New survey maps 237 methods for embedding physics priors in robot learning
Jan_R_Peters · x · 2026-10-03
Jan Peters and colleagues have released a new survey, "Embedding Physics Priors in Robot Learning," covering 237 methods and 329 references on the topic.
Core argument: purely data-driven methods excel in vision and NLP, but robotics is constrained by limited data, complex physical interaction, and reliability demands. Encoding physical laws as inductive biases can improve generalization, interpretability, and sample efficiency.
Unified taxonomy: approaches are classified by how physics is embedded — physics-guided inputs/data/representations, physics-encoded architectures, and physics-informed training losses — spanning models from single-layer perceptrons to generative foundation models. The paper is available on arXiv and serves as a field map for this growing area.
More from Embodied
- ETH's Anymal robot completes 10km autonomous alpine hike with VLM-guided ROAM system — ChongZzZhang · 2026-11-19
- iFixit tears down Apple HomeView: lots of glue and 'endless frustration' — Resident-Fall4351 · 2026-11-19
- Removing human video drops robot OOD success from 60% to 7%: generalization comes from human video — gui_penedo · 2026-11-19
- PSYONIC's prosthetic Ability Hand now used by 90+ robotics orgs, raising Series A — DhowCIO · 2026-11-19
- Simulator misspecification breeds overconfidence: sim-to-real study on SPEED+ — Nikolay Dolgov · 2026-11-19
- Doppel world model renders the robot body and beats action-only baselines across 8 embodiments — mangahomanga · 2026-11-19