ICRA 2026 Topic Tactile Force - Heungwoo/research GitHub Wiki
ICRA 2026 — Tactile & Force-Based Manipulation (Topic Analysis)
Venue: IEEE ICRA 2026 · Vienna, Austria · June 1–5, 2026 · compiled against the official PaperCept program (May 2026 snapshot). Scope: the 92 papers in this wiki's Tactile & Force cluster — vision-based tactile hardware, tactile/force representation learning, force control & estimation, in-hand/dexterous tactile, sim-to-real for tactile, visuo-tactile fusion, tactile for VLA/policies, and soft/compliant sensing.
Force-and-tactile is one of the largest single clusters at ICRA 2026 (92 papers here; "Force and Tactile Sensing" is one of the top program keywords, behind only the RL/IL/grasping learning themes). The venue's systems orientation shows: where ML venues debate VLA architecture, ICRA is where contact meets the robot. Three things are new in 2026. First, vision-based tactile sensors (VBTS) have splintered into a hardware design space rather than a single GelSight clone — translucent/marker-hybrid skins, UV-encoded elastomers, Moiré amplification, barometric/magnetic/acoustic/EIT transduction, and curved biomimetic geometries all appear. Second, sim-to-real for optical tactile has gone differentiable and physically calibrated (DOT-Sim, path-tracing Sim2Real, ETac, TacFlex). Third, force/tactile is being absorbed into learned policies and VLAs, including the headline idea of getting force awareness without a force sensor via distillation (FD-VLA). The list below covers all 92; the sub-trends are analytical groupings (papers carry several keywords and can belong to more than one).
Sub-trends
1. Vision-based tactile sensors & GelSight-style hardware
The single largest hardware thread is novel VBTS designs that push past the opaque-elastomer-plus-markers template. Balancing Marker and Markerless Modes ... with a Translucent Skin [ThI1I.164] directly tackles the marker/markerless trade-off (markers help force/shear, but occlude geometry) with a translucent skin. MoiréTac [TuI1I.54] generates dense interference patterns via overlapping micro-gratings to recover an analytical force-to-image relationship that sparse marker arrays lack. TransTac [WeI1I.326] and UVDtact [WeI1I.35] both exploit UV-encoded / UV-marker-embedded transparent elastomers — TransTac to transition between visuo-tactile and RGB-D depth modes, UVDtact for fingertip shape reconstruction and force estimation. MINT [ThI1I.223] is a hybrid VBTS+resistive-strain soft sensor for joint normal-force and texture perception; SuckTac [ThI1I.287] embeds a camera in a sucker for surface perception during suction. TacTape [ThAT3.7] is a tactile fiducial system with structured 3D texture to overcome VBTS's small local sensing area for localization. Calibration and curvature are first-class concerns: NLiPsCalib [TuI1I.160] is a calibration framework for high-fidelity 3D reconstruction of curved visuotactile sensors under non-uniform illumination.
2. Non-optical & large-area / novel transduction
A parallel hardware thread explores transduction beyond cameras, often chasing large-area coverage, low cost, or compliance preservation. EIT–Pneumatic Hybrid Robotic Skin [WeI1I.105] and Touch with Insight [WeI1I.184] both use electrical impedance tomography (the latter adds physics-aware data-driven learning); Magnet-Based Soft Robotic Skin [TuI2I.163] uses Hall-effect arrays plus a CNN tactile super-resolution model; Time-Division Multimodal Tactile Perception [ThI2LB.7] interleaves thermal and mechanical sensing in one layer. Acoustic morphological sensing appears twice: Acoustic Sensing for Universal Jamming Grippers [ThI1I.68] (sensing without compromising compliance) and Wearable, Fabric-Embedded Acoustic Waveguides [ThI1I.311] for meter-scale contact localization. Barometric (Constructing Contact Estimation Models for Barometric Tactile Sensors [TuI2I.47]), liquid-metal (3D Printable Soft Liquid Metal Sensors [ThBT3.5]), and low-cost conductive-fiber strain/touch sensors (Low Cost ... Strain and Touch Sensitive Fiber [WeI1I.226]) round out the cheap/manufacturable end. SpikeATac [TuI1I.262] is a standout multimodal finger combining a 16-taxel PVDF dynamic channel (4 kHz) with capacitive static sensing.
3. Tactile representation learning & perception
A cluster targets what to learn from tactile streams and how to make it transferable. SARL [TuI2I.216] is spatially-aware self-supervised representation learning for visuo-tactile perception; Built Different [ThI1I.236] uses tactile representations to bridge cross-embodiment capability differences in collaborative manipulation. How to Train Your Tactile Model [TuI2I.196] argues against CNN-per-sensor pipelines and for rapidly deployable models on multi-fingered hands; Tactile Recognition of Both Shapes and Materials [ThI1I.317] uses meta-learning with automatic feature optimization to fight tactile-data scarcity. Tacser [ThI2I.371] adds an action-conditioned latent filter for generalizable surface perception (moving beyond fixed-category texture classifiers), and TactEx [WeI1I.198] unifies vision/touch/language for explainable hardness estimation. Recurrent and predictive models appear in Tactile Object Recognition with RNNs [TuI2I.437] and Tactile Execution Monitoring [WeI1I.317] (time-series predictive encoding for skill-level monitoring). Tactile Elastography [WeI1I.4] recovers underlying elasticity distributions, and MFCC-Inspired Spectral Feature Extraction [TuI1I.181] handles social-robot touch interaction.
4. Force / torque control, estimation & sensorless force
Classical force control and force estimation remain strong, increasingly fused with learning. Learning-Guided Force-Feedback MPC [TuI1I.203] adds real-time force feedback and obstacle avoidance to MPC for deburring; Multifingered Force-Aware Control for Humanoid Robots [TuI1I.221] redistributes forces across torso/arm/wrist/fingers from tactile estimates; Reactive Slip Control in Multifingered Grasping [WeI2I.152] combines learned slip detection with model-based internal-force optimization. A notable estimation sub-thread is force without a dedicated F/T sensor: Estimating Force Interactions of Deformable Linear Objects from Their Shapes [TuI1I.85], Grasp-Independent Indirect Tool Force Estimation Using VBTS [TuI2I.378], CableSense [TuI2I.327] (MuJoCo-guided NN force estimation in cable-driven manipulators), Force Estimation ... Hydraulic Folded Pouch Actuator [WeI2I.187], Local Linearized Cosserat Rod Model [WeI1LB.14] for continuum-robot contact force, and the surgical Tri-Axial FBG-Based Force Sensor [TuI2I.122]. Sensor hardware itself includes the High-Stiffness Capacitive Torque Sensor [ThI2LB.16]. TacTip-Based Dynamic Contact Force Estimation [ThI1I.256] regresses force from sequential tactile images for force tracking. At the policy frontier, FD-VLA distills a force token from vision+state into a VLM, delivering force-aware contact-rich manipulation with no force sensor at deploy time.
5. Tactile in dexterous / in-hand manipulation
In-hand and multifingered work leans heavily on tactile feedback for state that vision can't see. Tactile-Driven Dexterous In-Hand Writing [ThI1I.373] uses extrinsic-contact sensing on a three-finger hand; Simultaneous Extrinsic Contact and In-Hand Pose Estimation [TuI2I.425] estimates both from distributed tactile sensing for peg insertion / tool use; UNIC [TuI2I.146] learns unified multimodal extrinsic contact estimation. TaSA [WeI1I.144] models tactile sensory attenuation to improve in-grasp manipulation; Learning Controlled Separation of Small Objects between Two Fingers [WeI2I.125] solves a novel count-down separation task with a tactile skin. High-Bandwidth Tactile-Reactive Control for Grasp Adjustment [WeI2I.148] and Shear-Based Grasp Control for Underactuated Tactile Hands [ThI2I.416] (Pisa/IIT SoftHand with microTac fingertips) close the loop fast. ActiveSPN [TuI2I.381] uses active soft polyhedral networks for in-finger pose estimation, and Grasp Like Humans [WeI1I.53] learns multifingered grasping from human proprioceptive sensorimotor integration. Modular Actuator for ... Proprioceptive and Kinesthetic Feedback [ThI2I.345] and the Low-Dimensional Tactile Glove [TuI1LB.19] address the hardware/teleop interface for dexterous tactile control.
6. Sim-to-real & differentiable tactile simulation
2026's clearest methodological shift is differentiable, physically-calibrated optical-tactile simulation. DOT-Sim [ThI1I.176] models the elastomer with the Material Point Method and learns a residual-image optical rendering, calibrating from a handful of demos in minutes. Automatic Physically-Based Sim2Real for Tactile Images through Differentiable Path-Tracing Rendering [ThI2I.308] attacks the optical refraction gap directly; ETac [TuBT1.5] is a lightweight tactile simulation framework for learning dexterous manipulation (fidelity/cost trade-off); TacFlex [WeI1I.420] simulates multi-mode tactile imprints across coating patterns and sensor configs. ConTact [ThI2I.314] does contrastive tactile alignment for sim-to-real RL; Zero-Shot Sim2Real Transfer for Magnet-Based Tactile Sensor [WeI2I.201] closes the gap for u-skin-style magnetic sensors on insertion; Learning Dexterous Manipulation Skills from Imperfect Simulations [ThI2I.276] (DexScrew) handles contact-dynamics mismatch.
7. Visuo-tactile fusion & generation
Naive concatenation of vision and touch often fails, so 2026 papers focus on how to fuse. Multi-Modal Manipulation via Multi-Modal Policy Consensus [TuI1I.51] factorizes the policy into per-modality diffusion experts with a learned router so sparse tactile signals aren't drowned out by vision. Symmetry-Aware Fusion of Vision and Tactile ... via Bilateral Force Priors [ThI2I.81] addresses why naive visuo-tactile fusion underperforms on insertion. Generation/cross-modal hallucination is a growing trick: MultiDiffSense [ThI2I.133] is a unified diffusion model generating multimodal visuo-tactile images conditioned on object shape and contact pose; ViTacGen [TuI1I.348] generates touch from vision for robotic pushing when real tactile hardware is unavailable. TacUMI [WeI2I.188] is a multimodal universal manipulation interface; FreeTacMan [TuI2I.177] is a robot-free, wearable visuo-tactile data-collection rig. TactEx [WeI1I.198] and InvariantCloud [ThI1I.348] (globally-invariant indexed point cloud for 6-DoF tactile pose tracking) sit at the fusion/perception boundary.
8. Tactile/force for learned policies & VLAs
Beyond fusion, tactile is increasingly inside the policy. Tactile-Conditioned Diffusion Policy for Force-Aware Manipulation [TuI1I.229] makes applied force a controlled output rather than a passive observation; PoCoDP3 [ThI2I.397] is a pose- and contact-aware visual-tactile 3D diffusion policy; ManipForce [ThI2I.257] introduces a Frequency-Aware Multimodal Transformer fusing asynchronous RGB and high-frequency F/T inside a diffusion policy. Multimodal Diffusion Forcing for Forceful Manipulation [WeI1I.209] exploits the interplay between modalities rather than mapping images→actions. ViTac-Tracing [ThI1I.158] is visual-tactile imitation learning for deformable-object tracing, and Touch2Insert [TuI1I.219] does zero-shot peg insertion by touching peg/hole intersections. TranTac [WeI1I.169] embeds a contact-sensitive IMU in gripper tips for transient-signal insertion. FD-VLA is the VLA-side flagship (force-token distillation, no sensor at inference).
9. Soft / compliant sensing & applications
A long tail applies tactile/force to compliant and applied settings. Soft grippers and compliant wrists: Contact Detection and Manipulation with a Shape-Memory Alloy Soft Gripper [TuAT4.1], ShapeForce [TuI2I.224] (low-cost soft robotic wrist), FORTE [TuI2I.420] (fin-ray fingers with internal air channels for force+slip), INTACT-GRIP [WeI1I.255] (inflatable tactile gripper). Agriculture and fragile handling: DexFruit [WeI1I.437] (gentle fruit handling + Gaussian-splatting damage inspection), RICE [ThI1I.163] (reactive interaction in cluttered canopies), Intelligent Mechanical Characterization of Date Fruits [ThI1LB.17]. Medical/wearable: MINT [ThI1I.223], Multi-Modal Sensing in Colonoscopy [TuI2I.392], Haptics of Pulse Palpation [TuI1I.434], CPR Training Glove [WeI2I.107]. Robust skin/whole-arm and other embodiments: No Need to Look! [WeI2I.44] (grasping with a sensitive-skin-covered arm), Tactile Hide and Seek [ThI2I.192] (bimanual tactile-only search), SlipSense [ThI1I.150] (slip detection on legged robots), Omnidirectional Dual-Arm Aerial Manipulator [WeI2I.142] (proprioceptive contact localization for roof landing), Ultra-Fast Incipient Slip Detection with Hyperdimensional Computing [WeI2I.408] (PapillArray). Active/exploratory sensing: Active Tactile Exploration for ... Pose and Shape [WeI1I.113], Grasp, Slide, Roll [ThI1I.41] (contact-mode comparison for shape reconstruction), Autonomous Exploration for Shape Reconstruction [WeI2I.401]. Skill transfer and learning: Few-Shot Transfer of Tool-Use Skills [WeI2I.47], Bootstrapping Self-Supervised ... Insertion [WeI1I.197], Vi-TacMan [WeI2I.211] (articulated-object manipulation via vision+touch).
Standout deep-dives
-
FreeTacMan [TuI2I.177] (arXiv 2506.01941, OpenDriveLab). A robot-free, wearable dual visuo-tactile gripper worn on human fingers, with optical pose tracking, used to collect 10k trajectories across 50 contact-rich tasks. The headline: imitation policies trained on its visuo-tactile data reach an average success rate 50% higher than vision-only counterparts, making the case that the bottleneck for contact-rich learning is data-collection ergonomics, not algorithms.
-
SpikeATac [TuI1I.262] (arXiv 2510.27048). A multimodal tactile finger that pairs a 16-taxel PVDF dynamic channel sampled at 4 kHz with capacitive static sensing, capturing the very onset and breaking of contact. Combined with an RLHF-with-tactile-reward learning recipe, it enables in-hand manipulation of fragile/deformable objects — stopping quickly and delicately on contact — which the authors frame as a previously unachieved dexterous task.
-
FORTE [TuI2I.420] (arXiv 2506.18960). Tactile force and slip sensing built into 3D-printed fin-ray fingers via internal air channels rather than mounted sensors. It estimates grasping forces over 0–8 N with ~0.2 N average error and detects slip within 100 ms, grasping raspberries and potato chips at 92% success with 93% slip-detection accuracy — a compelling low-latency, manufacturable answer for delicate manipulation.
-
DOT-Sim [ThI1I.176] (arXiv 2604.27367). A differentiable optical tactile simulator that models the soft elastomer with the Material Point Method and learns a residual-image optical rendering relative to the real idle state. It calibrates from a small number of real demonstrations in minutes and supports much larger, non-linear deformations than prior simulators — exemplifying 2026's move to physically-calibrated, differentiable tactile sim.
-
Multi-Modal Manipulation via Multi-Modal Policy Consensus [TuI1I.51] (arXiv 2509.23468). Factorizes the policy into per-modality diffusion experts (vision, touch, …) and learns a router that computes consensus weights, so sparse-but-critical tactile signals retain influence and new modalities can be added without retraining the whole policy. Evaluated on RLBench plus real occluded-picking, in-hand spoon reorientation, and puzzle insertion, beating feature-concatenation baselines — a clean structural answer to "vision drowns out touch."
-
ManipForce [ThI2I.257] (arXiv 2509.19047). A handheld rig capturing high-frequency F/T + RGB from human demos, with a Frequency-Aware Multimodal Transformer that embeds asynchronous RGB and F/T by frequency/modality and fuses them via cross-attention inside a diffusion policy. It reaches 83% average success across six contact-rich tasks (gear assembly, box flipping, battery insertion), substantially over RGB-only.
-
TranTac [WeI1I.169] (arXiv 2509.16550). A data-efficient, low-cost approach embedding a single contact-sensitive 6-axis IMU in the gripper's elastomer tips. With vision, it hits 79% average success on grasping+insertion and ~70% on unseen objects (USB plug, metal key), reportedly outperforming both vision-only and 6D F/T-sensor-augmented policies — a reminder that transient contact dynamics, not steady-state force, often carry the decisive signal.
-
FD-VLA (arXiv 2602.02142). The VLA flagship of this cluster: a Force Distillation Module maps a learnable query (conditioned on vision+state) to a force token aligned in training with the latent of the real force signal, then injected into a pretrained VLM. Force awareness is delivered as a learned token with no force/torque sensor at inference, and the authors report the distilled token outperforming real sensor readings — reframing force sensing for VLAs as a representation problem rather than a hardware requirement.
Complete paper list (92)
| Code | Title | arXiv |
|---|---|---|
| ThAT3.7 | TacTape: Real-Time High-Accuracy Tactile Fiducial System with Structured 3D Texture for Vision-Based Tactile Sensors | — |
| ThBT3.5 | 3D Printable Soft Liquid Metal Sensors for Delicate Manipulation Tasks | 2509.17389 |
| ThI1I.150 | SlipSense: Multimodal Sensing for Online Slip Detection in Legged Robots | — |
| ThI1I.158 | ViTac-Tracing: Visual-Tactile Imitation Learning of Deformable Object Tracing | 2603.18784 |
| ThI1I.163 | RICE: Reactive Interaction Controller for Cluttered Canopy Environment | 2506.10383 |
| ThI1I.164 | Balancing Marker and Markerless Modes in Vision-Based Tactile Sensors with a Translucent Skin | 2512.06829 |
| ThI1I.176 | DOT-Sim: Differentiable Optical Tactile Simulation with Precise Real-To-Sim Physical Calibration | 2604.27367 |
| ThI1I.223 | MINT: A Vision-Based Soft Sensor for Mutual Integration of Normal Interaction Force and Texture Perception | — |
| ThI1I.236 | Built Different: Tactile Perception to Overcome Cross-Embodiment Capability Differences in Collaborative Manipulation | 2409.14896 |
| ThI1I.256 | TacTip-Based Dynamic Contact Force Estimation with Sequential Tactile Images and Its Applications to Robotic Force Tracking | — |
| ThI1I.287 | SuckTac: Camera-Based Tactile Sucker for Unstructured Surface Perception and Interaction | 2511.02294 |
| ThI1I.311 | Wearable, Fabric-Embedded Acoustic Waveguides for Meter-Scale Contact Localization and Force Sensing | — |
| ThI1I.317 | Tactile Recognition of Both Shapes and Materials with Automatic Feature Optimization-Enabled Meta Learning | 2603.08423 |
| ThI1I.348 | InvariantCloud: A Globally Invariant, Uniquely Indexed Point Cloud Framework for Robust 6-DoF Tactile Pose Tracking | — |
| ThI1I.373 | Tactile-Driven Dexterous In-Hand Writing Via Extrinsic Contact Sensing | — |
| ThI1I.41 | Grasp, Slide, Roll: Comparative Analysis of Contact Modes for Tactile-Based Shape Reconstruction | 2602.23206 |
| ThI1I.68 | Acoustic Sensing for Universal Jamming Grippers | 2603.00351 |
| ThI1LB.17 | Intelligent Mechanical Characterization of Date Fruits for Automated Harvesting Grippers | — |
| ThI2I.133 | MultiDiffSense: Diffusion-Based Multi-Modal Visuo-Tactile Image Generation Conditioned on Object Shape and Contact Pose | 2602.19348 |
| ThI2I.192 | Tactile Hide and Seek: Bimanual Object Blind Search and Retrieval Via Tactile-Only Feedback | — |
| ThI2I.196 | How to Train Your Tactile Model: Tactile Perception with Multi-Fingered Robot Hands | 2604.00744 |
| ThI2I.257 | ManipForce: Force-Guided Policy Learning with Frequency-Aware Representation for Contact-Rich Manipulation | 2509.19047 |
| ThI2I.276 | Learning Dexterous Manipulation Skills from Imperfect Simulations (DexScrew) | 2512.02011 |
| ThI2I.308 | Automatic Physically-Based Sim2Real for Tactile Images through Differentiable Path-Tracing Rendering | — |
| ThI2I.314 | ConTact: Contrastive Tactile Alignment for Sim-To-Real Robotic Manipulation | — |
| ThI2I.345 | Modular Actuator for Multimodal Proprioceptive and Kinesthetic Feedback of Robotic Hands | — |
| ThI2I.371 | Tacser and Action-Conditioned Latent Filter for Generalizable Robotic Surface Perception | — |
| ThI2I.397 | PoCoDP3: Pose and Contact-Aware Visual-Tactile Policy for Contact-Rich 3D Manipulation | — |
| ThI2I.416 | Shear-Based Grasp Control for Multi-Fingered Underactuated Tactile Robotic Hands | 2503.17501 |
| ThI2I.81 | Symmetry-Aware Fusion of Vision and Tactile Sensing Via Bilateral Force Priors for Robotic Manipulation | 2602.13689 |
| ThI2LB.16 | High-Stiffness Capacitive Torque Sensor Based on a Hybrid Scott-Russell and Parallelogram Mechanism | — |
| ThI2LB.7 | Time-Division Multimodal Tactile Perception for Physical AI and Robotic Hands | — |
| TuAT4.1 | Contact Detection and Manipulation with a Shape-Memory Alloy Based Soft Gripper | — |
| TuBT1.5 | ETac: A Lightweight and Efficient Tactile Simulation Framework for Learning Dexterous Manipulation | 2604.20295 |
| TuI1I.160 | NLiPsCalib: An Efficient Calibration Framework for High-Fidelity 3D Reconstruction of Curved Visuotactile Sensors | 2603.09319 |
| TuI1I.181 | MFCC Inspired Spectral Feature Extraction for Robust Touch Interaction in Social Robots | — |
| TuI1I.203 | Learning-Guided Force-Feedback Model Predictive Control with Obstacle Avoidance for Robotic Deburring | 2604.06133 |
| TuI1I.219 | Touch2Insert: Zero-Shot Peg Insertion by Touching Intersections of Peg and Hole | 2603.03627 |
| TuI1I.221 | Multifingered Force-Aware Control for Humanoid Robots | 2603.08142 |
| TuI1I.229 | Tactile-Conditioned Diffusion Policy for Force-Aware Robotic Manipulation | 2510.13324 |
| TuI1I.262 | SpikeATac: A Multimodal Tactile Finger with Taxelized Dynamic Sensing for Dexterous Manipulation | 2510.27048 |
| TuI1I.348 | ViTacGen: Robotic Pushing with Vision-To-Touch Generation | 2510.14117 |
| TuI1I.434 | Haptics of Pulse Palpation: Simulation and Validation through Novel Sensor-Actuator System (I) | — |
| TuI1I.51 | Multi-Modal Manipulation Via Multi-Modal Policy Consensus | 2509.23468 |
| TuI1I.54 | MoiréTac: A Dual-Mode Visuotactile Sensor for Multidimensional Perception Using Moiré Pattern Amplification | 2509.12714 |
| TuI1I.85 | Estimating Force Interactions of Deformable Linear Objects from Their Shapes | 2602.01085 |
| TuI1LB.19 | Low-Dimensional Tactile Glove for Visuo-Tactile Robot Hand Control: A Preliminary Study | — |
| TuI2I.122 | A Tri-Axial FBG-Based Force Sensor at the Tool Tip of a Continuum Manipulator for Single-Port Access Surgery | — |
| TuI2I.146 | UNIC: Learning Unified Multimodal Extrinsic Contact Estimation | 2601.04356 |
| TuI2I.163 | Magnet-Based Soft Robotic Skin Using a 3D-Printed Multi-Lattice Structure and CNN-Based Tactile Super-Resolution | — |
| TuI2I.177 | FreeTacMan: Robot-Free Visuo-Tactile Data Collection System for Contact-Rich Manipulation | 2506.01941 |
| TuI2I.216 | SARL: Spatially-Aware Self-Supervised Representation Learning for Visuo-Tactile Perception | 2512.01908 |
| TuI2I.224 | ShapeForce: Low-Cost Soft Robotic Wrist for Contact-Rich Manipulation | 2511.19955 |
| TuI2I.327 | CableSense: MuJoCo Simulation-Guided Neural Networks for Force Estimation in Cable-Driven Manipulators | — |
| TuI2I.378 | Grasp Independent Indirect Tool Force Estimation Using Vision-Based Tactile Sensors | — |
| TuI2I.381 | ActiveSPN: Active Soft Polyhedral Networks with Pose Estimation for In-Finger Object Manipulation | — |
| TuI2I.392 | Multi-Modal Sensing in Colonoscopy: A Data-Driven Approach | — |
| TuI2I.420 | FORTE: Tactile Force and Slip Sensing on Compliant Fingers for Delicate Manipulation | 2506.18960 |
| TuI2I.425 | Simultaneous Extrinsic Contact and In-Hand Pose Estimation Via Distributed Tactile Sensing | 2512.23856 |
| TuI2I.437 | Tactile Object Recognition with Recurrent Neural Networks through a Perceptive Soft Gripper | — |
| TuI2I.47 | Constructing Contact Estimation Models for Barometric Tactile Sensors | — |
| WeI1I.105 | EIT–Pneumatic Hybrid Robotic Skin for Practical and Accurate Force Map Reconstruction | 2605.28468 |
| WeI1I.113 | Active Tactile Exploration for Rigid Body Pose and Shape Estimation | 2510.13595 |
| WeI1I.144 | TaSA: Two-Phased Deep Predictive Learning of Tactile Sensory Attenuation for Improving In-Grasp Manipulation | 2602.05468 |
| WeI1I.169 | TranTac: Leveraging Transient Tactile Signals for Contact-Rich Robotic Manipulation | 2509.16550 |
| WeI1I.184 | Touch with Insight: Physics-Aware Data-Driven Learning for EIT-Based Tactile Sensing | — |
| WeI1I.197 | Bootstrapping Self-Supervised Learning of Binary Classification Using Error Bounds: A Case Study on a Robotic Insertion Task | — |
| WeI1I.198 | TactEx: An Explainable Multimodal Robotic Interaction Framework for Human-Like Touch and Hardness Estimation | 2602.18967 |
| WeI1I.209 | Multimodal Diffusion Forcing for Forceful Manipulation | 2511.04812 |
| WeI1I.226 | Low Cost, Easily Manufactured, Highly Flexible Strain and Touch Sensitive Fiber for Robotics Applications | — |
| WeI1I.255 | INTACT-GRIP: An Inflatable Tactile Gripper for Soft Manipulation and High-Resolution Texture Mapping | — |
| WeI1I.317 | Tactile Execution Monitoring of Robotic Manipulation Via Time-Series Based Predictive Encoding | — |
| WeI1I.326 | TransTac: Visuo-Tactile Modality Transition Via Ultraviolet-Encoded Transparent Elastomers | — |
| WeI1I.35 | UVDtact: UV Marker-Embedded Fingertip-Like Vision-Based Tactile Sensor for Shape Reconstruction and Force Estimation | — |
| WeI1I.4 | Tactile Elastography | — |
| WeI1I.420 | TacFlex: Multi-Mode Tactile Imprints Simulation for Visuotactile Sensors with Coating Patterns | — |
| WeI1I.437 | DexFruit: Dexterous Manipulation and Gaussian Splatting Inspection of Fruit | 2508.07118 |
| WeI1I.53 | Grasp Like Humans: Learning Generalizable Multi-Fingered Grasping from Human Proprioceptive Sensorimotor Integration | 2509.08354 |
| WeI1LB.14 | Local Linearized Cosserat Rod Model for Contact Force Estimation in Flexible Medical Instruments and Continuum Robots | — |
| WeI2I.107 | A Closed-Loop CPR Training Glove with Integrated Tactile Sensing and Haptic Feedback | 2603.05793 |
| WeI2I.125 | Learning Controlled Separation of Small Objects between Two Fingers with a Tactile Skin | — |
| WeI2I.142 | Omnidirectional Dual-Arm Aerial Manipulator with Proprioceptive Contact Localization for Landing on Slanted Roofs | 2602.10703 |
| WeI2I.148 | High-Bandwidth Tactile-Reactive Control for Grasp Adjustment | 2509.15876 |
| WeI2I.152 | Reactive Slip Control in Multifingered Grasping: Hybrid Tactile Sensing and Internal-Force Optimization | 2602.16127 |
| WeI2I.187 | Force Estimation and Position Control of a Hydraulic Folded Pouch Actuator for Soft Robotics | — |
| WeI2I.188 | TacUMI: A Multi-Modal Universal Manipulation Interface for Contact-Rich Tasks | 2601.14550 |
| WeI2I.201 | Zero-Shot Sim2Real Transfer for Magnet-Based Tactile Sensor on Insertion Tasks | 2505.02915 |
| WeI2I.211 | Vi-TacMan: Articulated Object Manipulation Via Vision and Touch | 2510.06339 |
| WeI2I.401 | Autonomous Exploration for Shape Reconstruction and Measurement Via Informative Contact-Guided Planning | — |
| WeI2I.408 | Ultra-Fast Lightweight Incipient Slip Detection Using Hyperdimensional Computing with the PapillArray Tactile Sensor | — |
| WeI2I.44 | No Need to Look! Locating and Grasping Objects by a Robot Arm Covered with Sensitive Skin | 2508.17986 |
| WeI2I.47 | Few-Shot Transfer of Tool-Use Skills Using Human Demonstrations with Proximity and Tactile Sensing | 2507.13200 |
Related
- Tactile VLA cross-paper review (architecture × sensor HW × 2026 trends)
- ICRA 2026 Survey
- Dexterous Manipulation review
- FD-VLA (force awareness without a force sensor)
← Back to ICRA-2026-VLA-Manipulation-Survey