Shell-Type Soft Jig for Robotic Disassembly

• The paper demonstrates that the shell-type soft jig achieves up to 25° tolerance in pull-out disassembly by combining a rigid shell and inflatable soft chambers for secure, compliant holding.
• It integrates a bottom jamming system with lateral centering via pressurized silicone membranes, offering enhanced robustness against positioning and alignment errors.
• Experimental results indicate stable force profiles and repeatable performance across varied objects, highlighting its potential for adaptive, non-damaging robotic disassembly.

Searching arXiv for the main paper and closely related soft-jig work to ground the article. A shell-type soft jig is a flexible holding tool for robotic disassembly that combines a rigid surrounding shell, inflatable soft side chambers, and a bottom jamming jig to hold assembled objects during pull-out operations such as extracting a shaft from a bearing or separating plugged-in components. It is intended to hold the base object securely enough that a robot can remove a target subcomponent, while remaining compliant enough to tolerate positioning error and avoid damage. The defining idea is soft caging inside a rigid shell: the object can be inserted with clearance, initially supported from below, and then centered and softly fixed by inward-bulging silicone membranes pressurized within the shell (Kiyokawa et al., 17 Sep 2025).

1. Definition and task setting

The shell-type soft jig was proposed for robotic disassembly, specifically for pull-out operations in which a robot grasps and transports an assembled object, places it into a jig, releases and re-approaches, grasps the component to be removed, relies on the jig to secure the remaining object, and then performs a pull-out motion (Kiyokawa et al., 17 Sep 2025). The canonical example is extracting a shaft from a bearing, but the reported experiments span ten assembled objects: shaft–bearing, motor–pulley, USB–adapter, LAN–hub, AC–switch, pulley–shaft, AC–adapter, wire–board, battery–charger, and USB–computer connector (Kiyokawa et al., 17 Sep 2025).

The assumed system consists of a single-arm manipulator, a two-finger gripper, and a pre-planned disassembly trajectory, with the jig improving robustness of execution rather than solving planning itself (Kiyokawa et al., 17 Sep 2025). This division of roles is important: the shell-type soft jig is not a planning method or a perception algorithm, but a fixturing device meant to reduce sensitivity to recognition, planning, and control errors (Kiyokawa et al., 17 Sep 2025).

The motivation follows from the limitations of conventional rigid fixtures, vises, and clamps. These typically must be designed for each object geometry, are not easily reusable across many product forms, rigidly constrain pose so that small perception or execution errors can cause failure, can induce excessive forces or slipping when pull-out direction is imperfectly aligned, and may damage fragile housings or internal parts because they are hard and unforgiving (Kiyokawa et al., 17 Sep 2025). The target, therefore, is not merely strong holding, but holding that is shape-adaptive, soft, self-aligning, and robust to recognition, planning, and control errors (Kiyokawa et al., 17 Sep 2025).

A broader context exists in adaptive fixturing by jamming-based soft jigs. "Soft-Jig-Driven Assembly Operations" introduced a jamming-based general-purpose assembly fixture whose malleable membrane deforms around a part and then rigidifies under vacuum (Kiyokawa et al., 2020). "Soft-Jig: A Flexible Sensing Jig for Simultaneously Fixing and Estimating Orientation of Assembly Parts" extended the soft-jig concept by integrating optical sensing into a bead-and-oil membrane structure (Sakuma et al., 2021). The shell-type soft jig differs from these mainly in its explicit surrounding shell and balloon-based lateral centering for pull-out disassembly (Kiyokawa et al., 17 Sep 2025).

2. Structural concept and caging principle

The shell-type soft jig combines two subsystems: a bottom jamming jig and an outer shell with inflatable soft chambers (Kiyokawa et al., 17 Sep 2025). The bottom jamming jig is used first to support and fix the object after placement, using depressurization of a deformable membrane in the style of prior soft-jig concepts (Kiyokawa et al., 17 Sep 2025). The outer subsystem consists of a rigid shell surrounding the object laterally, with air chambers inside the shell covered by silicone membranes; when pressurized, these membranes bulge inward and press the object from multiple directions (Kiyokawa et al., 17 Sep 2025).

It is called “shell-type” because the object is surrounded by a shell-like enclosure rather than merely supported from below or pinched from two sides (Kiyokawa et al., 17 Sep 2025). The shell provides a geometric boundary, lateral guidance, and multi-directional inward pressing when the balloons inflate (Kiyokawa et al., 17 Sep 2025). This enclosure differentiates the design from a plain jamming pad or a vise (Kiyokawa et al., 17 Sep 2025).

The placement and holding sequence is: the object is inserted into the shell cavity with enough clearance to avoid rigid-shell contact, the bottom jamming jig provides initial fixation, the robot releases and re-grasps the removable component, the side balloons inflate, the silicone membranes bulge inward, and contact from several directions guides the object toward the center and into an upright posture (Kiyokawa et al., 17 Sep 2025). The shell thus acts not only as a clamp but as a centering enclosure (Kiyokawa et al., 17 Sep 2025).

The paper formulates this using caging-related conditions. It defines

$C_\mathrm{free-ICS} \neq \emptyset,$

$q_\mathrm{obj} \in C_\mathrm{free-ICS},$

$C^{'}_\mathrm{free-ICS} = \emptyset.$

Here, $q_\mathrm{obj}$ is the object configuration, $C_\mathrm{free\text{-}ICS}$ is the space where there is no rigid contact, and $C^{'}_\mathrm{free-ICS}$ is the space where there is no soft contact (Kiyokawa et al., 17 Sep 2025). The intended interpretation is that the rigid shell leaves a nonempty free configuration space for insertion, but after inflation there is no free configuration that avoids soft contact, so the object is always enclosed by the membranes (Kiyokawa et al., 17 Sep 2025). This is the formal basis for the jig’s caging-based holding (Kiyokawa et al., 17 Sep 2025).

A plausible implication is that the shell-type soft jig occupies an intermediate category between rigid form closure and unconstrained soft support. The shell contributes boundary geometry and gross pose limitation, while the inflatable membranes eliminate residual internal freedom without imposing hard, precisely aligned contact.

3. Mechanical operation and contact behavior

The central mechanics idea is soft caging within a rigid shell (Kiyokawa et al., 17 Sep 2025). A rigid shell alone may leave an object with a free internal pose and uncertain alignment, so it can still shift during pull-out (Kiyokawa et al., 17 Sep 2025). By contrast, when the soft chambers inflate, they occupy the free space between object and shell, create distributed contact around the object, eliminate free internal motion, passively align the object toward the center and upright posture, and provide resisting forces during pull-out while allowing slight compliance (Kiyokawa et al., 17 Sep 2025).

The paper does not provide detailed force equations, constitutive membrane equations, or explicit friction models (Kiyokawa et al., 17 Sep 2025). Its mechanical description remains qualitative but specific. The inflated silicone membranes create distributed normal contact forces over multiple sides of the object; compliance allows the contact area to conform to local geometry; distributed contact increases stability relative to narrow hard-jaw contact; the object can undergo small passive pose adjustments under load during extraction rather than abrupt force spikes; this helps the constrained part follow the pulling direction enough to avoid gripper slip or jamming; and the surrounding shell geometry prevents large excursions or tipping, keeping the object approximately centered and upright (Kiyokawa et al., 17 Sep 2025).

This hybridization of rigid and soft functions has precedents in soft robotics more broadly. A lobster-inspired hybrid actuator encloses a soft chamber with articulated rigid shells so that the shell protects the actuator, constrains motion, and makes behavior easier to analyze (Chen et al., 2020). Although that work addresses bending actuation rather than fixturing, it similarly treats the shell as the motion-programming structure and the soft interior as the force-generating element (Chen et al., 2020). The shell-type soft jig applies an analogous division of roles to disassembly holding (Kiyokawa et al., 17 Sep 2025).

Adaptation to different shapes arises from membrane deformation, multiple contact modules, and the use of a generic cavity rather than object-specific machined nests (Kiyokawa et al., 17 Sep 2025). Secure holding comes from enclosure from multiple directions, increased contact area due to membrane conformity, centering by symmetric inward pressure, bottom support from the jamming jig, and caging-like elimination of free motion after inflation (Kiyokawa et al., 17 Sep 2025). Softness comes from silicone membranes rather than hard jaws, pneumatic pressure rather than rigid over-constraint, and passive deformation during load (Kiyokawa et al., 17 Sep 2025).

This mechanical combination is central to the intended tradeoff: enough force and enclosure to hold during disassembly, but enough compliance to tolerate error and avoid component damage (Kiyokawa et al., 17 Sep 2025).

4. Design, materials, and implementation

The soft chamber material is Dragon Skin 10 silicone rubber from Smooth-On, chosen for its relatively high tensile strength and 10% modulus of elasticity (Kiyokawa et al., 17 Sep 2025). The rigid shell is a separate rigid structure, with resin structures mentioned in the fabrication description (Kiyokawa et al., 17 Sep 2025).

The paper reports trial-and-error dimensions for the shell/chamber structure: width of the inner cavity of the shell $4.0\,\mathrm{mm}$, wall thickness $2.0\,\mathrm{mm}$, and height $62\,\mathrm{mm}$ (Kiyokawa et al., 17 Sep 2025). These dimensions affect expansion amount, contact area, membrane stiffness, and maximum deformation (Kiyokawa et al., 17 Sep 2025).

The number of side modules is four, chosen as the maximum number that could simultaneously contact the object surface under the task and size constraints; more modules could improve holding, but would increase pneumatic complexity and reduce room for expansion (Kiyokawa et al., 17 Sep 2025). Fabrication is modular: rigid shell and soft chamber parts are manufactured separately, the soft chambers are integrated into the rigid shell, pneumatic inlets are embedded, and the shell module is combined with the bottom jamming jig (Kiyokawa et al., 17 Sep 2025). This supports individual module replacement, adjustment of module count for different object sizes, and easier scaling than monolithic fabrication (Kiyokawa et al., 17 Sep 2025).

The pneumatic system uses an external compressor, electro-pneumatic proportional valves, a DC power supply, a control unit, pneumatic routing to the shell chambers, and a vacuum pump for the bottom jamming jig (Kiyokawa et al., 17 Sep 2025). Chamber pressure can be precisely regulated, though the actual pressure values used in experiments are not reported (Kiyokawa et al., 17 Sep 2025). The behavior is therefore best described as quasi-passive: pneumatically actuated but not closed-loop contact-controlled (Kiyokawa et al., 17 Sep 2025).

Two robot setups were used. Robustness evaluation employed a KUKA LBR iiwa 14 R820 with a Robotiq Hand-E. Holding-force evaluation used a UR5e with a Robotiq 2F-85, relying on the UR5e’s integrated 3-axis wrist force/torque sensor, with force data recorded through socket communication with an external PC (Kiyokawa et al., 17 Sep 2025).

Related soft-jig work clarifies the design lineage. "Soft Regrasping Tool Inspired by Jamming Gripper" used Dragon Skin FX-Pro, a 1 mm membrane, 1 mm glass beads, a bag volume of approximately $296\,\mathrm{cm}^3$, a surface curvature radius of about $q_\mathrm{obj} \in C_\mathrm{free-ICS},$0, and a vacuum level of roughly $q_\mathrm{obj} \in C_\mathrm{free-ICS},$1 to create reconfigurable cavities for regrasping (Kiyokawa et al., 17 Sep 2025). "Soft-Jig-Driven Assembly Operations" used a 160 mm diameter jamming-based soft jig with a 1 mm Dragon Skin FX-Pro membrane and approximately $q_\mathrm{obj} \in C_\mathrm{free-ICS},$2 vacuum (Kiyokawa et al., 2020). These works indicate that shell-type fixturing emerged within a broader program of general-purpose soft fixturing, though the shell-type soft jig replaces cavity formation by shell-guided pneumatic centering (Kiyokawa et al., 17 Sep 2025).

5. Experimental methodology and comparative evaluation

The shell-type soft jig was evaluated against two baselines: a vise and a jamming-gripper-inspired soft jig (Kiyokawa et al., 17 Sep 2025). The ten test objects were chosen to cover varied geometry and connection conditions relevant to disassembly (Kiyokawa et al., 17 Sep 2025).

Robustness was measured by allowable angular deviation from the ideal extraction direction (Kiyokawa et al., 17 Sep 2025). The robot executed pull-out trajectories with angular deviations of $q_\mathrm{obj} \in C_\mathrm{free-ICS},$3, $q_\mathrm{obj} \in C_\mathrm{free-ICS},$4, $q_\mathrm{obj} \in C_\mathrm{free-ICS},$5, $q_\mathrm{obj} \in C_\mathrm{free-ICS},$6, $q_\mathrm{obj} \in C_\mathrm{free-ICS},$7, and $q_\mathrm{obj} \in C_\mathrm{free-ICS},$8, and the outcome was recorded as success or failure (Kiyokawa et al., 17 Sep 2025). From this, the maximum allowable angular deviation was identified for each object/jig combination (Kiyokawa et al., 17 Sep 2025).

Holding performance was evaluated for three representative objects—shaft–bearing, motor–pulley, and USB–adapter—using 10 trials under angular deviations of $q_\mathrm{obj} \in C_\mathrm{free-ICS},$9, $C^{'}_\mathrm{free-ICS} = \emptyset.$0, and $C^{'}_\mathrm{free-ICS} = \emptyset.$1 (Kiyokawa et al., 17 Sep 2025). The measured quantity was the absolute value of vertical wrist force during pull-out, plotted as mean and standard deviation over the 10 trials (Kiyokawa et al., 17 Sep 2025). The intended interpretation was that large abrupt peaks followed by return to zero indicate failure or slip, whereas moderate sustained force with low variability indicates stable extraction (Kiyokawa et al., 17 Sep 2025).

The paper does not provide separate numerical metrics for alignment accuracy in millimeters or degrees, damage occurrence rate, or insertion force (Kiyokawa et al., 17 Sep 2025). These aspects are instead assessed qualitatively from success under misalignment, observed stable centering and upright guidance, absence of excessive force in wrist-force measurements, and discussion of failure cases (Kiyokawa et al., 17 Sep 2025).

A concise comparison of the principal robustness outcomes is useful:

Fixture type	Typical robustness outcome	Representative limitation
Vise	Usually $C^{'}_\mathrm{free-ICS} = \emptyset.$2 to $C^{'}_\mathrm{free-ICS} = \emptyset.$3	Rigid over-constraint and gripper slip
Jamming jig baseline	Often up to $C^{'}_\mathrm{free-ICS} = \emptyset.$4	Low membrane stiffness, limited stability
Shell-type soft jig	Up to $C^{'}_\mathrm{free-ICS} = \emptyset.$5 for six objects	Geometry- and force-capacity-dependent failures

This table compresses reported outcomes only; the detailed object-wise differences remain essential for interpretation (Kiyokawa et al., 17 Sep 2025).

6. Results, limitations, and significance

The most important quantitative result is robustness to extraction-direction error. The vise baseline generally succeeded only under near-perfect alignment. Representative results were: shaft–bearing, motor–pulley, and USB–adapter successful at $C^{'}_\mathrm{free-ICS} = \emptyset.$6 and failing beyond; LAN–hub successful only at $C^{'}_\mathrm{free-ICS} = \emptyset.$7; AC–adapter successful at $C^{'}_\mathrm{free-ICS} = \emptyset.$8; wire–board successful only at $C^{'}_\mathrm{free-ICS} = \emptyset.$9; battery–charger successful at $q_\mathrm{obj}$0; and USB–computer successful at $q_\mathrm{obj}$1 (Kiyokawa et al., 17 Sep 2025). Failures were attributed to rigid over-constraint, so that misalignment caused object slip from the gripper (Kiyokawa et al., 17 Sep 2025).

The jamming-jig baseline improved tolerance somewhat but remained limited. Representative outcomes were success up to $q_\mathrm{obj}$2 for shaft–bearing, motor–pulley, AC–adapter, battery–charger, and USB–computer; only up to $q_\mathrm{obj}$3 for USB–adapter; and only $q_\mathrm{obj}$4 for LAN–hub and wire–board (Kiyokawa et al., 17 Sep 2025). The limitation was attributed to low membrane stiffness and insufficient holding stability at larger angular errors (Kiyokawa et al., 17 Sep 2025).

The shell-type soft jig achieved the best tolerance. It succeeded up to $q_\mathrm{obj}$5 for shaft–bearing, motor–pulley, USB–adapter, AC–adapter, battery–charger, and USB–computer; up to $q_\mathrm{obj}$6 for LAN–hub and wire–board; and did not succeed for AC–switch, while pulley–shaft also failed with the proposed jig (Kiyokawa et al., 17 Sep 2025). Compared with the vise’s typical $q_\mathrm{obj}$7–$q_\mathrm{obj}$8 range and the jamming jig’s frequent $q_\mathrm{obj}$9 limit, the shell-type jig often maintained success to $C_\mathrm{free\text{-}ICS}$0 (Kiyokawa et al., 17 Sep 2025).

Force measurements reinforce that interpretation. For the vise at $C_\mathrm{free\text{-}ICS}$1 angular deviation, the measured vertical force sharply increased and then quickly returned to near $C_\mathrm{free\text{-}ICS}$2, indicating buildup under misalignment followed by slip and failure (Kiyokawa et al., 17 Sep 2025). With the proposed jig, approximately $C_\mathrm{free\text{-}ICS}$3 of force was applied during extraction, no excessive force was observed, all 10/10 trials were successful, and standard deviation was relatively small (Kiyokawa et al., 17 Sep 2025). The reported interpretation is that the object could adapt slightly under the shell balloons, enough force was sustained for extraction, repeatability was high, and the abrupt overconstraint failure seen in the vise was avoided (Kiyokawa et al., 17 Sep 2025).

The failure analysis is equally important. LAN–hub failed at larger angular errors because the gripper missed the locking tab, showing that the jig cannot compensate for connector-specific unlocking requirements (Kiyokawa et al., 17 Sep 2025). AC–switch exceeded the jig’s holding-force capacity, indicating a hard limit of the balloon-side holding mechanism (Kiyokawa et al., 17 Sep 2025). Pulley–shaft failed because the large pulley prevented proper balloon contact with the shaft and the high center of gravity caused tipping before inflation (Kiyokawa et al., 17 Sep 2025). Wire–board failed because slight tilting bent the wire inside the socket, preventing removal (Kiyokawa et al., 17 Sep 2025).

These limitations clarify the scope of the concept. The shell-type soft jig is particularly suitable for hand-sized objects, pull-out disassembly tasks, assemblies that can be extracted without unlocking complex latches, objects with accessible side surface for balloon contact, and tasks where some compliance during extraction is beneficial (Kiyokawa et al., 17 Sep 2025). It assumes that the object fits inside the shell, the shell can contact it from multiple sides, extraction force is not too high, geometry does not block balloon contact, disassembly is mainly a pull-out action, and the object remains reasonably upright before stabilization (Kiyokawa et al., 17 Sep 2025).

The authors explicitly identify future directions: improved balloon design, better durability, different friction materials, greater modular scalability, improved pressure regulation and control strategies, and adapting modular balloon units to different object sizes (Kiyokawa et al., 17 Sep 2025). A plausible implication is that future work will need to address the central unresolved tradeoff already visible in the experiments: increasing holding force without losing the compliance that gives the jig its tolerance and damage-mitigating character.

In the broader literature, shell-type soft jigging can be understood as one branch of a more general shift from rigid, object-specific fixtures toward reusable compliant fixtures. Jamming-based soft jigs support assembly and regrasping by conformal support or cavity formation (Kiyokawa et al., 2020, Kiyokawa et al., 17 Sep 2025), sensing-enabled soft jigs add pose estimation to fixation (Sakuma et al., 2021), and shell-type soft jigs extend this trajectory by adding rigid-shell caging and inflatable lateral centering specifically for disassembly (Kiyokawa et al., 17 Sep 2025). Within that progression, the shell-type soft jig’s distinctive contribution is to show that a shell enclosure plus soft balloons can materially expand tolerance to pull-out misalignment while preserving gentle, shape-adaptive holding (Kiyokawa et al., 17 Sep 2025).