Exploring Proximity and Tactile Sensing for Robotic Manipulation

Eric T. Chang, Bianca Aumann, Zhanpeng He, Mohamed Elnoor, Thomas Liang, Matei Ciocarlie, Nima Fazeli, Joshua Migdal, Tye Brady, Taskin Padir

November 2026

TL;DR ToFTouch puts three 4x4 multizone time-of-flight depth sensors next to an optical pressure patch on one gripper finger, so the robot can sense objects before and during contact. Closing the gripper early on ToF readings cut grasp cycle time by 7.4%, and ToF depth correction picked items 9/10 times at 2 cm depth errors where picking at the nominal depth failed every time.

What they do

Tactile sensors only respond once contact has happened, and wrist cameras are often occluded right where the fingers meet the object. ToFTouch adds proximity sensing at the fingertip: a 45 x 30 x 10 mm module with three multizone time-of-flight (ToF) sensors next to an optical pressure patch. The authors mount it on a Robotiq gripper on a Franka FR3 and test it on three tasks: pre-grasp closing, picking from clutter and inserting into tight clutter.

How it works

Proximity. Three VL53L8CX sensors, each a 4x4 depth grid at 60 Hz, are folded on a flexible PCB to face 0, 45 and 90 degrees from vertical, 48 zones in total. Readings below 1-2 cm are unreliable.
Touch. A 3 x 3 cm patch of four LEDs and four photodiodes under a coated elastomer. Each LED-photodiode pair is sampled in turn, giving a 16-value signal at 300 Hz. Only the raw signals are used, with no force model.
Control. For picking, a PID loop on the downward ToF sensor sets the final gripper depth; no camera is used for control.

Results

Pre-grasp: closing the gripper once the ToF sees the table at 130 mm cut cycle time from 3.533 s to 3.271 s over ten trials (7.4%).
Extraction with the object 2 cm above or below the nominal pick depth: 9/10 picks with ToF, 0/10 without. At -1 cm, ToF did worse (8/10 against 10/10); the main failure was sensor overheating.
Measured depth corrections had standard deviations of at most 0.4 cm but did not match the true offsets (1.6 cm at zero offset), which the authors put down to no calibration and misaligned mounting.
Insertion (a kinesthetically taught trajectory): ToF readings pick out the tops of neighboring boxes and the free cavity, and the photodiode signals leave their baseline at about the same time as the load cell registers contact (x-axis force reaching about -20 N).

Limitations

The authors call it an initial characterization: few trials per condition, no statistical analysis and no comparison with other proximity or tactile sensors. The ToF sensors are uncalibrated and overheat. The insertion was taught kinesthetically, so the sensor only observed it rather than closing a control loop, and multi-fingered hands are left to future work.

Figures

From Figure 1: Exploded view of ToFTouch, with a U.S. penny for scale. The upper inset shows ToF points accumulated d... — From Figure 1: Exploded view of ToFTouch, with a U.S. penny for scale. The upper inset shows ToF points accumulated during a grasp, red from the downward sensor and blue from the sensor angled 45 degrees toward the object.

From Figure 2: Left: ToFTouch mounted on one finger of a Robotiq parallel jaw gripper. Right: the gripper above the c... — From Figure 2: Left: ToFTouch mounted on one finger of a Robotiq parallel jaw gripper. Right: the gripper above the clutter used for the extraction experiments.

From Figure 4: Readings from the three ToF sensors (red, green, blue) during insertion into tight clutter, at four mo... — From Figure 4: Readings from the three ToF sensors (red, green, blue) during insertion into tight clutter, at four moments in chronological order (A-D).

From Figure 5: Load-cell forces (top) and ToFTouch tactile signals, one per LED-photodiode pair (bottom), during the ... — From Figure 5: Load-cell forces (top) and ToFTouch tactile signals, one per LED-photodiode pair (bottom), during the same insertion. Letters match Figure 4.

Abstract

Pre-touch sensing, or sensing information about an object before contact, is an attractive modality for robot hands as it can allow a robot to anticipate and respond to a contact before it happens. In this work, we introduce ToFTouch, a multimodal tactile sensor combining three multizone (4$\times$4 resolution) time-of-flight sensors with an optics-based pressure sensing array. We mount this sensor on a parallel jaw gripper and demonstrate its utility on three distinct tasks: (i) pre-grasp adjustment, (ii) item extraction from clutter, and (iii) item insertion into tight spaces. For pre-grasp adjustment, ToF feedback reduced cycle time by 7.4$\%$ by enabling anticipatory gripper closure during approach. For extraction, ToF sensing enabled successful retrieval across all tested depth offsets ($\pm$2 cm) deviating from the nominal pick point, including conditions where baseline approaches failed, with standard deviations below 0.4 cm. For insertion into dense clutter, we explore the proximity and tactile information provided from ToFTouch during an insertion trajectory. The results demonstrate the utility of multizone ToF sensing in a manipulation environment.

Summaries are written automatically from each paper's text. Figures and videos come from the paper and its project page; all credit goes to the authors.

Matt Strong