Identifying Terrain Physical Parameters from Vision -- Towards Physical-Parameter-Aware Locomotion and Navigation

TL;DR

This paper introduces a self-supervised learning framework that estimates environmental physical parameters like friction and stiffness from vision, enabling robots to anticipate terrain properties for improved locomotion and navigation.

Contribution

It presents a novel cross-modal self-supervised approach that trains a physical decoder in simulation and applies it to real-world vision-based terrain property estimation.

Findings

Outperforms existing baseline methods in simulation and real-world tests

Enables dense prediction of physical properties from images in diverse environments

Allows fast adaptation to new terrains for robotic navigation

Abstract

Identifying the physical properties of the surrounding environment is essential for robotic locomotion and navigation to deal with non-geometric hazards, such as slippery and deformable terrains. It would be of great benefit for robots to anticipate these extreme physical properties before contact; however, estimating environmental physical parameters from vision is still an open challenge. Animals can achieve this by using their prior experience and knowledge of what they have seen and how it felt. In this work, we propose a cross-modal self-supervised learning framework for vision-based environmental physical parameter estimation, which paves the way for future physical-property-aware locomotion and navigation. We bridge the gap between existing policies trained in simulation and identification of physical terrain parameters from vision. We propose to train a physical decoder in simulation to predict friction and stiffness from multi-modal input. The trained network allows the labeling of real-world images with physical parameters in a self-supervised manner to further train a visual network during deployment, which can densely predict the friction and stiffness from image data. We validate our physical decoder in simulation and the real world using a quadruped ANYmal robot, outperforming an existing baseline method. We show that our visual network can predict the physical properties in indoor and outdoor experiments while allowing fast adaptation to new environments.