Multiple Hands Make Light Work: Enhancing Quality and Diversity using MAP-Elites with Multiple Parallel Evolution Strategies
We propose MAP-Elites-Multi-ES (MEMES), a novel ES-based QD algorithm that exploits large-scale parallelisation to improve the search for quality and diversity. MEMES builds on the original intuition from ME-ES but maintains multiple parallel and independent ES-processes, also called emitters, by leveraging fast parallel evaluations. Exploration and exploitation is distributed between the parallel ES emitters, which together simultaneously perform optimization. MEMES also proposes a new dynamic and automatic reset of the independent ES emitters for continuous and efficient improvement of the population.
Ant-Trap
MEMES
PGA-ME
ME
ES
Ant-Uni
MEMES
PGA-ME
ME
NSR-ES
Hexapod-Omni
MEMES
ME
CMA-ME
NS-ES