arXiv subjects

Hojung Choi

Publications and source records attributed to Hojung Choi.

9 recordsLinked to original sources

arXiv2026

Multisensory Continual Learning: Adapting Pretrained Visuomotor Policies to Force

Robot manipulation often relies on sensory feedback beyond vision, particularly in contact-rich settings where force, tactile, or audio signals reveal interaction states that are not directly observable from images. However, these modalities are often hardware- and task-specific, and large-scale multisensory robot datasets remain scarce. As a result, it is impractical to pretrain policies with every sensor they may encounter. We study multisensory continual learning: adapting a pretrained robot policy to new tasks with newly introduced modalities while preserving performance under the original sensor suite. We propose MultiSensory World Model (MuSe), which incorporates limited multisensory data into pretrained vision-only policies through multi-stage fusion, multisensory future prediction, and experience replay over pretraining data. We instantiate MuSe by augmenting a pretrained vision-only policy with force-torque sensing and evaluate it on real-world manipulation tasks. Our experiments show that MuSe performs strongly on contact-rich finetuning tasks while preserving, and in some cases improving, performance on the original pretraining tasks. These results suggest that a modest multisensory dataset can improve general robot capabilities beyond the finetuning distribution. Project website: https://jadenvc.github.io/multisensory-continual-learning/

cs.RO↗

arXiv2026

In-the-Wild Compliant Manipulation with UMI-FT

Many manipulation tasks require careful force modulation. With insufficient force the task may fail, while excessive force could cause damage. The high cost, bulky size and fragility of commercial force/torque (F/T) sensors have limited large-scale, force-aware policy learning. We introduce UMI-FT, a handheld data-collection platform that mounts compact, six-axis force/torque sensors on each finger, enabling finger-level wrench measurements alongside RGB, depth, and pose. Using the multimodal data collected from this device, we train an adaptive compliance policy that predicts position targets, grasp force, and stiffness for execution on standard compliance controllers. In evaluations on three contact-rich, force-sensitive tasks (whiteboard wiping, skewering zucchini, and lightbulb insertion), UMI-FT enables policies that reliably regulate external contact forces and internal grasp forces, outperforming baselines that lack compliance or force sensing. UMI-FT offers a scalable path to learning compliant manipulation from in-the-wild demonstrations. We open-source the hardware and software to facilitate broader adoption at:https://umi-ft.github.io/.

cs.RO↗

arXiv2025

Multimodal Sensing for Robot-Assisted Sub-Tissue Feature Detection in Physiotherapy Palpation

Robotic palpation relies on force sensing, but force signals in soft-tissue environments are variable and cannot reliably reveal subtle subsurface features. We present a compact multimodal sensor that integrates high-resolution vision-based tactile imaging with a 6-axis force-torque sensor. In experiments on silicone phantoms with diverse subsurface tendon geometries, force signals alone frequently produce ambiguous responses, while tactile images reveal clear structural differences in presence, diameter, depth, crossings, and multiplicity. Yet accurate force tracking remains essential for maintaining safe, consistent contact during physiotherapeutic interaction. Preliminary results show that combining tactile and force modalities enables robust subsurface feature detection and controlled robotic palpation.

cs.RO↗

arXiv2025

CoinFT: A Coin-Sized, Capacitive 6-Axis Force Torque Sensor for Robotic Applications

We introduce CoinFT, a capacitive 6-axis force/torque (F/T) sensor that is compact, light, low-cost, and robust with an average root-mean-squared error of 0.16N for force and 1.08mNm for moment when the input ranges from 0~14N and 0~5N in normal and shear directions, respectively. CoinFT is a stack of two rigid PCBs with comb-shaped electrodes connected by an array of silicone rubber pillars. The microcontroller interrogates the electrodes in different subsets in order to enhance sensitivity for measuring 6-axis F/T. The combination of features of CoinFT enables various contact-rich robot interactions across different embodiment domains including drones, robot end-effectors, and wearable haptic devices. We demonstrate the utility of CoinFT through two representative applications: a multi-axial contact-probing experiment in which a CoinFT mounted beneath a hemispherical fingertip measures 6-axes of force and torque representative of manipulation scenarios, and an attitude-based force-control task on a drone. The design, fabrication, and firmware of CoinFT are open-sourced at https://coin-ft.github.io/.

cs.RO↗

arXiv2025

DexForce: Extracting Force-informed Actions from Kinesthetic Demonstrations for Dexterous Manipulation

Imitation learning requires high-quality demonstrations consisting of sequences of state-action pairs. For contact-rich dexterous manipulation tasks that require dexterity, the actions in these state-action pairs must produce the right forces. Current widely-used methods for collecting dexterous manipulation demonstrations are difficult to use for demonstrating contact-rich tasks due to unintuitive human-to-robot motion retargeting and the lack of direct haptic feedback. Motivated by these concerns, we propose DexForce. DexForce leverages contact forces, measured during kinesthetic demonstrations, to compute force-informed actions for policy learning. We collect demonstrations for six tasks and show that policies trained on our force-informed actions achieve an average success rate of 76% across all tasks. In contrast, policies trained directly on actions that do not account for contact forces have near-zero success rates. We also conduct a study ablating the inclusion of force data in policy observations. We find that while using force data never hurts policy performance, it helps most for tasks that require advanced levels of precision and coordination, like opening an AirPods case and unscrewing a nut.

cs.RO↗

arXiv2024

Tactile-Informed Action Primitives Mitigate Jamming in Dense Clutter

It is difficult for robots to retrieve objects in densely cluttered lateral access scenes with movable objects as jamming against adjacent objects and walls can inhibit progress. We propose the use of two action primitives -- burrowing and excavating -- that can fluidize the scene to un-jam obstacles and enable continued progress. Even when these primitives are implemented in an open loop manner at clock-driven intervals, we observe a decrease in the final distance to the target location. Furthermore, we combine the primitives into a closed loop hybrid control strategy using tactile and proprioceptive information to leverage the advantages of both primitives without being overly disruptive. In doing so, we achieve a 10-fold increase in success rate above the baseline control strategy and significantly improve completion times as compared to the primitives alone or a naive combination of them.

cs.RO↗

arXiv2022

Deep Learning Classification of Touch Gestures Using Distributed Normal and Shear Force

When humans socially interact with another agent (e.g., human, pet, or robot) through touch, they do so by applying varying amounts of force with different directions, locations, contact areas, and durations. While previous work on touch gesture recognition has focused on the spatio-temporal distribution of normal forces, we hypothesize that the addition of shear forces will permit more reliable classification. We present a soft, flexible skin with an array of tri-axial tactile sensors for the arm of a person or robot. We use it to collect data on 13 touch gesture classes through user studies and train a Convolutional Neural Network (CNN) to learn spatio-temporal features from the recorded data. The network achieved a recognition accuracy of 74% with normal and shear data, compared to 66% using only normal force data. Adding distributed shear data improved classification accuracy for 11 out of 13 touch gesture classes.

cs.RO↗

arXiv2022

Scalably manufactured high-index atomic layer-polymer hybrid metasurfaces for high-efficiency virtual reality metaoptics in the visible

Metalenses, which exhibit superior light-modulating performance with sub-micrometer-scale thicknesses, are suitable alternatives to conventional bulky refractive lenses. However, fabrication limitations, such as a high cost, low throughput, and small patterning area, hinder their mass production. Here, we demonstrate the mass production of low-cost, high-throughput, and large-aperture visible metalenses using an argon fluoride immersion scanner and wafer-scale nanoimprint lithography. Once a 12-inch master stamp is imprinted, hundreds of centimeter-scale metalenses can be fabricated. To enhance light confinement, the printed metasurface is thinly coated with a high-index film, resulting in drastic increase of conversion efficiency. As a proof of concept, a prototype of a virtual reality device with ultralow thickness is demonstrated with the fabricated metalens.

physics.optics↗

arXiv2022

Perceived Intensities of Normal and Shear Skin Stimuli using a Wearable Haptic Bracelet

Our aim is to provide effective interaction with virtual objects, despite the lack of co-location of virtual and real-world contacts, while taking advantage of relatively large skin area and ease of mounting on the forearm. We performed two human participant studies to determine the effects of haptic feedback in the normal and shear directions during virtual manipulation using haptic devices worn near the wrist. In the first study, participants performed significantly better while discriminating stiffness values of virtual objects when the feedback consisted of normal displacements compared to shear displacements. Participants also commented that they could detect normal cues much easier than shear, which motivated us to perform a second study to find the point of subjective equality (PSE) between normal and shear stimuli. Our results show that shear stimuli require a larger actuator displacement but less force than normal stimuli to achieve perceptual equality for our haptic bracelets. We found that normal and shear stimuli cannot be equalized through skin displacement nor the interaction forces across all users. Rather, a calibration method is needed to find the point of equality for each user where normal and shear stimuli create the same intensity on the user's skin.

cs.RO↗