🤖 AI Summary
This paper addresses the challenge of remotely verifying ownership of robot control policies solely from external observations (e.g., videos or motion-capture data), a critical open problem in robotic policy intellectual property protection. To bridge the “observability gap” arising from inaccessible internal states in physical systems, we propose the first remote watermarking detection framework tailored for embodied agents. We formalize asynchronous, noisy external observations as *glimpse sequences* and introduce Colored Noise Coherency (CoNoCo)—a novel watermarking method that exploits inherent policy stochasticity to embed provably invariant, robust watermarks in the action spectral domain. We theoretically prove that watermark embedding preserves the marginal action distribution and does not degrade policy performance. Extensive experiments on both simulation and real-world robotic platforms demonstrate high detection robustness across diverse modalities, including multi-view videos and motion-capture data.
📝 Abstract
The success of machine learning for real-world robotic systems has created a new form of intellectual property: the trained policy. This raises a critical need for novel methods that verify ownership and detect unauthorized, possibly unsafe misuse. While watermarking is established in other domains, physical policies present a unique challenge: remote detection. Existing methods assume access to the robot's internal state, but auditors are often limited to external observations (e.g., video footage). This ``Physical Observation Gap'' means the watermark must be detected from signals that are noisy, asynchronous, and filtered by unknown system dynamics. We formalize this challenge using the concept of a extit{glimpse sequence}, and introduce Colored Noise Coherency (CoNoCo), the first watermarking strategy designed for remote detection. CoNoCo embeds a spectral signal into the robot's motions by leveraging the policy's inherent stochasticity. To show it does not degrade performance, we prove CoNoCo preserves the marginal action distribution. Our experiments demonstrate strong, robust detection across various remote modalities, including motion capture and side-way/top-down video footage, in both simulated and real-world robot experiments. This work provides a necessary step toward protecting intellectual property in robotics, offering the first method for validating the provenance of physical policies non-invasively, using purely remote observations.