DMotion: Robotic Visuomotor Control with Unsupervised Forward Model Learned from Videos
This addresses the challenge of reducing data labeling demands in robotic control, though it is incremental as it builds on unsupervised learning methods.
The paper tackles the problem of learning accurate forward models for robotic visuomotor control without labeled data by proposing DMotion, which trains from videos only by disentangling agent motion, achieving superior performance in Grid World and simulated robot environments.
Learning an accurate model of the environment is essential for model-based control tasks. Existing methods in robotic visuomotor control usually learn from data with heavily labelled actions, object entities or locations, which can be demanding in many cases. To cope with this limitation, we propose a method, dubbed DMotion, that trains a forward model from video data only, via disentangling the motion of controllable agent to model the transition dynamics. An object extractor and an interaction learner are trained in an end-to-end manner without supervision. The agent's motions are explicitly represented using spatial transformation matrices containing physical meanings. In the experiments, DMotion achieves superior performance on learning an accurate forward model in a Grid World environment, as well as a more realistic robot control environment in simulation. With the accurate learned forward models, we further demonstrate their usage in model predictive control as an effective approach for robotic manipulations.