Keeping Contact: Task-Adaptive Friction Margins for Bimanual Non-Prehensile Manipulation
Authors: Qian Ying Lee, Lin Yang, Domenico Campolo
Organizations: School of Mechanical and Aerospace Engineering, Nanyang Technological University (NTU), Singapore. · R3 Institute (R3I) Singapore, Nanyang Technological University (NTU), Singapore
Maintaining contact in non-prehensile manipulation requires more than nominal friction cone feasibility: a force barely inside the cone and one well inside it are both feasible, but they provide different margins against slip and contact loss. This paper uses signed friction distance to measure friction margin in force space and derives a task-adaptive margin from the force correction needed to balance pivot moment deviations in bimanual tilting. The resulting contact margins are embedded in a sequence of second-order cone programs (SOCPs), each giving a globally optimal contact force solution at a tilt angle. The optimizer balances the object about an environmental pivot while keeping robot and pivot contact forces away from the nominal friction boundary. Optimizer-level results show interpretable force redistribution and adaptation to object property changes. Real-world experiments demonstrate successful execution under object property variations and model mismatch, while fixed margin and no-margin comparisons illustrate the importance of adapting the margin to task demand. These results support signed friction distance margins as a force planning mechanism for maintaining contact in bimanual non-prehensile manipulation. Project page: https://keeping-contact.github.io/
Figures & tables
Fig. 1: Bimanual non-prehensile tilting. Inset illustrates two feasible contact forces in the local force plane. Translucent circles show equal force perturbations around each force tip: a near boundary force can leave the friction cone and slip, whereas a force with finite margin remains feasible.
Fig. 2: Signed friction distance in the contact force plane. The nominal boundary ψμ(f)=0 is the friction limit, while ψμ(f)≥ρ keeps the planned force a finite signed distance inside the cone. When ft=0 , the same constraint induces a minimum normal force fn,min . The color bar schematically indicates increasing task force demand Ftask and margin ρ .
Fig. 3: Bimanual tilting model. Two robot contacts apply forces to the object while an environmental pivot supplies a passive reaction.
Fig. 4: Experimental setup. (a) Kinova Gen3 robot arm with an ATI Mini40 force/torque sensor at the wrist and a hemispherical end-effector covered with friction tape. (b) Hole pattern plate for varying the object CoM. (c) Calibration weights used for varying the object mass.
Condition
Varied factor
Mass (kg)
CoM offset (mm)
Margins
Nominal
–
1.9
0
ρi(θ)
Light object
Mass
1.7
0
ρi(θ)
Heavy object
Mass
2.2
0
ρi(θ)
Toward pivot CoM
CoM
1.9
−6.3
ρi(θ)
Away from pivot CoM
CoM
1.9
+6.3
ρi(θ)
No margin
Margin
1.9
0
0
TABLE I: Object property and margin evaluation conditions. Mass and CoM offset define the optimizer model. Object property variations retain the same contact-specific margin law and parameters, while the no-margin condition uses nominal properties with ρi=0 . CoM offsets are relative to the nominal.
Fig. 5: Nominal contact force profiles and margin effects. Optimized columns compare robot contact forces with and without margins. Shading indicates their difference. Orange curves show contact-specific margins ρi(θ) . Dotted lines mark fn,min and θtip . Insets show force vectors, with open and filled markers denoting no margin and margin solutions. The experiment column overlays all trials and their means.
Fig. 6: Representative tilting sequences under nominal, no margin, and fixed margin CoM mismatch conditions. The nominal trial completes the tilt with both contacts maintained, while removing the margin leads to contact loss. Both fixed margin trials exhibit severe slip: the bottom left terminates before completing the tilt, and the bottom right loses the left contact as it slips below the box.
Fig. 7: Adaptation to object property variations. Rows show mass and CoM variations. Columns show ρΣ , optimized and experimental FΣ , all in newtons. Mass changes scale the profiles, CoM changes shift their minima. Markers indicate profile minima. Vertical dotted lines mark θtip .
Condition
Success trials
Force std. dev. (N)
Violation rate (%)
Nominal
5/5
0.46
6.9
Light object
5/5
0.50
3.9
Heavy object
5/5
0.40
4.0
Toward pivot CoM
5/5
0.38
3.3
Away from pivot CoM
5/5
0.36
2.9
No margin
0/5
–
90.7
TABLE II: Real-world experiment outcomes. Successful trials complete the tilt with both robot contacts maintained. Force standard deviation summarizes trial-to-trial variation in experimental contact forces and is omitted for failed conditions. Violation rate is the percentage of samples with fn≤0.5N or ∣ft∣/(μfn)>1 . † Omitted as failed trials were terminated early for safety. FM: fixed margin.
Fig. 8: Friction mismatch at robot contacts. Dashed and solid lines denote the planning and measured friction cones, respectively. Red shading marks the region feasible only under the planning coefficient. Dotted lines show margin boundaries at 30∘ intervals. Optimized forces and experimental samples are colored by tilt angle. Inset magnifies one boundary’s intersection with the measured cone.
Fig. 9: Task-adaptive and fixed margins under CoM mismatch with matched average combined force magnitude. Left: task-level margins, with red (blue) shading where the fixed margin is lower (higher) than the task-adaptive margin. Force plane panels show optimized profiles (top) and task-adaptive experimental samples (bottom) for both contacts. Inset magnifies the optimized forces at θ=10∘ . Line style denotes formulation and color denotes tilt angle.
Fig. 10: Left: optimized friction utilization ηmax with and without margins. Right: experimental robot contact forces over trials for the with margin (top row) and no margin (bottom row) conditions. Points show experimental forces, curves show optimized forces, with color indicates tilt angle. Shading shows the robot–object friction cone, and crosses mark fn≤0.5N , indicating contact loss.
Non-prehensile planar manipulation, including pushing and press-and-slide, is critical for diverse robotic tasks, but notoriously challenging due to hybrid contact mechanics, under-actuation, and asymmetric friction limits that traditionally necessitate computationally expensive iterative control. In this paper, we propose a mode-aware framework for planar manipulation with one or two robotic arms based on contact topology selection and reduced-order kinematic modeling. Our core insight is that complex wrench-twist limit surface mechanics can be abstracted into a discrete library of physically intuitive models. We systematically map various single-arm and bimanual contact topologies to simple non-holonomic formulations, e.g. unicycle for simplified press-and-slide motion. By anchoring trajectory generation to these reduced-order models, our framework computes the required object wrench and distributes feasible, friction-bounded contact forces via a direct algebraic allocator. We incorporate manipulator kinematics to ensure long-horizon feasibility and demonstrate our fast, optimization-free approach in simulation across diverse single-arm and bimanual manipulation tasks. Supplementary videos and additional information are available at: https://sites.google.com/view/pushpressslide
Melih Özcan, Umut Orguner, Ozgur S. Oguz
Dept. of Electrical and Electronics Engineering, Middle East Technical University, Ankara, Turkiye. · Dept. of Computer Engineering, Bilkent University, Ankara, Turkiye.
Contact-implicit trajectory optimization formulates contact-rich manipulation as a single constrained program; however, that single program run collapses onto one local optimum out of many equally valid contact modes, grasps, or push directions. As a consequence, the resulting manipulation strategy is reluctant to change and sensitive to initialization. In order to promote robust manipulation, this paper investigates how contact-implicit solvers can discover diverse contact-rich strategies. Our approach derives a variation of Consensus Alternating Direction Method of Multipliers (ADMM) combined with Stein variational inference methods to output a set of distinct contact-rich solutions. We find that applying the Stein repulsive force to ADMM's split variable (rather than its primal form) allows for effective coverage over the set of feasible contact strategies without prematurely stalling the solver. We demonstrate the effectiveness of our approach on a variety of contact-rich manipulation tasks, including pushing, grasping, and multi-robot handover. Last, we find the proposed solver is simpler in form and capable of discovering unique contact modes when compared with existing solvers. Videos and code with examples are found in https://anon-website-submission.github.io/stein-admm-website/.
Hrishikesh Sathyanarayan, Christian Hughes, Ian Abraham
Department of Mechanical Engineering, Yale University, New Haven, CT, USA. · Department of Electrical Engineering, University of Sydney, NSW, Australia.
Generalizable contact-rich manipulation requires robots to preserve intended task behavior while adapting its physical realization to changing contact conditions. However, interaction forces can vary substantially with small changes in surface geometry, orientation, and friction, making policies trained directly on raw force measurements difficult to transfer beyond demonstrated conditions. We introduce FACE, a contact-factorized imitation learning framework that separates intended task behavior from environment-dependent contact factors. Our representation expresses interaction forces in normalized, contact-relative coordinates, while a learned contact-normal estimator and an online friction estimator infer the local contact normal and effective friction scale. Together, these estimators enable force observations to be encoded and policy outputs to be decoded into physical motion and force commands during execution. In this way, FACE adapts execution to current contact conditions while preserving the intended task behavior, without updating the policy parameters. We evaluate FACE on real-robot contact-rich manipulation under unseen variations in surface properties and geometry, demonstrating robust generalization across contact conditions through controlled comparisons with variants that adapt prior approaches to our setting. Videos and additional materials can be found on the project page: https://rcilab.khu.ac.kr/face.
Jiho Hong, Daeun Song, Sanghyun Kim +1
Kyung Hee University · Ewha Womans University · University of Central Florida