Self-supervised Learning Of Visual Pose Estimation Without Pose Labels By Classifying LED States

2025-09-12

AI Summary

This work addresses the challenge of self-supervised relative pose estimation for ground robots using only monocular RGB images—without pose labels, CAD models, or appearance priors. We propose a novel proxy-task framework leveraging LED state classification as a self-supervised signal. Specifically, we design an LED state recognition network and integrate it with monocular depth-ambiguity correction and joint relative orientation regression modules, trained end-to-end on video sequences captured during random multi-robot motion. Our key contribution is the first use of visible-light LED states as a weakly supervised proxy for pose learning, eliminating reliance on ground-truth poses, external localization systems, or robot geometric priors. Experiments demonstrate that our method achieves accuracy comparable to state-of-the-art (SOTA) approaches requiring pose labels or CAD models, while exhibiting strong cross-domain generalization and native support for collaborative multi-robot pose estimation.

Technology Category

• Intelligent Robots: State Estimation
• Computer Vision: Multi-modal Vision
• Machine Learning: Multimodal Learning

Application Category

• Responsible Web: Machine-in-the-loop, human agency and autonomy
• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• Graph Algorithms and Modeling for the Web: Representation, reconstruction, and subgraph or motif discovery in Web-related graphs

Abstract

We introduce a model for monocular RGB relative pose estimation of a ground robot that trains from scratch without pose labels nor prior knowledge about the robot's shape or appearance. At training time, we assume: (i) a robot fitted with multiple LEDs, whose states are independent and known at each frame; (ii) knowledge of the approximate viewing direction of each LED; and (iii) availability of a calibration image with a known target distance, to address the ambiguity of monocular depth estimation. Training data is collected by a pair of robots moving randomly without needing external infrastructure or human supervision. Our model trains on the task of predicting from an image the state of each LED on the robot. In doing so, it learns to predict the position of the robot in the image, its distance, and its relative bearing. At inference time, the state of the LEDs is unknown, can be arbitrary, and does not affect the pose estimation performance. Quantitative experiments indicate that our approach: is competitive with SoA approaches that require supervision from pose labels or a CAD model of the robot; generalizes to different domains; and handles multi-robot pose estimation.

Problem

• Self-supervised pose estimation without labeled data
• Monocular RGB relative pose prediction for robots
• Learning from LED states instead of pose labels

Innovation

• Self-supervised learning without pose labels
• LED state classification for visual training
• Monocular RGB relative pose estimation