaches to these problems. We illustrate the promise of ALE by developing and benchmarking domain-independent agents designed using well-established AI techniques for both reinforcement learning and planning. In doing so, we also propose an evaluation methodology made possible by ALE, reporting empirical results on over 55 different games. All of the software, including the benchmark agents, is publicly available.

Marc G. Bellemare, Yavar Naddaf, Joel Veness, Michael Bowling · 2012

Related benchmarks

Task	Dataset	Result
Reinforcement Learning	Atari 57	Atlantis29028.1	21
Atari Game Playing	Asterix	Game Score620	11
Atari Game Playing	Seaquest	Game Score156	6
Atari Game Playing	H.E.R.O.	Game Score6087	6
Reinforcement Learning	Space Invaders Atari 2600 (test)	Average Total Reward271	5
Reinforcement Learning	Beam Rider Atari 2600 (test)	Average Episode Reward996	5
Reinforcement Learning	Breakout Atari 2600 (test)	Average Total Reward5.2	5
Reinforcement Learning	Enduro Atari 2600 (test)	Average Total Reward129	5
Reinforcement Learning	Pong Atari 2600 (test)	Average Total Reward-19	5
Reinforcement Learning	Q*bert Atari 2600 (test)	Average Total Reward614	5