Source-linked AI summary
Hybrid Roller-Jamming Gripper for Object Acquisition and Retention Under Pose Uncertainty
Yijie Ren, Guillaume Gourmelen, Hiroyasu Iwata
TL;DR
Pose uncertainty can produce off-centre or partial contact that makes object acquisition difficult, while roller and jamming grippers address different stages of grasping. This paper combines inward roller intake with vacuum-induced granular-jamming stiffening and evaluates the hybrid across offset conditions and ablations. It achieved 812/840 successes overall, including 215/216 planar-offset and 597/624 orientation-offset successes, while the authors identify calibrated object-specific settings and limited sensing as important scope boundaries.
Problem
Off-centre or partial initial contact makes reliable grasp acquisition difficult, while granular jamming depends on sufficient contact before stiffening can secure an object.
Method
The paper presents a hybrid gripper that uses rolling to develop contact and then granular jamming to stabilise the grasp, evaluated with offset trials and ablations.
Results
812/840 overall successes were achieved across eight objects, including 215/216 planar-offset and 597/624 orientation-offset successes.
Takeaways & Limitations
The ablation results support complementary roles for rolling in acquisition after partial contact and jamming in post-capture retention.
Takeaways & Limitations
Object-specific calibration and indirect motor-current sensing limit the system to a mechanism-validation platform rather than a fully autonomous grasping system.
Abstract
from arXiv · showhide
In household manipulation, pose uncertainty often results in off-centre or partial initial contact, making reliable object acquisition difficult. Roller-based grippers can actively draw objects inward but often provide limited post-capture stability, whereas granular-jamming grippers require sufficient contact before jamming to achieve strong retention. This paper presents a hybrid roller-jamming gripper that integrates active object intake and post-capture retention within a single gripper. The proposed gripper uses inward roller rotation to increase contact and draw the object toward the gripper centre, followed by vacuum-induced granular jamming to stiffen the rollers and stabilise the grasp. The paper also presents a simplified geometric analysis of the gripper and a bench-level characterisation of the prototype's force capability. The gripper prototype was mounted on a 7-DoF robotic arm and evaluated using eight test objects. Furthermore, controlled planar position and orientation offsets were applied, with each condition repeated three times. The main evaluation comprised 840 grasp trials, including 216 planar-offset trials and 624 orientation-offset trials. Overall, the gripper succeeded in 812/840 trials: 215/216 planar-offset trials and 597/624 orientation-offset trials. The ablation evaluation comprised 162 trials on three objects. The roller-only and jamming-only conditions achieved 54/81 and 24/81 successes, respectively, showing their different contributions. These results provide initial mechanism-level evidence that hybrid roller-jamming is a promising strategy for improving acquisition and retention after imperfect first contact.
I. INTRODUCTION
Household pose uncertainty makes reliable acquisition difficult after off-centre or partial contact. The paper motivates a hybrid gripper that combines active rolling intake with granular-jamming retention.
- Motivation: Off-centre or partial initial contact makes reliable grasp acquisition difficult in cluttered household environments.Home objects vary in shape, size, stiffness, surface properties, and fragility, while observations and pose estimates are often incomplete or noisy.
- Related work: Compliant and underactuated grippers passively adapt contact geometry while avoiding explicit control of every joint.The related-work discussion identifies compliance and underactuation as widely adopted strategies for robust grasping across diverse objects.
- Related work: Granular jamming remains compliant during approach and stiffens under vacuum, improving stability after contact through frictional reinforcement and geometric interlocking.The paper also notes that sufficient contact must be established before jamming can effectively secure the object.
- Problem gap: Actively driven rollers draw objects inward, translate them within the grasp, or reorient them while maintaining contact.Rolling contact can convert partial engagement into a more centred, secure grasp, but roller-based systems do not inherently increase post-capture stiffness or retention.
- Proposed approach: The proposed gripper uses the same spherical membrane rollers for active intake and subsequent vacuum-induced stiffening.The study combines rolling during acquisition with granular jamming during retention and evaluates the design through geometric characterisation, force tests, offset trials, and ablations.
II. GRIPPER DESIGN AND SYSTEM ARCHITECTURE
The gripper is a two-finger system that combines a shared closing mechanism with independently driven roller fingers. Its architecture integrates active rolling and granular jamming within the same end-effector.
- System architecture: The two-finger architecture comprises a base module and two opposing roller-finger modules.The base module opens and closes both fingers, while each roller-finger module rotates its roller and connects it to the vacuum system.
- System integration: The prototype keeps pneumatic, electronic, and control hardware off-board during experimental evaluation.This arrangement supports operation of the gripper through the experimental robot-arm setup without making the gripper specific to that platform.
- Actuation: Three actuated degrees of freedom provide one shared finger-opening DoF and one independent DoF for each roller.The main dimensions and mass are summarised in the paper’s mechanical-specification table.
- Base module: A single XM430 actuator drives synchronous, self-locking finger motion through a 1:40 worm-gear transmission.The transmission engages two worm wheels to move the fingers symmetrically in opposite directions.
- Roller fingers: Each roller uses an XL320 actuator and a 1:20 worm-gear transmission, allowing inward rotation for intake and outward rotation for release.The rollers are bearing-supported and measure approximately 80 mm in outer diameter and 73 mm in axial width.
4) Design choices:
The prototype design uses spherical silicone membranes, mixed granular filling, and sealed rotary vacuum interfaces. These choices supported smoother contact, retained airflow during evacuation, and enabled repeated experimental use.
- Design choices: Spherical membrane rollers replaced cylindrical rollers because cylindrical membranes folded near central contact and reduced contact formation.The spherical design produced smoother deformation and more stable contact development.
- Membrane fabrication: Each roller membrane is a 1 mm-thick Dragon Skin 30 spherical shell with an approximately 80 mm outer diameter and dimpled surface.The membrane is cast in a two-part mould whose core-to-cavity gap determines thickness; the dimples are intended to improve flexibility.
- Granular filling: Mixed filling with ground coffee and approximately 50 polypropylene balls was selected to retain adaptability while reducing clumping and preserving airflow during evacuation.Each roller-finger module has an approximate mass of 0.12 kg.
- Vacuum interface: Clamped ABS end rings create an airtight seal while providing the roller drive interface and a rotary-union connection for vacuum feedthrough.The rotary union allows vacuum application while the roller remains free to rotate.
- Durability and maintenance: No membrane failures occurred during approximately 1000 grasp trials, and the membrane and infill could be replaced in approximately 10 minutes.The replaceability follows from securing the membrane with clamped end rings rather than permanent bonding.
III. GEOMETRIC ANALYSIS AND BENCH CHARACTERISATION
The gripper is analysed using a simplified rigid-envelope model and a nominal geometric estimate for small-object enclosure. The model is explicitly limited by deformation, friction, asymmetry, and three-dimensional effects.
- Nominal closing geometry: The model represents two length-L rigid links rotating symmetrically about pivots separated by d_p, with roller radius R and spacing d_c.
- Nominal closing geometry: The nominal inter-roller gap g(θ) distinguishes finite separation, tangential contact, and geometric overlap.
- Minimum supported feature size: The minimum supported circular-feature radius is derived by modelling the feature as tangent to both roller envelopes and the support plane at θ_min.
- Minimum supported feature size: r_min ≈ 1.9 mm for R = 40 mm and d_c(θ_min) = 35 mm, corresponding to a nominal diameter of approximately 3.8 mm.
- Model scope: The geometric estimate is a reference rather than an experimentally verified graspability limit.
C. Effect of Front-View Tilt on Minimum Supported Circular-Feature Size
The tilted-contact analysis extends the nominal circular-feature geometry to front-view roller tilt, while bench tests provide single-trial force measurements for the prototype.
- Effect of front-view tilt: The tilted model rotates the fully closed roller pair by Δθ while maintaining nominal centre spacing and tangency assumptions.
- Effect of front-view tilt: As Δθ → 0, the tilted expression recovers the zero-tilt result d^2/(16R).
- Model limitations: The tilted geometric trend excludes membrane deformation, asymmetric or out-of-plane contact, and the full roll–pitch–yaw offsets used experimentally.
- Bench characterisation: Each bench test was performed once, so the reported forces are single-trial maxima rather than statistical force estimates.
- Bench characterisation: 8.4 N, 25.6 N, and 27.7 N were the maximum measured forces for roller traction, grasp slip-out resistance, and clamping force, respectively.
B. State Definitions
The controller sequences contact detection, inward intake, clamping, vacuum jamming, and release through a five-state finite-state machine. Intake draws objects inward before jamming, while retention is evaluated through a fixed motion routine.
- CLOSE: CLOSE uses a motor-current threshold and dwell condition as an indirect indication of initial contact.
- INTAKE: During INTAKE, inward-rotating rollers draw the object toward the gripper centre and increase membrane engagement before jamming.
- HOLD: HOLD increases clamping until its current threshold is maintained for the prescribed dwell time, then transitions to JAM.
- JAM: JAM applies vacuum to both membrane rollers for a fixed 2-s interval before lifting and transport begin.
- State definitions: The finite-state machine contains CLOSE, INTAKE, HOLD, JAM, and RELEASE states.
- Experimental procedure: The experiments distinguish acquisition from post-capture retention using a fixed retention routine after grasping.
C. Test Objects
The evaluation uses eight diverse objects and object-specific calibrated parameters under controlled offset conditions. Success requires acquisition, lifting, and retention through the complete post-grasp motion routine.
- Test objects: Eight objects represent slender, spherical, flat, deformable, thin-walled, textile, and irregular grasping challenges.
- Calibrated parameters: Grasp height, intake duration, and HOLD settings were calibrated separately for each object before repeated experiments.
- Calibrated parameters: Object-specific parameters remained fixed across offset conditions and repeated trials, with a separate height used for orientation-offset tests.
- Success criterion: A successful trial required acquisition, lifting clear of the table, and retention throughout the complete post-grasp motion routine.
- Retention routine: The retention routine applies ±75 mm x/y translations, ±30 mm z translations, and ±15° yaw, pitch, and roll rotations after lifting.
- Planar offset robustness: Planar offsets were sampled on a 3 × 3 grid with maximum axis offsets equal to 20% of the corresponding object extents.
G. Orientation Offset Robustness Test
Orientation offsets were tested across 26 non-zero roll-pitch-yaw combinations, with three repetitions per object. The gripper achieved 597/624 successful trials, or 95.7%, under this protocol.
- Test protocol: 624 orientation-offset trials covered 26 non-zero roll-pitch-yaw conditions, repeated three times for each of eight objects.The paper sheet used reduced offset magnitudes of ±5°, while the other conditions used ±10° offsets.
- Test protocol: The orientation-test grasp height was calibrated separately where needed and then held fixed across all orientation conditions and repetitions.This adjustment addressed reduced table clearance during tilted approaches.
B. Planar Offset Robustness Results
Under controlled planar offsets, the hybrid gripper achieved 215/216 successes, while ablations showed that rolling and jamming contributed differently to acquisition and retention across objects.
- Planar-offset results: 215/216 planar-offset trials succeeded, yielding a 99.5% success rate across the tested offset conditions.Seven objects completed all 27 trials, while the gallon jug completed 26/27 trials and accounted for the only failure.
- Ablation results: 80/81 hybrid, 54/81 roller-only, and 24/81 jamming-only trials succeeded in the three-object ablation evaluation.The roller-only condition disabled jamming, while the jamming-only condition disabled roller intake.
- Mechanism interpretation: Inward roller motion often drew objects toward the membrane centre and increased the contact area available before stiffening.The repeated planar-offset results indicate recovery from the tested offsets before jamming was applied.
- Ablation interpretation: The gallon jug performed better with jamming only, whereas the pen and cable performed better with rolling only.The jug primarily required retention during lifting, while the thin pen and cable required roller intake for initial acquisition.
- Mechanism interpretation: Rolling mainly supports acquisition and contact formation, while jamming improves post-capture retention in the complete hybrid sequence.The hybrid sequence first develops contact through rolling and then stabilises the grasp through jamming.
C. Interpretation of the Orientation Offset Results
Orientation offsets expose greater sensitivity to three-dimensional contact conditions than planar offsets, especially for rigid or asymmetric objects. The paper sheet is a separate scope case because collision avoidance reduced its tested orientation range.
- Orientation sensitivity: Orientation offsets made rolling less effective when changed roller contact produced one-sided or shallow engagement before jamming.The controller may begin jamming while contact remains limited or uneven.
- Object-dependent effects: Pressure-gauge and gallon-jug failures were associated with rigid, asymmetric shapes that changed contact substantially across approach orientations.Their membranes could deform around the initial contact without moving the object into a better position.
- Scope of the result: The paper sheet was tested over ±5° rather than ±10° because the low-height approach risked gripper–table collision.Its reliable performance therefore applies only within the reduced collision-free orientation range.
- Object-dependent effects: Marbles, plastic cups, and T-shirts remained easier because their geometry allowed more consistent contact across tested orientations.Pens and cables were intermediate cases, with some larger-angle acquisition failures from one-sided contact.
- Scope of the result: Three trials per offset condition and a tabletop evaluation limit interpretation primarily to the tested conditions and objects.The object set was not a standardised benchmark, and cluttered or constrained approaches were not evaluated.
B. Mechanical Limitations
The prototype’s mechanical design and modelling impose practical limits on balanced contact, object scope, and the fidelity of geometric predictions. The study nevertheless reports successful hybrid grasps while identifying calibration and environmental constraints.
- Mechanical limitations: Independent finger adaptation is unavailable, so asymmetric or one-sided contact can reduce balanced contact formation.The limitation is particularly relevant to thin objects such as paper.
- Mechanical limitations: The bulky, heavy prototype restricts working space, while non-uniform roller motion may arise under high deformation or off-centre loading.These effects make the design less suitable for large or heavy objects requiring strong, symmetric power-grasp contact.
- Control and sensing limitations: Motor-current feedback cannot identify roller contact or grasp stability, and fixed-duration intake and jamming limit adaptation to unseen objects.Object-specific calibration means the system is a mechanism-validation platform rather than a fully autonomous grasping system.
- Modelling limitations: The geometric analysis omits membrane deformation, granular redistribution, vacuum stiffness, friction, contact pressure, and three-dimensional contact.Its relationships therefore serve as nominal geometric references rather than experimentally validated grasping limits.
- Supported scope: The evaluation achieved 812 successful grasps across 840 trials, while the hybrid ablation condition achieved 80/81 successes.The conclusion characterises these results as mechanism-level evidence for combining active intake with post-contact stiffening.