Don't Bet on Luck Alone: Enhancing Behavioral Reproducibility of Quality-Diversity Solutions in Uncertain Domains

Luca, Grillotti,; Flageat, Manon; Lim, Bryan; Cully, Antoine

doi:10.1145/3583131.3590498

Cited by 3 publications

(2 citation statements)

References 15 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Performance metrics To evaluate the ability of a policy to solve a task given a skill z, we estimate the expected undiscounted return, denoted R(z), by averaging the return over 10 rollouts, as defined by ; Grillotti et al (2023). First, we use R(z) to compute performance profiles on Figure 4, which quantify for a given return R, the proportion of skills in the skill space that have an expected return larger than R, after filtering out the skills that are not achieved by the policy.…”

Section: Evaluation Metricsmentioning

confidence: 99%

“…In this section, we illustrate how to compute and read the distance and performance profiles in Figure 4. In the Quality-Diversity community, there is a consensus that the best evaluation metric is the "distance/performance profile" (Flageat et al, 2022;Grillotti et al, 2023;Grillotti & Cully, 2022a;Batra et al, 2023). This metric is also being used in skill learning for robotics .…”

Section: D3 Evaluation Metrics Detailsmentioning

confidence: 99%

See 1 more Smart Citation

Relevance-guided unsupervised discovery of abilities with quality-diversity algorithms

Luca

Cully

2022

Proceedings of the Genetic and Evolutionary Computation Conference

View full text Add to dashboard Cite

Quality-Diversity algorithms provide efficient mechanisms to generate large collections of diverse and high-performing solutions, which have shown to be instrumental for solving downstream tasks. However, most of those algorithms rely on a behavioural descriptor to characterise the diversity that is hand-coded, hence requiring prior knowledge about the considered tasks. In this work, we introduce Relevance-guided Unsupervised Discovery of Abilities; a Quality-Diversity algorithm that autonomously finds a behavioural characterisation tailored to the task at hand. In particular, our method introduces a custom diversity metric that leads to higher densities of solutions near the areas of interest in the learnt behavioural descriptor space. We evaluate our approach on a simulated robotic environment, where the robot has to autonomously discover its abilities based on its full sensory data. We evaluated the algorithms on three tasks: navigation to random targets, moving forward with a high velocity, and performing half-rolls. The experimental results show that our method manages to discover collections of solutions that are not only diverse, but also well-adapted to the considered downstream task.

show abstract

Section: Evaluation Metricsmentioning

confidence: 99%

Section: D3 Evaluation Metrics Detailsmentioning

confidence: 99%