PROBE: Proprioceptive Obstacle Detection and Estimation while Navigating in Clutter

2025-05-17

AI Summary

To address the challenge of perceiving occluded obstacles in visually constrained, confined environments (e.g., rubble search-and-rescue), this paper proposes a purely proprioceptive, end-to-end method for estimating the geometric parameters of rectangular obstacles. Leveraging only joint torques and whole-body motion history—without external sensors such as cameras or LiDAR—the approach employs a Transformer architecture to model spatiotemporal dynamics and kinematics, enabling real-time estimation of occluded rectangular obstacles’ position, orientation, and dimensions directly in the SE(2) space. It requires no prior map or visual input. Evaluated jointly in Isaac Gym simulation and on a physical Unitree Go1 quadruped robot, the method achieves centimeter-level accuracy in size estimation and degree-level precision in pose prediction, significantly enhancing robustness for vision-free navigation. To our knowledge, this is the first work to enable online, proprioception-only geometric parameter estimation of obstacles.

Technology Category

• Intelligent Robots: State Estimation
• Search and Optimization: Learning to Search
• Computer Vision: Low Level & Physics-based Vision

Application Category

• Graph Algorithms and Modeling for the Web: Efficient manipulation of static and dynamic Web-related graphs
• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• Systems and Infrastructure for Web, Mobile and WoT: Applied ML and AI for Web-based mobile applications

Abstract

In critical applications, including search-and-rescue in degraded environments, blockages can be prevalent and prevent the effective deployment of certain sensing modalities, particularly vision, due to occlusion and the constrained range of view of onboard camera sensors. To enable robots to tackle these challenges, we propose a new approach, Proprioceptive Obstacle Detection and Estimation while navigating in clutter PROBE, which instead relies only on the robot's proprioception to infer the presence or absence of occluded rectangular obstacles while predicting their dimensions and poses in SE(2). The proposed approach is a Transformer neural network that receives as input a history of applied torques and sensed whole-body movements of the robot and returns a parameterized representation of the obstacles in the environment. The effectiveness of PROBE is evaluated on simulated environments in Isaac Gym and with a real Unitree Go1 quadruped robot.

Problem

• Detects occluded obstacles using robot proprioception only
• Predicts obstacle dimensions and poses in cluttered environments
• Uses Transformer network with torque and movement history

Innovation

• Uses proprioception for obstacle detection
• Employs Transformer neural network
• Predicts obstacle dimensions and poses