Soft Jig: Adaptive Fixturing in Robotics
- Soft jigs are deformable, compliant fixtures that adapt contact geometry and mechanical impedance using jamming, pneumatic caging, or friction-based interfaces.
- They enable versatile assembly, regrasping, and disassembly by replacing rigid jigs with sensor-embedded and reconfigurable designs, achieving high success rates in trials.
- They trade off absolute positioning precision for adaptability, leveraging integrated sensing and tunable friction to support diverse robotic manipulation tasks.
Searching arXiv for recent and foundational papers on “soft jig” and closely related fixture concepts. Searching for "soft jig" on arXiv. Searching arXiv for soft jig / soft fixturing / jamming-based fixture papers. A soft jig is a deformable, compliant, or variable-stiffness fixture that holds, aligns, or supports an object by adapting its contact geometry and mechanical impedance to the object or task, rather than by imposing a fixed rigid cavity or clamp. In current arXiv literature, the term encompasses granular-jamming fixtures for assembly, optical sensing fixtures that simultaneously estimate orientation, reconfigurable regrasping tools, textile and membrane interfaces for soft robots, and, in a broader architectural sense, modular jigs whose behavior is defined by finite-state control and demonstration rather than by rigid part-specific programming (Kiyokawa et al., 2020, Sakuma et al., 2021, Sasaki et al., 13 Jun 2025). The common theme is the replacement of dedicated hard fixturing with contact mechanisms that tolerate geometric variation, uncertainty, and deformation while still providing sufficient repeatability for assembly, disassembly, sensing, or transport (Kiyokawa et al., 17 Sep 2025, Kiyokawa et al., 17 Sep 2025).
1. Conceptual scope and relation to rigid fixturing
Conventional rigid jigs in factories are hard, machined fixtures tailored to a specific part geometry. They provide precise locating surfaces and high stiffness, but require custom design and fabrication for each part or posture, which is costly and time-consuming in high-mix, low-volume production (Sakuma et al., 2021, Kiyokawa et al., 2020). Soft jigs were introduced to address this lack of versatility by replacing rigid datum surfaces with deformable membranes, jamming media, compliant shells, or other adaptive interfaces that conform to arbitrary shapes and can be reconfigured without mechanical rebuild (Kiyokawa et al., 2020, Kiyokawa et al., 17 Sep 2025).
The literature uses the term in several related senses. In the most literal sense, a soft jig is a compliant fixture made of silicone membranes, granular media, textiles, or inflatable chambers that directly contacts and holds a part (Sakuma et al., 2021, Kiyokawa et al., 17 Sep 2025). In a second sense, the soft jig is a variable-stiffness support or regrasping aid whose geometry is actively reshaped and then locked, as in jamming-based cavity formation for precise dropping and regrasping (Kiyokawa et al., 17 Sep 2025). In a third, broader sense, the term extends to task-specific robotic jigs whose “softness” lies in modularity, ROS integration, and FSM-based reprogrammability, rather than in continuum mechanics (Sasaki et al., 13 Jun 2025). This suggests that “soft jig” denotes a family of adaptive fixturing strategies rather than a single material platform.
A recurrent contrast with rigid fixtures is that soft jigs prioritize adaptability over maximal intrinsic positioning precision. The 2020 work on soft-jig-driven assembly explicitly characterizes soft jigs as having higher easiness of fixing and versatility, but lower intrinsic parts-positioning precision than rigid jigs (Kiyokawa et al., 2020). Subsequent papers preserve this trade-off while progressively adding sensing, optimized cavity geometry, caging-based alignment, or structured compliance to recover repeatability and robustness (Sakuma et al., 2021, Kiyokawa et al., 17 Sep 2025, Kiyokawa et al., 17 Sep 2025).
2. Physical mechanisms and mechanical models
A dominant mechanism is jamming. In "Soft-Jig: A Flexible Sensing Jig for Simultaneously Fixing and Estimating Orientation of Assembly Parts" (Sakuma et al., 2021), the fixture is a thin silicone membrane filled with transparent beads and optical oil; fixation is achieved by bead-based jamming induced by discharging only oil. In the earlier assembly-oriented soft jig (Kiyokawa et al., 2020), the bag is covered with a malleable silicone membrane of thickness $\SI{1}{mm}$, diameter $\SI{160}{mm}$, filled with $\SI{450}{g}$ of approximately $\SI{1}{mm}$ glass beads in an internal volume of $\SI{296}{cm^3}$, and jammed at about $\SI{90}{kPa}$ confining pressure. In the 2025 soft regrasping tool, a silicone membrane of thickness $1$ mm and Shore A hardness , filled with about of $1$ mm glass beads, is stamped with a triangular-pyramid-shaped tool and then jammed by a pressure drop of about $\SI{160}{mm}$0, leaving a stable cavity for gravity-driven placement (Kiyokawa et al., 17 Sep 2025). Across these systems, the unjammed state provides shape conformity; the jammed state provides load-bearing fixation.
A second mechanism is pneumatic caging. The shell-type soft jig for robotic disassembly places an object on a bottom jamming jig and then encloses it with balloon chambers embedded in a rigid shell (Kiyokawa et al., 17 Sep 2025). The shell diameter is chosen so that an object can initially lie in a rigid-contact-free region, while inflation removes the soft-contact-free region: $\SI{160}{mm}$1 The tested chamber geometry has inner cavity width $\SI{160}{mm}$2 mm, shell wall thickness $\SI{160}{mm}$3 mm, and chamber height $\SI{160}{mm}$4 mm (Kiyokawa et al., 17 Sep 2025). The object is therefore not rigidly clamped into a single pose; rather, it is softly caged, centered, and stabilized under distributed membrane contact.
A third mechanism is frictional soft contact. The deformable tip mount for vine robots is a textile cap whose holding force derives from circumferential contact pressure and friction, modeled as
$\SI{160}{mm}$5
where $\SI{160}{mm}$6 is effective circumferential contact pressure, $\SI{160}{mm}$7 is the friction coefficient, $\SI{160}{mm}$8 is radius, and $\SI{160}{mm}$9 is contact length (Suulker et al., 2024). The cap behaves as a compliant sleeve: it grips by elastic tension rather than by a rigid interface, and its low bending stiffness allows squeezing and bending with the host robot. A related but more explicitly modeled interface appears in the soft-rigid hybrid finger with inflatable silicone pockets (Ly et al., 7 Mar 2026). There, membrane bulging under pressure modifies contact area and stiffness, and the effective friction coefficient follows a pressure-dependent model: $\SI{450}{g}$0 The paper distinguishes three regimes: rigid-rim contact at low pressure, mixed rigid-soft contact at intermediate pressure, and fully soft contact at high pressure (Ly et al., 7 Mar 2026). This formulation makes explicit a central soft-jig principle: holding performance can be improved by increasing interface friction and contact area, rather than by increasing normal force alone.
Variable-stiffness chains and fabrics extend the same principle to slender or palm-like fixtures. Compliant Joint Jamming provides a stiffness range from $\SI{450}{g}$1 to $\SI{450}{g}$2, a 4x increase, while retaining passive residual compliance (Westermann et al., 6 Feb 2025). The palm-shape gripper based on 3D-printed fabric jamming uses interlocked octahedral lattices inside vacuumized membranes to form broad adaptive contact surfaces that are soft during enclosure and stiff during retention (Zhao et al., 2023). A related continuum template is the monolithic bead-jamming pneumatic arm, which bends by approximately $\SI{450}{g}$3 at $\SI{450}{g}$4 when unjammed, then loses roughly $\SI{450}{g}$5 of range of motion when fully jammed, with an empirically identified load threshold of about $\SI{450}{g}$6 for when jamming becomes beneficial (Yao et al., 5 Feb 2025). These architectures generalize soft jigs from benchtop fixtures to compliant, reconfigurable support structures.
3. Sensing, estimation, and programmable jig behavior
An important development is the merger of fixturing and sensing. The sensing soft jig (Sakuma et al., 2021) uses a silicone membrane with internal white convex markers on a dark background, transparent beads, refractive-index-tuned oil, and two internal cameras about $\SI{450}{g}$7 mm below the base plate with a baseline of $\SI{450}{g}$8 mm. Marker centers are detected with Laplacian of Gaussian, refined to sub-pixel accuracy, triangulated using stereo geometry, and converted to a 3D marker cloud. A plane is then fit by SVD,
$\SI{450}{g}$9
and the normal vector $\SI{1}{mm}$0 is used as the principal normal vector of the object’s bottom surface (Sakuma et al., 2021). After calibration, the system estimates the orientation of cylindrical objects with diameter larger than $\SI{1}{mm}$1 mm with an RMSE of less than $\SI{1}{mm}$2 degrees. The soft jig therefore compensates for the pose uncertainty that its own compliance introduces.
The same paper embeds the jig into a robot calibration chain. Hand–eye calibration, PnP using four circular markers on the jig, and robot-controlled tilt trajectories are used to align jig coordinates with robot coordinates (Sakuma et al., 2021). The resulting fixture is not merely universal in shape; it is instrumented so that the robot can infer a principal normal vector directly from the deformed support surface. This is a distinct departure from rigid fixtures, where shape is fixed and pose is assumed, rather than estimated from fixture deformation.
A different form of softness appears in chemical experiment automation (Sasaki et al., 13 Jun 2025). Here the jigs are physically rigid or electromechanical, but modular and behaviorally reconfigurable. The system records a joint trajectory of end-effector poses and jig states,
$\SI{1}{mm}$3
and replays it by proportional pose control and state-difference commands,
$\SI{1}{mm}$4
The experimental jigs include a pipetting jig, bottle-mounting jig, flow-plumbing jig, and bottle-holder jig, all controlled as ROS nodes and finite-state machines (Sasaki et al., 13 Jun 2025). In two simple tasks, bottle manipulation and pipetting each achieved 100% success in 10 executions, with Hausdorff-distance-based motion reproducibility on the order of $\SI{1}{mm}$5 mm and $\SI{1}{mm}$6 mm in position, respectively, and three full polymer-synthesis trials were all successful (Sasaki et al., 13 Jun 2025). Although these jigs are not soft in the material sense, the paper makes explicit a broader architectural notion of soft tooling: behavior is taught and re-timed by demonstration rather than hard-coded around a single protocol.
4. Assembly, disassembly, and regrasping
Soft jigs first appeared in robotic assembly as general-purpose fixtures for parts of varying shape and posture. In soft-jig-driven assembly (Kiyokawa et al., 2020), a motor and a plate were fixed in multiple postures on a jamming-based silicone fixture, and all tested motor and plate postures achieved a 100% success rate over 10 trials. Fixing performance under external pushing was evaluated using the jig displacement metric
$\SI{1}{mm}$7
defined as $\SI{1}{mm}$8 of the robot’s $\SI{1}{mm}$9 push distance. Success cases averaged about $\SI{296}{cm^3}$0 displacement, whereas failure cases averaged about $\SI{296}{cm^3}$1 (Kiyokawa et al., 2020). The same system completed a belt-drive-unit-style assembly experiment in which the soft jig held a motor in a posture that rigid jigs could not easily fixture because of pins on the bottom surface (Kiyokawa et al., 2020).
Regrasping extends the soft-jig concept from support to pose reduction. The jamming-inspired regrasping tool (Kiyokawa et al., 17 Sep 2025) stamps a triangular-pyramid cavity into an unjammed membrane, then chooses stamping depth by optimizing a weighted combination of grasp feasibility and stability margin: $\SI{296}{cm^3}$2 The generated cavity matches the ideal shape with RMSE $\SI{296}{cm^3}$3 mm, and 20 drop trials per object on ten mechanical parts produced placement success rates exceeding 80% for most objects and above 90% for cylindrical ones (Kiyokawa et al., 17 Sep 2025). A trial was counted as successful when position error was within $\SI{296}{cm^3}$4 mm and orientation error within $\SI{296}{cm^3}$5 (Kiyokawa et al., 17 Sep 2025). This work formalizes a central soft-jig idea: once placement reliably yields a stable pose, subsequent regrasping can proceed with reduced uncertainty.
In robotic disassembly, the shell-type soft jig (Kiyokawa et al., 17 Sep 2025) combines a bottom jamming seat with four balloon modules around the object. For six of ten tested assemblies, the proposed fixture succeeded at angular pull-out deviations from $\SI{296}{cm^3}$6 through $\SI{296}{cm^3}$7, whereas the vise typically succeeded only at $\SI{296}{cm^3}$8 and sometimes $\SI{296}{cm^3}$9, and the jamming-gripper-inspired soft jig was often limited to around $\SI{90}{kPa}$0 (Kiyokawa et al., 17 Sep 2025). In force measurements for shaft–bearing, motor–pulley, and USB–adapter pull-out, the proposed jig showed approximately $\SI{90}{kPa}$1 pull-out forces without the sharp spikes seen in the vise case (Kiyokawa et al., 17 Sep 2025). The compliant cage therefore increases tolerance to recognition, planning, and control errors by allowing micro-motion and self-alignment during extraction.
A parallel line of work uses soft-jig principles directly in the end-effector. The finray-effect gripper for connector insertion behaves as a passive RCC-like fixture: it provides form-closure at the fingertip, high stiffness along the insertion axis, and low lateral stiffness for mechanical search (Hartisch et al., 2023). Its best experimental tolerance window was up to $\SI{90}{kPa}$2 mm in $\SI{90}{kPa}$3, with tool speeds of $\SI{90}{kPa}$4–$\SI{90}{kPa}$5, 84 trials with 2 failures in repeatability testing, and an assembly time from first contact of about $\SI{90}{kPa}$6 s (Hartisch et al., 2023). The compliant-joint-jamming anthropomorphic finger likewise acts as a soft alignment fixture during peg insertion; in a peg-in-hole task with 1 mm total clearance, the compliant gripper succeeded in 3 of 5 trials while the rigid one succeeded in 0 of 5, corresponding to a 60% higher success rate for the compliant design (Westermann et al., 6 Feb 2025). In both cases, stiffness anisotropy or residual compliance substitutes for force sensing and high-precision feedback.
5. Soft-robot interfaces and adaptive contact surfaces
Soft jigs also appear as interfaces mounted on highly deformable robots. The deformable tip mount for vine robots is explicitly describable as a soft jig: a textile “beanie” that encapsulates the leading section of the everting tube, grips by friction and elastic tension, and carries a payload such as a camera while allowing continued growth (Suulker et al., 2024). Experiments showed that a $\SI{90}{kPa}$7 cm diameter robot fitted with the cap and camera could pass through a $\SI{90}{kPa}$8 cm gate, and that protrusions measuring approximately $\SI{90}{kPa}$9 cm wide $1$0 cm long and $1$1 cm wide $1$2 cm long could evert through the cap without jamming or tearing (Suulker et al., 2024). The cap also remained attached during tip reorientation driven by navigation pouches. In fixturing terms, the cap provides a repeatable payload pose without imposing a rigid envelope.
Broad adaptive contact surfaces offer another interpretation. The palm-shape variable-stiffness gripper (Zhao et al., 2023) encloses objects with wide pneumatic palms, then increases stiffness via vacuum jamming of a 3D-printed fabric. For the main fabric design, the maximum gripping force reached $1$3 and the grip-to-pinch force ratio reached $1$4; for the layer-jamming variant, the maximum gripping force reached $1$5 and the grip-to-pinch force ratio reached $1$6 (Zhao et al., 2023). Pinch forces remained low, around $1$7 for the PET-lined fabric design, $1$8 for the fabric-only design, and $1$9 for the layer-jamming design at 0 actuator pressure (Zhao et al., 2023). The device handled 12 daily objects, including a ping pong ball, shuttlecock, bottle, pitaya, and wet hydrogel sphere, illustrating how a soft jig can be gentle during engagement yet strong during retention.
Pressure-tunable friction extends this idea to hybrid contacts. The hybrid finger with inflatable silicone pockets shows that increasing internal pressure increases the effective coefficient of friction, allowing secure lifting of heavy or slippery objects, while fragile items such as eggs, tofu, fruits, and paper cups can be held by increasing friction rather than normal force (Ly et al., 7 Mar 2026). In paper-cup tests, a 1 g payload achieved an optimal region around 2, 3, with roundness ratio 4, whereas a 5 g payload favored 6, 7, with roundness ratio 8 (Ly et al., 7 Mar 2026). At 9, the reported success rate reached 100% across all tested objects (Ly et al., 7 Mar 2026). For soft-jig design, the implication is that fixture reliability can be shifted from force-dominant clamping toward pressure-controlled frictional contact.
Harsh-environment handling motivates still another variant. The fire-resistant torus gripper based on wire jamming replaces membranes and granular media with titanium-alloy beads and tungsten wire, enabling compliant enclosure of sharp or burning debris (Tadakuma et al., 2019). In comparison with a conventional film jamming gripper, the wire-jamming torus achieved about 1.4× the gripping force, while pressing force at larger displacements remained lower (Tadakuma et al., 2019). The system was demonstrated gripping roughly 2 kg of burning concrete rubble. This broadens the soft-jig idea from gentle handling to damage-tolerant fixturing in environments where silicone membranes would fail.
6. Limits, trade-offs, and research directions
The principal limitation of soft jigs remains the trade-off between adaptability and absolute precision. The early assembly paper states directly that soft jigs provide lower intrinsic parts-positioning precision than rigid jigs (Kiyokawa et al., 2020). The sensing soft jig partially compensates for this by estimating a principal normal vector from membrane deformation, but its accuracy degrades for small contact areas, such as 0 mm cylinders, and for large tilt angles where part of the bottom surface lifts away from the membrane (Sakuma et al., 2021). Similarly, the regrasping jig reports failures caused by membrane elasticity, high friction, and geometric constraints, especially for complex parts and lightweight cylindrical objects that bounce or stall on cavity slopes (Kiyokawa et al., 17 Sep 2025).
Holding capacity and task geometry also impose hard limits. In the shell-type soft jig, AC–switch and pulley–shaft could not be successfully disassembled; the former required holding forces beyond the jig’s capacity, and the latter combined obstructing geometry with a high center of gravity (Kiyokawa et al., 17 Sep 2025). In connector insertion, part-specific fingertip notches and stiffness tuning are essential; overly soft fingers slip, whereas overly stiff ones reduce tolerance windows and risk damage (Hartisch et al., 2023). In the hybrid friction finger, the pressure that increases friction can also increase local indentation on soft or thin-walled objects, so the optimal operating point depends on both payload and material stiffness (Ly et al., 7 Mar 2026).
Material durability and reconfiguration overhead remain active concerns. The finray gripper reports rib buckling and cracking at high infill directions and densities (Hartisch et al., 2023). The bead-jamming arm is limited to 1 tendon tension, and full jamming reduces its range of motion from roughly 2–3 to about 4–5 (Yao et al., 5 Feb 2025). The chemical automation framework achieves high reproducibility, but it does not adapt to environmental variation, and protocol changes currently require a full re-demonstration rather than compositional reuse (Sasaki et al., 13 Jun 2025). These results indicate that soft jigs often trade dedicated mechanical tooling for dependence on material life, calibration, or demonstration workflows.
Current research directions therefore converge on hybridization. The sensing soft jig points toward better optical materials, denser markers, and eventual 6-DOF pose estimation (Sakuma et al., 2021). The regrasping tool points toward sharper cavity edges, improved membrane durability, lubrication, and vibration-assisted settling (Kiyokawa et al., 17 Sep 2025). The disassembly shell suggests more sophisticated balloon geometries, adjustable module counts, and richer pressure control (Kiyokawa et al., 17 Sep 2025). Across the broader literature, a plausible implication is that future soft jigs will increasingly combine compliant or jamming-based mechanics with embedded sensing, tunable friction, modular actuation, and task-level programmability, preserving the defining advantage of soft fixturing—adaptation to uncertain geometry—while reducing the penalties in precision and load capacity (Sakuma et al., 2021, Kiyokawa et al., 17 Sep 2025, Ly et al., 7 Mar 2026).
