Observe and Look Further: Achieving Consistent Performance on Atari

TL;DR

This paper introduces a new deep RL algorithm that overcomes key challenges like reward processing, long-term reasoning, and exploration, achieving human-level performance on most Atari games and solving Montezuma's Revenge's first level.

Contribution

The paper presents a novel algorithm with a transformed Bellman operator, an auxiliary loss for stability, and human demonstrations to improve exploration in Atari games.

Findings

Exceeds average human performance on 40 out of 42 Atari games.

First deep RL algorithm to solve the first level of Montezuma's Revenge.

Effectively handles diverse reward scales and long-term planning.

Abstract

Despite significant advances in the field of deep Reinforcement Learning (RL), today's algorithms still fail to learn human-level policies consistently over a set of diverse tasks such as Atari 2600 games. We identify three key challenges that any algorithm needs to master in order to perform well on all games: processing diverse reward distributions, reasoning over long time horizons, and exploring efficiently. In this paper, we propose an algorithm that addresses each of these challenges and is able to learn human-level policies on nearly all Atari games. A new transformed Bellman operator allows our algorithm to process rewards of varying densities and scales; an auxiliary temporal consistency loss allows us to train stably using a discount factor of $\gamma = 0.999$ (instead of $\gamma = 0.99$) extending the effective planning horizon by an order of magnitude; and we ease the exploration problem by using human demonstrations that guide the agent towards rewarding states. When tested on a set of 42 Atari games, our algorithm exceeds the performance of an average human on 40 games using a common set of hyper parameters. Furthermore, it is the first deep RL algorithm to solve the first level of Montezuma's Revenge.