Object-sensitive Deep Reinforcement Learning

TL;DR

This paper introduces a method to incorporate object recognition into deep reinforcement learning, improving performance on Atari games and providing visual explanations of agent actions.

Contribution

It presents a novel approach to integrate object recognition with deep reinforcement learning, enhancing both performance and interpretability.

Findings

Achieved state-of-the-art results on Atari games

Proposed object saliency maps for visual explanations

Demonstrated adaptability to existing deep RL frameworks

Abstract

Deep reinforcement learning has become popular over recent years, showing superiority on different visual-input tasks such as playing Atari games and robot navigation. Although objects are important image elements, few work considers enhancing deep reinforcement learning with object characteristics. In this paper, we propose a novel method that can incorporate object recognition processing to deep reinforcement learning models. This approach can be adapted to any existing deep reinforcement learning frameworks. State-of-the-art results are shown in experiments on Atari games. We also propose a new approach called "object saliency maps" to visually explain the actions made by deep reinforcement learning agents.