Contact-mode decision making
Robots must decide when to grasp, push, drag, brace, lean, pin, wedge, pivot, or exploit the environment, and how to transition between those strategies under uncertainty.
Workshop Theme
Recent advances in robot learning, control, perception, and planning have expanded the ability of mobile robots to move through complex environments and interact with the physical world. Legged robots, humanoids, and mobile manipulators are increasingly demonstrated in tasks that require whole-body motion, forceful interaction, and adaptation to uncertain environments.
Many real-world loco-manipulation problems cannot be solved through grasping alone. Robots must combine prehensile skills, such as grasping and carrying, with non-prehensile skills, such as pushing, dragging, pivoting, bracing, or using sustained body-object contact. The contact strategy is directly coupled to locomotion, balance, onboard sensing, planning, and whole-body control.
This workshop brings contact-rich loco-manipulation into sharper focus as a central problem for robot learning. It connects researchers working on legged locomotion, mobile manipulation, non-prehensile manipulation, whole-body control, contact-aware planning, onboard perception, and sim-to-real transfer.
Core Challenges
Whole-body coordination under contact
Contacts made by arms, legs, torso, base, objects, and the environment affect balance, mobility, force regulation, and task success.
Reasoning under uncertainty
Robots need to act with uncertain friction, mass, compliance, center of mass, contact location, slip, object geometry, terrain, and partial observability.
Generalization across embodiments
The field needs principles and representations that transfer across arms, mobile manipulators, quadrupeds with arms, humanoids, soft robots, and dexterous hands.
Benchmarks, metrics, and infrastructure
Progress requires shared task families, metrics, simulation assets, and real-world protocols that capture robustness, safety, recoverability, contact efficiency, and generalization.
Suggested Invited Speakers
Speaker invitations are in progress and will be updated as confirmations arrive.
Suggested speaker
Alan Fern
Oregon State University
Robot learning, planning, reinforcement learning, and decision making under uncertainty.
Suggested speaker
Tao Pang
Robotics and AI Institute
Contact-rich manipulation, planning, control, and robust physical interaction.
Suggested speaker
Pulkit Agrawal
MIT
Robot learning, legged mobility, manipulation, and sim-to-real transfer.
Workshop Format
Focused invited talks
A small number of invited talks will frame the central open problems and leave time for direct Q&A around contact-mode selection, whole-body coordination, and uncertainty.
Lightning talks and posters
Accepted workshop papers will be introduced through short spotlights and then discussed in an interactive poster session.
Panel and community artifact
Speakers, organizers, authors, and attendees will contribute questions and discussion points for a panel and a post-workshop open-problems report.
Call for Papers
Short workshop papers
We plan to invite 2-4 page workshop papers on contact-rich loco-manipulation, including new results, early-stage ideas, benchmark proposals, position papers, open problems, negative results, and failure analyses.
Prehensile and non-prehensile skills
Whole-body control
Contact-mode representations
Multimodal sensing
Sim-to-real transfer
Audience
Researchers in robot learning, legged locomotion, mobile manipulation, non-prehensile manipulation, contact-rich control, perception, planning, and benchmarking.