Embedding Physics Priors in Robot Learning: A Survey
A survey reviews physics-embedded robot learning, proposing a unified taxonomy and arguing for its benefits over purely data-driven methods.
This survey reviews the field of physics-embedded robot learning, which integrates physics priors into learning algorithms to address challenges in robotics like limited data and the need for reliable operation. The authors propose a unified taxonomy that classifies approaches according to their physics embedding: physics-guided inputs, data, and representations; physics-encoded model architectures; and physics-informed training loss functions. The paper argues that physics priors provide a relevant robotics-specific inductive bias that complements data-driven learning, paving the way for more generalizable, data-efficient, and trustworthy robotic systems.
- Survey reviews physics-embedded robot learning across a broad range of applications.
- A unified taxonomy classifies approaches by how physics is embedded (inputs, architectures, loss functions).
- Physics priors are argued to complement data-driven learning for more generalizable robotic systems.
- The survey covers methods from simple models to generative foundation models.
Full article232 words · extracted from huggingface.co · click to collapse
The rapid progress of artificial intelligence is reshaping robotics and accelerating the adoption of learning-based approaches. While purely data-driven methods have achieved remarkable success in computer vision and natural language processing, robotics remains constrained by limited data, complex real-world interactions, and the need for reliable operation. These challenges have motivated the exploration of physics-embedded robot learning, which embeds physics priors into learning algorithms. By encoding the underlying physical laws and constraints, physics priors can complement limited data with robotics-specific inductive biases, potentially improving generalization, interpretability, and sample efficiency. However, the literature on physics-embedded robot learning remains fragmented across terminology, methodologies, and application domains, making it difficult to assess this growing body of work. This survey reviews physics-embedded robot learning across a broad range of physics priors, robotics applications, and machine learning models, from single-layer perceptrons to generative foundation models. We adopt a unified taxonomy that classifies existing approaches according to their physics embedding: physics-guided inputs, data, and representations; physics-encoded model architectures; and physics-informed training loss functions. Building on this taxonomy, we review methods for robot dynamics learning, trajectory planning, prediction, control, and estimation, together with the corresponding open-source software ecosystem. We identify key open challenges, and outline promising future research directions. Overall, we argue that physics priors provide a particularly relevant robotics-specific inductive bias, complementing rather than replacing data-driven learning, and paving the way toward more generalizable, data-efficient, and trustworthy robotic systems.
Text extracted automatically; images, tables and formatting may be missing. Original: https://huggingface.co/papers/2609.22319