Research Analyzer
← Back ICRA 2026

Ultra-Fast Lightweight Incipient Slip Detection Using Hyperdimensional Computing with the PapillArray Tactile Sensor

Jingtao Zhang, Yi Liu, Yanxun Lu, Stephen J. Redmond, Changhong Wang

PDF

AI summary

Key figure (auto-extracted from paper)
A hyperdimensional computing framework enables ultra-fast, lightweight incipient slip detection on edge devices, achieving 91.78% accuracy with 0.42 μs FPGA inference and a 0.375 kB model size.
Hyperdimensional computing incipient slip detection tactile sensing FPGA acceleration edge robotics PapillArray sensor

Problem

Existing learning-based slip detection methods suffer from high computational demands and detection latency, hindering real-time deployment on resource-constrained robotic platforms.

Approach

The authors map graphical, spatial, and temporal tactile data from a PapillArray sensor into high-dimensional binary vectors using a novel encoding scheme, then classify slip states via a context-driven training and inference strategy accelerated on an FPGA.

Key results

  • 91.78% offline slip detection accuracy
  • 0.375 kB model size for edge deployment
  • 0.42 μs FPGA inference latency (10^4× CPU speedup)
  • Validated online grip-force control in robotic experiments

Why it matters

Enables real-time, resource-efficient tactile perception for delicate robotic grasping and dexterous manipulation on edge hardware.

Abstract

Timely detection of incipient slip is critical for delicate robotic grasping and dexterous manipulation. However, existing learning-based methods suffer from detection latency and high computational demands. In this paper, we present an ultra-fast lightweight incipient slip detection framework based on hyper- dimensional (HD) computing, using the PapillArray optical tac- tile sensor. Our approach introduces a novel graphical-spatial- temporal HD encoding scheme coupled with a context-driven training and inference strategy, achieving a slip detection accu- racy of 91.78% in offline evaluation. The resulting model is ex- ceptionally compact and highly edge-compatible, with a size of only 0.375 kB. Furthermore, hardware acceleration on an FPGA enables inference within 0.42 microseconds, representing an over 104 × speedup compared to optimized CPU implementations. Online robotic experiments involving grip-force control based on the proposed slip detection method further validate its practical effectiveness. This work offers a practical and scalable solution for real-time slip detection in robotic manipulation tasks.

Index terms

Force and Tactile Sensing Hardware-Software Integration in Robotics Machine Learning for Robot Control

Related papers