Problem
Research questions and friction points this paper is trying to address.
Tactile Sensing
Visual-Tactile Fusion
Force Modulation
Multimodal and Force-Matched Imitation Learning With a See-Through Visuotactile Sensor
🤖 AI Summary
To address low force control accuracy and insufficient vision–tactile fusion in contact-intensive robotic manipulation—leading to poor task success rates—this paper proposes a translucent multimodal visuo-tactile perception and imitation learning framework. We introduce a novel tactile force-matching mechanism for precise replication of demonstrated forces, alongside a data-driven vision–tactile modality switching strategy enabling adaptive transition between approach and contact phases. Further, we enhance modeling of relative motion (e.g., sliding, slipping) via force signal decoupling, trajectory re-planning, and a multimodal policy fusion network. Evaluated on four door-opening tasks, force matching improves success rate by 62.5%, modality switching by 30.3%, and overall visuo-tactile input integration by 42.5%. Results validate the dual advantages of translucent sensing: high-fidelity data acquisition and real-time closed-loop feedback.
📝 Abstract
Contact-rich tasks continue to present many challenges for robotic manipulation. In this work, we leverage a multimodal visuotactile sensor within the framework of imitation learning (IL) to perform contact-rich tasks that involve relative motion (e.g., slipping and sliding) between the end-effector and the manipulated object. We introduce two algorithmic contributions, tactile force matching and learned mode switching, as complimentary methods for improving IL. Tactile force matching enhances kinesthetic teaching by reading approximate forces during the demonstration and generating an adapted robot trajectory that recreates the recorded forces. Learned mode switching uses IL to couple visual and tactile sensor modes with the learned motion policy, simplifying the transition from reaching to contacting. We perform robotic manipulation experiments on four door-opening tasks with a variety of observation and algorithm configurations to study the utility of multimodal visuotactile sensing and our proposed improvements. Our results show that the inclusion of force matching raises average policy success rates by 62.5%, visuotactile mode switching by 30.3%, and visuotactile data as a policy input by 42.5%, emphasizing the value of see-through tactile sensing for IL, both for data collection to allow force matching, and for policy execution to enable accurate task feedback.
Innovation
Methods, ideas, or system contributions that make the work stand out.
Tactile-Visual Sensor
Haptic Force Matching
Learning Mode Switching
🔎 Similar Papers
No similar papers found.
