Egocentric Visual Self-Modeling for Autonomous Robot Dynamics Prediction and Adaptation

TL;DR

This paper introduces a novel egocentric visual self-model for robots that learns their dynamics from a single camera in a self-supervised way, enabling autonomous detection of damage and adaptation without prior knowledge.

Contribution

It presents the first task-agnostic visual self-model learned solely from first-person camera input, capable of detecting damage and adapting behavior across different robot configurations.

Findings

Successfully learned dynamic self-model from visual input

Enabled autonomous damage detection and adaptation

Validated generalizability across multiple robot types

Abstract

The ability of robots to model their own dynamics is key to autonomous planning and learning, as well as for autonomous damage detection and recovery. Traditionally, dynamic models are pre-programmed or learned from external observations. Here, we demonstrate for the first time how a task-agnostic dynamic self-model can be learned using only a single first-person-view camera in a self-supervised manner, without any prior knowledge of robot morphology, kinematics, or task. Through experiments on a 12-DoF robot, we demonstrate the capabilities of the model in basic locomotion tasks using visual input. Notably, the robot can autonomously detect anomalies, such as damaged components, and adapt its behavior, showcasing resilience in dynamic environments. Furthermore, the model's generalizability was validated across robots with different configurations, emphasizing its potential as a universal tool for diverse robotic systems. The egocentric visual self-model proposed in our work paves the way for more autonomous, adaptable, and resilient robotic systems.