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

We developed an approach that allows robots to learn their own dynamics using only a first-person camera view, without any prior knowledge! 🎥💡

🦿 Tested on a 12-DoF robot, the self-supervised model showcased the capabilities of basic locomotion tasks.

🔧 The robot can detect damage and adapt its behavior autonomously. Resilience! 💪

🌍 The model proved its versatility by working across different robot configurations!

🔮 This egocentric visual self-model could be the key to unlocking a new era of autonomous, adaptable, and resilient robots.

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 their 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.

Authors: Yuhang Hu, Boyuan Chen, Hod Lipson
Affiliation: Columbia University, Duke University

Paper Link: https://arxiv.org/pdf/2207.03386.pdf
Website: yuhang-hu.com

Видео Egocentric Visual Self-Modeling for Autonomous Robot Dynamics Prediction and Adaptation канала Yuhang Hu