Evaluation metrics for humanoid standing-up control

Develop standardized and validated evaluation metrics for humanoid standing-up control that capture core performance aspects and can be applied consistently across diverse postures and terrains.

Background

The authors note that there is no agreed-upon set of metrics to evaluate humanoid standing-up performance comprehensively. They propose several metrics (success rate, feet movement, motion smoothness, and energy) to initiate progress, but emphasize that the broader design of evaluation criteria remains unresolved in the field.

Establishing widely accepted metrics would facilitate fair comparison, rigorous analysis, and progress tracking across different methods and environments.

References

While the design of evaluation metrics for humanoid standing-up control remains an open question, we aim to make a step forward by proposing the following metrics:

Learning Humanoid Standing-up Control across Diverse Postures (2502.08378 - Huang et al., 12 Feb 2025) in Simulation Experiments (Section 5) – Evaluation Metrics subsubsection