Attention-based Deep Reinforcement Learning for Multi-view Environments

TL;DR

This paper introduces an attention-based deep reinforcement learning approach that dynamically focuses on multiple environment views to improve decision-making in complex, partially observable environments, demonstrated on racing and 3D obstacle tasks.

Contribution

It proposes a novel attention mechanism for multi-view reinforcement learning, enabling dynamic view importance weighting for better policy learning.

Findings

Enhanced performance in TORCS racing simulations

Effective handling of partial observability in complex environments

Improved decision-making accuracy with multi-view attention

Abstract

In reinforcement learning algorithms, it is a common practice to account for only a single view of the environment to make the desired decisions; however, utilizing multiple views of the environment can help to promote the learning of complicated policies. Since the views may frequently suffer from partial observability, their provided observation can have different levels of importance. In this paper, we present a novel attention-based deep reinforcement learning method in a multi-view environment in which each view can provide various representative information about the environment. Specifically, our method learns a policy to dynamically attend to views of the environment based on their importance in the decision-making process. We evaluate the performance of our method on TORCS racing car simulator and three other complex 3D environments with obstacles.