Benchmarking Quality-Diversity Algorithms on Neuroevolution for Reinforcement Learning
This provides a tool for researchers in reinforcement learning and robotics to compare and improve algorithms, but it is incremental as it standardizes existing methods rather than introducing new ones.
The authors tackled the lack of standardized benchmarks for Quality-Diversity algorithms in deep neuroevolution for reinforcement learning by creating a benchmark suite with tasks, environments, and metrics like coverage and QD-score, and they made the source code publicly available.
We present a Quality-Diversity benchmark suite for Deep Neuroevolution in Reinforcement Learning domains for robot control. The suite includes the definition of tasks, environments, behavioral descriptors, and fitness. We specify different benchmarks based on the complexity of both the task and the agent controlled by a deep neural network. The benchmark uses standard Quality-Diversity metrics, including coverage, QD-score, maximum fitness, and an archive profile metric to quantify the relation between coverage and fitness. We also present how to quantify the robustness of the solutions with respect to environmental stochasticity by introducing corrected versions of the same metrics. We believe that our benchmark is a valuable tool for the community to compare and improve their findings. The source code is available online: https://github.com/adaptive-intelligent-robotics/QDax