Robust Dual View Deep Agent

TL;DR

This paper introduces a dual-view deep reinforcement learning architecture that enhances robustness and reduces training parameters by splitting inputs, demonstrated effectively in a Doom-based environment.

Contribution

It proposes a novel dual-stream input architecture within A3C that improves robustness and reduces parameters without sacrificing performance.

Findings

Achieves similar performance to single-input agents

Reduces training parameters by nearly 30%

Demonstrates robustness in a 3D Doom environment

Abstract

Motivated by recent advance of machine learning using Deep Reinforcement Learning this paper proposes a modified architecture that produces more robust agents and speeds up the training process. Our architecture is based on Asynchronous Advantage Actor-Critic (A3C) algorithm where the total input dimensionality is halved by dividing the input into two independent streams. We use ViZDoom, 3D world software that is based on the classical first person shooter video game, Doom, as a test case. The experiments show that in comparison to single input agents, the proposed architecture succeeds to have the same playing performance and shows more robust behavior, achieving significant reduction in the number of training parameters of almost 30%.