Hybrid Roller-Jamming Gripper for Object Acquisition and Retention Under Pose Uncertainty
Listen
Radio episode about this paper
Transcript
Introduction to the show: ident: Robotics Radio. Generated commentary on the latest robotics and control papers.
Rosa: I'm Rosa.
Dev: And I'm Dev, here with Taro.
Rosa: We're looking at the paper: "Hybrid Roller-Jamming Gripper for Object Acquisition and Retention Under Pose Uncertainty".
Dev: Here's the gist of the paper. Hybrid roller-jamming grippers are presented to address object acquisition and retention challenges that arise from pose uncertainty in household manipulation tasks.
Rosa: The proposed mechanism integrates active object intake via inward roller rotation with post-capture stabilization through vacuum-induced granular jamming.
Dev: This hybrid strategy is designed to improve grasp acquisition and retention after imperfect first contact.
Rosa: Gripper Design and Architecture The proposed gripper adopts a two-finger architecture composed of a base module and two opposing roller-finger modules.
Title and authors: Rosa: So we're looking at this paper called "Hybrid Roller-Jamming Gripper for Object Acquisition and Retention Under Pose Uncertainty." It's about tackling those tricky moments when you pick something up, especially if it's not perfectly lined up.
Dev: Right, Rosa. The core idea here is combining two different actions: active rolling for getting the object in place and then jamming with vacuum to keep it there once you’ve got it. That hybrid strategy aims to fix problems where things get messy during the initial contact phase of a manipulation task.
Taro: I'm curious how this hybrid approach handles those moments when the object isn't quite where you expect it at first, like being off-centre or having a slight tilt.
Rosa: Well, what the paper presents is that by using inward roller rotation, you can actually draw the object closer to the centre region of those rollers before you apply the jamming force. That should help increase the contact area available right when you need it most.
Dev: And then, once that engagement happens, you switch to vacuum-induced granular jamming. This stiffens up those rollers and helps stabilize that grasp, which is something roller-based grippers usually struggle with on their own because they don't naturally increase stiffness after the initial touch.
Taro: So it sounds like rolling sorts out the acquisition problem—getting close—and jamming solves the retention problem once you’re engaged. How does this play out when things are off in terms of orientation, not just position?
Rosa: The paper found that these results were more sensitive to three dimensional contact conditions than they were to just planar position offsets. It suggests that for orientation issues, the rolling motion helps bring the object into a better alignment before jamming is applied.
Dev: And when we look at the performance metrics, they tested eight hundred forty grasp trials overall, with a success rate of ninety-six point seven percent in the main evaluation under those offset conditions. But it's important to see where it fails when you strip things down to just one mechanism.
Taro: I saw the ablation results were quite telling. When they only used the roller-only condition, it only got fifty-four out of eighty-one successes, and jamming only achieved twenty-four out of those same eighty-one trials. That clearly shows that rolling is mainly for acquisition and contact formation, while jamming is key for improving post-capture retention.
Rosa: Exactly. And the paper pointed out that for some objects, like a gallon jug they tested, the jamming-only condition actually performed better than the roller-only condition. That tells us we need to be smart about when to use which mechanism depending on what we're holding.
Title and authors: Dev: It also highlighted that for simpler things like a pen or cable, rolling alone was effective, suggesting that object properties really dictate whether you lean into the rolling part or rely more on the jamming part. That adaptability is what makes this hybrid strategy interesting for real-world use.
Taro: But looking at the design itself, they mentioned some limitations that we have to keep in mind. They noted that their geometric analysis only represents rigid links and undeformed circular envelopes for those rollers, so it doesn't perfectly model how the membrane actually deforms or how the granular material redistributes during jamming.
Rosa: And they also flagged that the current prototype is relatively bulky and heavy, which restricts the available working space around the object. Plus, they mentioned that their control strategy relies partly on time-driven control in certain phases, which isn't ideal for a fast real-time loop.
Dev: Those are fair points about scaling up. They also mentioned that because it's a prototype, the system is limited by object-specific calibration and doesn't fully account for the complexity of real objects. The authors themselves pointed out that future work should focus on integrating tactile sensing to support event driven state transitions, which would help with those calibration issues.
Taro: So the paper points toward a path forward where we move away from these fixed timing controls and towards something more reactive based on actual sensor feedback, like tactile or vacuum pressure readings. That seems like a natural progression for making it robust enough for cluttered or obstacle constrained environments.
Rosa: Exactly. The implication here is that if we can get the system to react to the contact state rather than relying on pre-set durations, we might solve those issues with pose uncertainty much more reliably in actual household manipulation tasks.
Dev: Looking at this paper, "Hybrid Roller-Jamming Gripper for Object Acquisition and Retention Under Pose Uncertainty," it shows a clear way to layer active acquisition with passive stabilization. It proves that combining inward roller rotation to draw the object in with vacuum-induced granular jamming for stiffness is a viable mechanism for handling imperfect first contacts.
Taro: I think what this means for us is that we can design grasping systems where the first phase of interaction focuses purely on bringing the object into a stable engagement, and the second phase focuses entirely on maintaining that engagement through physical stiffening.
Title and authors: Rosa: And it does give us concrete numbers to work with, like that ninety-six point seven percent success rate in their main evaluation across those planar and orientation offset trials. It gives us a baseline to measure against other approaches we might test later.
Dev: We have to keep the real-time aspects in mind though; since they used fixed durations for intake and jam, that suggests the hardware control needs refinement if we want it running on a tight loop, like what I usually deal with.
Taro: And that's where the next step is crucial: moving towards event driven state transitions. If the system can sense when contact quality changes, it can decide whether to keep rolling or start jamming sooner, rather than following a predetermined timer.
Rosa: So that’s a good direction for future research—making the control logic smarter about when to switch from acquisition mode to retention mode based on what the sensors actually tell it.
Dev: That sounds like where the safety case gets stronger—when you have better real-time feedback, you know exactly how much latency or failure rate you can tolerate before a grasp fails completely.
Taro: So that's the path: using tactile and vacuum pressure sensing to support those state transitions, moving away from the time driven control mentioned in the paper.
Rosa: It’s a solid piece of work showing that for household manipulation under pose uncertainty, you don't need one perfect mechanism; you need a coordinated hybrid strategy like this roller-jamming system.
Dev: The paper on "Hybrid Roller-Jamming Gripper for Object Acquisition and Retention Under Pose Uncertainty" gives us a practical blueprint for how to address the specific challenges of off-centre or partial initial contact using complementary physical actions.
Taro: It really emphasizes that acquisition and retention aren't separate goals; they are sequential phases of engagement, and this gripper models that sequence very well with its rolling followed by jamming mechanism.
Rosa: So, to wrap up the main idea: the hybrid roller-jamming gripper achieves success across eight hundred forty trials by using rolling to improve engagement during acquisition and vacuum jamming to stiffen the grasp for retention.
Dev: It’s a promising approach for when you're dealing with position and orientation offsets because it addresses both sides of the problem sequentially.
Taro: For me, it highlights how important it is to consider object geometry when designing these systems, as we saw that sensitivity to three dimensional contact conditions versus planar position offsets.
Rosa: It’s a good summary of what this work achieves, showing how combining active intake and post-capture stiffening can be a way forward for robust grasping systems.
The paper's summary: Rosa: So we're looking at this paper called "Hybrid Roller-Jamming Gripper for Object Acquisition and Retention Under Pose Uncertainty." It basically shows how combining active rolling to get an object in place with vacuum jamming to keep it there helps when you don't know exactly where something is.
Dev: Right, the authors are proposing this hybrid strategy because they see that if you just rely on one thing, like just rolling, it struggles when the object is slightly off centre.
Taro: And if you only jam, it doesn't help with the initial pickup phase at all. This hybrid approach is designed to handle those first messy contacts gracefully.
Rosa: The main point they make is that inward roller rotation draws the object toward the center region of those rollers before you even try to jam it down.
Dev: So, if you think about acquisition, rolling helps pull it in, and then jamming takes over for stabilization once contact is made. It’s a two-step process.
Taro: The results show that rolling mainly supports getting the object close and forming initial contact, while the jamming part is what really improves how long you can keep it there after capture.
Rosa: They did some ablation tests that were pretty telling about this division of labor. The roller-only condition only got a few successes, while the jamming-only condition performed much worse overall.
Dev: It shows that for many things, you need both actions working together to get a good result. And they found specific objects where one action was clearly better than the other depending on the object itself.
Taro: For instance, they learned that some objects really benefit from jamming even if the rolling part doesn't do much work for acquisition. That means we have to be smart about which tool we deploy first based on what we’re holding.
Rosa: The paper also looked at how orientation—that tilt or angle—affects things, and they found that the system is more sensitive to those three dimensional contact conditions than it is to just a simple flat position offset.
Dev: That means for orientation problems, the rolling motion helps align things before jamming stabilizes them; you can't just rely on jamming when you're significantly tilted.
Taro: The numbers they got across eight hundred forty trials showed a success rate of about ninety seven point seven percent overall in their main evaluation, which is pretty solid for this kind of complex manipulation.
Rosa: But the paper isn't perfect either; they admit that their geometric analysis is just a simplified model, and it doesn't perfectly capture how those rollers actually deform or how the granular material moves during jamming.
Dev: And they also mentioned that the current prototype is pretty bulky and heavy, which limits where you can even put it relative to the object you’re trying to grab.
Taro: Because of those limitations, they pointed toward future work that needs to integrate tactile sensing and vacuum pressure sensors so the system can make decisions based on what it actually feels in real time.
Rosa: That’s where things get interesting for me—moving away from fixed timers and towards event driven state transitions based on actual sensor feedback.
Dev: If the system can sense when the contact quality changes, it can decide whether to keep rolling or start jamming sooner, instead of just following a pre-set duration.
Taro: That sounds like a major step for real world application; making the control logic reactive to the physical situation rather than relying on timing.
Rosa: So this paper gives us a blueprint for how to layer acquisition with retention using two distinct physical actions, and it points toward smarter, sensor-driven control for future versions of these grippers.
The paper's improvements: Taro: So, to wrap up on that, we’re talking about how they suggest improving this hybrid system for real use outside of a perfect lab setting.
Rosa: Right, the authors look at what they found and say that if you integrate tactile and vacuum pressure sensing, you can support event driven state transitions instead of relying on fixed timings for intake and jamming.
Dev: That means the gripper can actually sense when the contact quality is good enough to switch from rolling to jamming automatically, which would drastically improve robustness in a real-time loop.
Taro: If it can react based on what it feels, that addresses the issue of object specific calibration because it doesn't need perfect pre-calibration for every single item.
Rosa: That makes sense; if the AI can sense the change in surface texture or pressure, it knows exactly when to apply that granular jamming force.
Dev: It also means we get a much better safety case because you know the system is reacting to physical events rather than just following a pre-programmed schedule.
Taro: I think that’s what makes the difference between this prototype and something ready for real household manipulation; it moves from being pre-set to being adaptive.
Rosa: It shifts the focus from perfect geometric modeling to smart sensing, which is exactly what field robotics needs when things get messy outside of a controlled environment.
Dev: So, if we can solve that sensor integration problem, we could potentially design these grippers for a much wider range of objects without needing custom calibrations for every single one.
Taro: That opens up the door to tackling those cluttered or obstacle constrained environments where the object position is constantly changing and you need immediate adjustments.
Rosa: It’s about making the system less dependent on perfect initial geometry and more dependent on real time feedback, which is a huge step for autonomous systems in daily life.
Dev: So we're moving from a fixed control strategy to an intelligent, reactive one that uses sensor data to dictate the sequence of acquisition and retention actions.
Taro: That sounds like the next big hurdle for making these types of grippers truly reliable in dynamic settings where things are always misbehaving.
Conclusion: Rosa: So, to wrap up on this paper "Hybrid Roller-Jamming Gripper for Object Acquisition and Retention Under Pose Uncertainty," we've seen how combining rolling for acquisition and jamming for retention works well under pose uncertainty.
Dev: Right, the core message is that this hybrid approach tackles the problem of imperfect first contact by separating those two functions into distinct physical actions.
Taro: I think what this means for autonomy is that we can design systems where the initial engagement phase is purely about getting close, and then a separate mechanism handles stabilizing that grasp later.
Rosa: Exactly, and they showed through their results across eight hundred forty trials that the success rate was quite high, around ninety seven point seven percent overall.
Dev: But remember the caveat they mentioned—the geometric analysis is just a simplified model of rigid links and undeformed rollers, so we have to be careful about how it performs on actual deformable objects.
Taro: That's true; the limitations are real when you move this from the bench to a messy room where things aren't perfectly circular or rigid.
Rosa: For us in robotics, it gives us a solid baseline for how these complementary physical actions work together in practice.
Dev: And what I want to stress is that their focus on event driven state transitions is the real next step, moving control away from fixed durations toward sensor feedback like tactile pressure and vacuum readings.
Taro: If we can get that level of reactive control, those limitations they flagged about object specific calibration should become much less of a problem.
Rosa: It’s promising because it shows a path forward for grasping under position and orientation offsets without needing perfect pre-knowledge of the object's shape.
Dev: We'll keep an eye on how their future work integrates those tactile sensors, because that’s where the real robustness will come from in real world loops.
Yijie Ren, Guillaume Gourmelen, Hiroyasu Iwata
Waseda University
cs.RO
Submitted: 2026-08-21
Updated: 2026-08-21
Comments: 11 pages, 16 figures, under review
Journal ref: IEEE Access ( Volume: 14) 2026
DOI: 10.1109/ACCESS.2026.3738462
License: http://creativecommons.org/licenses/by/4.0/
Importance score: 77/100
The gist: Hybrid roller-jamming grippers are presented to address object acquisition and retention challenges that arise from pose uncertainty in household manipulation tasks.
Key concepts
- Active Object Intake (Inward Roller Rotation)
- This mechanism involves rotating the spherical rollers inward after initial contact. This rotation actively draws the object towards the center of the gripper, increasing contact area and improving engagement when objects are not perfectly centered.
- Vacuum-Induced Granular Jamming
- This technique uses vacuum pressure to create a stiffening effect on the rollers. Once an object is captured, this jamming mechanism locks the rollers into place, providing a strong mechanical grip that stabilizes the grasp and prevents slippage during holding.
- Hybrid Strategy for Pose Uncertainty
- The system employs a dual strategy: rolling is used primarily for acquisition (getting the object close), while jamming is used for retention (securing the object). This hybrid approach allows the gripper to compensate effectively when objects are positioned at angles or offsets from its ideal starting point.
Terminology
Summary
Hybrid roller-jamming grippers are presented to address object acquisition and retention challenges that arise from pose uncertainty in household manipulation tasks. p. 0 1
Hybrid Roller-Jamming Gripper for Object
Acquisition and Retention Under Pose Uncertainty
Yijie Ren, Guillaume Gourmelen, and Hiroyasu Iwata
Abstract—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 roller-only and jammingonly 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.Where the paper says it
Where the paper says it
Where the paper says it
Gripper Design and Architecture
The proposed gripper adopts a two-finger architecture composed of a base module and two opposing roller-finger modules. This system combines active rolling during acquisition and granular-jamming stiffening during retention. p. 1 Overall System Overview
The proposed gripper adopts a two-finger architecture composed of a base module and two opposing roller-finger modules, as shown in Fig. At the system level, the
design combines two functions within the same end-effector:
active rolling during acquisition and granular-jamming stiffening during retention. System Primary purpose Acquisition mechanism Retention mechanism Actuation architecture
RBO Hand 2 [3] Dexterous grasping through
compliance and underactuation
Pneumatic bending of compliant fingers and palm
around the object
Compliant enveloping contact maintained by pneumatic pressure
Seven pneumatic actuators grouped into
four control channels
Two-fingered hand with
jamming finger tips [10]
Adaptive grasping with
variable-stiffness fingertips
Finger closure with passive
fingertip conformity
Vacuum-jammed fingertips Four servomotors and a pneumatic jamming system
Simple Belted ParallelJaw Gripper [16]
Grasp-to-grasp in-hand translation and reorientation
Base parallel-jaw closure Force-controlled jaw clamping with belt and fingersurface contact
Four independent belt motors plus the
base-gripper actuator
Roller Grasper V3 [19] Full 6-DoF in-hand manipulation
Finger closure followed by
rolling pickup
Four-finger rolling contact,
assisted by an active palm
Four fingers with three active DoF each;
12 active DoF
Mecanum-wheel
hand [20]
Omnidirectional in-hand manipulation without finger gaiting
Coupled four-finger opening and closing
Force closure with springloaded wheel constraint
forces
Three-servo coupled reconfiguration and
two wheel-drive motors
Proposed roller-jamming
gripper
Grasp acquisition and retention
under pose offsets
Shared finger closure followed by inward rolling intake
Vacuum-induced granular
jamming of the same rollers
Three active DoF and pneumatic vacuum switching
TABLE II
KEY MECHANICAL SPECIFICATIONS OF THE GRIPPER. Item Value
Overall envelope (W×L×D) 70 × 228 × 177 mm
Total gripper mass 1. An XL320 actuator mounted on
each finger link drives the roller through a 1:20 worm-gear
transmission. This arrangement allows each roller to rotate
inward during object intake and outward during release.Where the paper says it
Where the paper says it
Where the paper says it
Geometric Analysis and Characterization
A simplified front-view model approximates the spherical membrane rollers as undeformed circular envelopes for geometric analysis. The nominal minimum supported circular-feature size is calculated using equation (3) based on the minimum finger angle θmin. p. 3 The fingers
rotate symmetrically about pivots separated by dp, with pivot-to-roller-centre
distance L, roller radius R, and roller-centre spacing dc. Nominal Minimum Supported Circular-Feature Size
To obtain a nominal geometric reference for small-object enclosure, the gripper is considered at its minimum finger angle
θmin, as shown in Fig. For the roller traction test, the load cell was connected by
a string to a 3D-printed spherical target. The rollers were
driven inward with vacuum disabled, and the maximum tensile
force measured before slip was 8. Bench Characterisation
Three types of preliminary bench test were conducted
to characterise the prototype: roller traction, grasp slip-out
resistance, and clamping force. Each test was performed once;
therefore, the reported values represent single-trial maximum
measurements rather than statistical force estimates. The supported
feature remains tangent to both undeformed roller envelopes and the support
plane. Bench Characterisation
Three types of preliminary bench test were conducted
to characterise the prototype: roller traction, grasp slip-out
resistance, and clamping force.Where the paper says it
Where the paper says it
Where the paper says it
Where the paper says it
Experimental Evaluation and Results
The main evaluation comprised 840 grasp trials, including 216 planar-offset trials and 624 orientation-offset trials. The gripper succeeded in 812/840 trials overall, showing a success rate of 96. p. 0 The
main evaluation comprised 840 grasp trials, including 216 planaroffset 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. Furthermore, controlled planar position and orientation offsets
were applied, with each condition repeated three times. The
main evaluation comprised 840 grasp trials, including 216 planaroffset trials and 624 orientation-offset trials. Each condition was repeated
three times, resulting in 78 orientation trials per object. Across
all eight objects, the gripper achieved 597 successes in 624
trials, corresponding to an overall orientation-offset success
rate of 95.Where the paper says it
Where the paper says it
Where the paper says it
Discussion of Performance
The planar-offset results indicate that inward roller motion drew the object closer to the centre region before jamming was applied. The ablation results suggest that rolling mainly supports object acquisition and contact formation, while jamming improves post-capture retention. p. 8 Rolling was therefore more important for gathering these
objects from partial contact. Overall, rolling mainly supports object acquisition and contact formation, while jamming improves post-capture retention. The complete hybrid sequence combines these functions
by first developing contact through rolling and then stabilising
the grasp through jamming. Interpretation of the Orientation Offset Results
The orientation results suggest that the gripper is more
sensitive to three-dimensional contact conditions than to planar
position offsets.Where the paper says it
Where the paper says it
Limitations and Future Work
The present study is limited by object-specific calibration and time-driven control strategies. The geometric analysis represents only nominal references and does not model membrane deformation or granular redistribution accurately. p. 9 It does not model membrane indentation, granular redistribution, vacuum-induced stiffness changes, friction,
contact pressure, or three-dimensional contact. The derived
relationships therefore provide nominal geometric references
rather than experimentally validated grasping limits. Hardware development will focus on reducing the size and mass of the gripper,
improving roller-motion uniformity, and introducing adaptive
or independently compliant finger closure. Tactile and vacuumpressure sensing will also be integrated to support event-driven
state transitions and reduce object-specific calibration.Where the paper says it
Where the paper says it
Conclusion
The hybrid roller-jamming gripper achieved 812 successful grasps across 840 trials, supporting the complementary roles of rolling in improving acquisition and jamming in improving post-capture retention. This approach is promising for grasping under position and orientation offsets. p. 8 Their rigid and asymmetric shapes produced large changes
in contact as the approach orientation varied. Under difficult
offsets, the membranes sometimes deformed around the initial
contact region without moving the object into a better position.Where the paper says it
--- Page 1 ---
<ref:2608.20962#pg2>Hybrid Roller-Jamming Gripper for Object Acquisition and Retention Under Pose UncertaintyThe 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. p. 0 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.Where the paper says it
--- Page 2 ---
A widely adopted strategy for robust grasping across diverse objects is to combine compliance and underactuation. Granular jamming is a representative example of this strategy. p. 0 Furthermore, mechanisms that remain compliant during
approach and then stiffen after contact are particularly attractive for robotic grasping because they combine gentle
initial interaction with improved stability after capture. Granular jamming is a representative example of this strategy.Where the paper says it
--- Page 3 ---
The main objectives of this work are as follows.
--- Page 4 ---
This paper presents a hybrid roller-jamming gripper in which the same spherical membrane rollers actively draw the object inward after contact and subsequently stiffen under vacuum for post-capture retention, as shown in Fig. 1.
--- Page 5 ---
Table I compares its acquisition, retention, and actuation mechanisms with those of other representative grippers and the proposed design.
--- Page 6 ---
The main dimensions and mass of the system are summarised in Table II.
--- Page 7 ---
This type of benchlevel evaluation is consistent with prior work that emphasises explicit external performance metrics for robot end-effectors.
--- Page 8 ---
A trial was classified as successful if the object was acquired, lifted clear of the table, and retained throughout the complete post-grasp motion routine.
--- Page 9 ---
The ablation tests used the pen, gallon jug, and cable.
--- Page 10 ---
The success rate for each test protocol was calculated as the number of successful trials divided by the total number of repeated trials.
--- Page 11 ---
In many successful trials, inward roller motion drew the object closer to the centre region of the spherical membranes, therefore increasing the contact area available before stiffening.
--- Page 11 ---
The gallon jug accounted for the only failure in the planar-offset protocol.
--- Page 12 ---
The paper sheet achieved reliable performance within the reduced ±5◦ orientation range.
--- Page 13 ---
The present prototype remains limited by object-specific calibration, partly time-driven control, a controlled tabletop environment, and idealised geometric modelling.
--- Page 14 ---
The current prototype is relatively bulky and heavy, which restricts the available working space around the object.
--- Page 15 ---
The controller uses motor-current feedback during CLOSE and HOLD, whereas INTAKE and JAM use fixed durations.
--- Page 16 ---
The geometric analysis represents the fingers as rigid links and the spherical membrane rollers as undeformed circular envelopes.
--- Page 17 ---
Future work will evaluate the gripper using standardised object sets, controlled variations in object properties, and cluttered or obstacle-constrained environments.
--- Page 18 ---
Tactile and vacuumpressure sensing will also be integrated to support event-driven state transitions and reduce object-specific calibration.
--- Page 19 ---
The present prototype remains limited by object-specific calibration, partly time-driven control, a controlled tabletop environment, and idealised geometric modelling.
--- Page 20 ---
The controller uses motor-current feedback during CLOSE and HOLD, whereas INTAKE and JAM use fixed durations.
--- Page 14 ---
The paper sheet achieved reliable performance within the reduced ±5◦ orientation range.
--- Page 17 ---
The present prototype remains limited by object-specific calibration, partly time-driven control, a controlled tabletop environment, and idealised geometric modelling.
--- Page 19 ---
Tactile and vacuumpressure sensing will also be integrated to support event-driven state transitions and reduce object-specific calibration.
--- Page 19 ---
The controller uses motor-current feedback during CLOSE and HOLD, whereas INTAKE and JAM use fixed durations.
Green — the paper says it, and the page is printed. Red — no passage carries it, so treat that sentence as unverified. The same check runs on your own text: check a summary against its source, $5.
Improvements for AI systems
-
Active Object Acquisition under Pose Uncertainty: The improved AI system can reliably acquire objects from off-centre or partial initial contact because it uses
inward roller rotation to increase contact and draw the object toward the gripper centre
before jamming. -
Robust Post-Capture Retention: The system gains strong retention after capture because it employs
vacuum-induced granular jamming to stiffen the rollers and stabilise the grasp,
overcoming limitations whereroller-based grippers require sufficient contact before jamming.
-
Pose Uncertainty Handling via Hybrid Strategy: The AI can handle pose uncertainty more effectively by implementing a hybrid strategy where
rolling is used to improve engagement during acquisition, and jamming is used to preserve that engaged state during retention.
-
Mechanism-Level Failure Diagnosis: The system can diagnose why grasping fails under different conditions by comparing the
roller-only and jamming-only conditions achieved 54/81 and 24/81 successes, respectively,
providing evidence on the complementary roles of rolling and jamming. -
Object Property Adaptation: The system can adapt its grasping strategy based on object characteristics; for example, it learns that
the gallon jug performed better with jamming only (ROLLER OFF), whereas the pen and cable performed better with rolling only (JAM OFF).
-
Tilt Compensation in Grasping: The AI can maintain performance under orientation offsets by recognizing the sensitivity of the system, as it learns that
under orientation offsets, contact was one-sided or too shallow for the object to be drawn inward before jamming.
Abstract
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.
Related papers
- Rethinking Actions: Continuity-Aware Priors for World-Model Planning
- MOSAIC: Modular Obstacle and Spatial-Aware Framework for Manipulation In Cluttered Scenes
- SAGAS: Semantic-Aware Graph-Assisted Stitching for Offline Temporal Logic Planning
- DexSeed: Expanding Human Hand-Object Interactions into Robot Demonstrations for Dexterous Policy Learning
- LaST0.5: A Generalist World Action Model with Interaction-Centric Latent Reasoning
- CIALLO: Learning Shared Intent Representations for Cross-Embodiment Manipulation