Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

HyperSurf: Quadruped Robot Leg Capable of Surface Recognition with GRU and Real-to-Sim Transferring

Sergei Satsevich, Yaroslav Savotin, Danil Belov, Elizaveta Pestova, Artem Erhov, Batyr Khabibullin, Artem Bazhenov, Vyacheslav Kovalev, Aleksey Fedoseev, Dzmitry Tsetserukou

TL;DR

HyperSurf tackles the challenge of surface recognition for safe quadruped locomotion by introducing a lab-scale one-leg stand to accelerate data collection and enable real-to-sim transfer for gait optimization in simulation. The approach builds on a GRU-based recognition pipeline and extends DogSurf, while also establishing a Gazebo-based digital twin workflow to close the sim-to-real loop via IMU data and calibrated surface models. Key contributions include a 1,000,000+ sample dataset across four surfaces, a 7.5x speedup in data-processing throughput, and high-fidelity real-to-sim transfer demonstrated by 98% correlation in angular velocities. The work enables rapid dataset creation, cost-effective experimentation, and practical pathways for deploying robust surface-aware quadruped locomotion in real-world and simulated environments with broad application potential across inspection, agriculture, and assistive robotics.

Abstract

This paper introduces a system of data collection acceleration and real-to-sim transferring for surface recognition on a quadruped robot. The system features a mechanical single-leg setup capable of stepping on various easily interchangeable surfaces. Additionally, it incorporates a GRU-based Surface Recognition System, inspired by the system detailed in the Dog-Surf paper. This setup facilitates the expansion of dataset collection for model training, enabling data acquisition from hard-to-reach surfaces in laboratory conditions. Furthermore, it opens avenues for transferring surface properties from reality to simulation, thereby allowing the training of optimal gaits for legged robots in simulation environments using a pre-prepared library of digital twins of surfaces. Moreover, enhancements have been made to the GRU-based Surface Recognition System, allowing for the integration of data from both the quadruped robot and the single-leg setup. The dataset and code have been made publicly available.

HyperSurf: Quadruped Robot Leg Capable of Surface Recognition with GRU and Real-to-Sim Transferring

TL;DR

HyperSurf tackles the challenge of surface recognition for safe quadruped locomotion by introducing a lab-scale one-leg stand to accelerate data collection and enable real-to-sim transfer for gait optimization in simulation. The approach builds on a GRU-based recognition pipeline and extends DogSurf, while also establishing a Gazebo-based digital twin workflow to close the sim-to-real loop via IMU data and calibrated surface models. Key contributions include a 1,000,000+ sample dataset across four surfaces, a 7.5x speedup in data-processing throughput, and high-fidelity real-to-sim transfer demonstrated by 98% correlation in angular velocities. The work enables rapid dataset creation, cost-effective experimentation, and practical pathways for deploying robust surface-aware quadruped locomotion in real-world and simulated environments with broad application potential across inspection, agriculture, and assistive robotics.

Abstract

This paper introduces a system of data collection acceleration and real-to-sim transferring for surface recognition on a quadruped robot. The system features a mechanical single-leg setup capable of stepping on various easily interchangeable surfaces. Additionally, it incorporates a GRU-based Surface Recognition System, inspired by the system detailed in the Dog-Surf paper. This setup facilitates the expansion of dataset collection for model training, enabling data acquisition from hard-to-reach surfaces in laboratory conditions. Furthermore, it opens avenues for transferring surface properties from reality to simulation, thereby allowing the training of optimal gaits for legged robots in simulation environments using a pre-prepared library of digital twins of surfaces. Moreover, enhancements have been made to the GRU-based Surface Recognition System, allowing for the integration of data from both the quadruped robot and the single-leg setup. The dataset and code have been made publicly available.
Paper Structure (17 sections, 6 figures, 2 tables)

This paper contains 17 sections, 6 figures, 2 tables.

Table of Contents

  1. Introduction
  2. Related Works
  3. System Overview
  4. Hardware Architecture
  5. HyperSurf Control System
  6. Conducting the Experiment
  7. Dataset Structure
  8. GRU-based Surface Recognition
  9. Creating a Digital Twin of HyperSurf System
  10. Preparation of the stand description file
  11. Kinematics Control Implementation
  12. Dataset Collection
  13. Novelty of HyperSurf System
  14. Boosting Dataset Collection Efficiency
  15. Optimizing Workforce for Dataset Collection
  16. ...and 2 more sections

Figures (6)

  • Figure 1: The experimental stand consists of a leg, fixing stand, control electronics, IMU sensor, and the surface sample under investigation.
  • Figure 2: HyperSurf system overview, including modules and interfaces developed for surface recognition.
  • Figure 3: Neural network architecture.
  • Figure 4: Confusion matrix for HyperSurf.
  • Figure 5: Digital twins of the one-leg stand in Gazebo simulator.
  • ...and 1 more figures