Embedded Files
Code is currently available at anonymous GitHub ↑, will be at GitHub after review.
Challenge:
Challenge:
Existing VLA systems lack force-awareness, leading to unsafe interactions in physical tasks involving contact or uncertainty. We see this challenge as a promising direction for safer deployment of VLA systems.
Existing VLA systems lack force-awareness, leading to unsafe interactions in physical tasks involving contact or uncertainty. We see this challenge as a promising direction for safer deployment of VLA systems.
We present CompliantVLA-adaptor:
We present CompliantVLA-adaptor:
a timely, effective, plug-and-play, and training-free solution
a timely, effective, plug-and-play, and training-free solution
compliantVLA_long.mp4Our method：
Overview of the CompliantVLA-adaptor. A VLM processes visual observations, language instructions, and real-time force feedback F to generate context-aware impedance parameters K, D. These parameters modulate a variable impedance controller (VIC) that executes actions generated by a VLA model, ensuring safe and adaptive contact-rich manipulation.
Other Videos：
wiping_with_failure_case.mp4Pi0.5
push.mp4Pi0.5
compliantVLA_failures_pi0.mp4Pi0
compliantVLA_failures_openvla-oft.mp4OpenVLA-oft
compliantVLA_failures_pi0_tomato.mp4Pi0
Page updated
Google Sites
Report abuse