Autoencoder-augmented Neuroevolution for Visual Doom Playing

TL;DR

This paper introduces a hybrid approach combining autoencoders and neuroevolution to enable effective reinforcement learning directly from raw pixel data in complex visual environments like VizDoom.

Contribution

It presents a novel method that trains an autoencoder alongside neuroevolution to handle high-dimensional visual inputs, improving scalability in reinforcement learning tasks.

Findings

Effective in VizDoom health-pack gathering task

Autoencoder training adapts during evolution

Outperforms traditional neuroevolution methods

Abstract

Neuroevolution has proven effective at many reinforcement learning tasks, but does not seem to scale well to high-dimensional controller representations, which are needed for tasks where the input is raw pixel data. We propose a novel method where we train an autoencoder to create a comparatively low-dimensional representation of the environment observation, and then use CMA-ES to train neural network controllers acting on this input data. As the behavior of the agent changes the nature of the input data, the autoencoder training progresses throughout evolution. We test this method in the VizDoom environment built on the classic FPS Doom, where it performs well on a health-pack gathering task.