Autoencoder-augmented Neuroevolution for Visual Doom Playing
This addresses a bottleneck in neuroevolution for visual reinforcement learning tasks, but it is incremental as it combines existing techniques.
The paper tackled the problem of scaling neuroevolution to high-dimensional inputs like raw pixels by training an autoencoder to create low-dimensional representations and using CMA-ES for controller evolution, achieving good performance on a health-pack gathering task in the VizDoom environment.
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.