Scaling Policy Gradient Quality-Diversity with Massive Parallelization via Behavioral Variations

TL;DR

This paper introduces ASCII-ME, a fast, scalable, and sample-efficient policy gradient-based algorithm for Quality-Diversity optimization that significantly reduces runtime and enables massive parallelization without centralized training.

Contribution

ASCII-ME is a novel method that scales Quality-Diversity optimization with policy gradients, avoiding centralized actor-critic training and enabling rapid, parallel solution generation.

Findings

Generates diverse high-performing neural network policies in under 250 seconds.

Operates five times faster than current state-of-the-art algorithms.

Maintains competitive sample efficiency with massive parallelization.

Abstract

Quality-Diversity optimization comprises a family of evolutionary algorithms aimed at generating a collection of diverse and high-performing solutions. MAP-Elites (ME), a notable example, is used effectively in fields like evolutionary robotics. However, the reliance of ME on random mutations from Genetic Algorithms limits its ability to evolve high-dimensional solutions. Methods proposed to overcome this include using gradient-based operators like policy gradients or natural evolution strategies. While successful at scaling ME for neuroevolution, these methods often suffer from slow training speeds, or difficulties in scaling with massive parallelization due to high computational demands or reliance on centralized actor-critic training. In this work, we introduce a fast, sample-efficient ME based algorithm capable of scaling up with massive parallelization, significantly reducing runtimes without compromising performance. Our method, ASCII-ME, unlike existing policy gradient quality-diversity methods, does not rely on centralized actor-critic training. It performs behavioral variations based on time step performance metrics and maps these variations to solutions using policy gradients. Our experiments show that ASCII-ME can generate a diverse collection of high-performing deep neural network policies in less than 250 seconds on a single GPU. Additionally, it operates on average, five times faster than state-of-the-art algorithms while still maintaining competitive sample efficiency.