Deep Visual Foresight for Planning Robot Motion

TL;DR

This paper introduces a method that combines deep video prediction with model-predictive control, allowing robots to learn and perform manipulation tasks using only unlabeled data without human supervision.

Contribution

It presents a novel approach that enables robots to learn predictive models and control strategies solely from unlabeled data, removing the need for calibration or precise sensing.

Findings

Robot successfully performs pushing tasks with unseen objects.

Method works without camera calibration or precise sensors.

Enables autonomous learning from unlabeled data.

Abstract

A key challenge in scaling up robot learning to many skills and environments is removing the need for human supervision, so that robots can collect their own data and improve their own performance without being limited by the cost of requesting human feedback. Model-based reinforcement learning holds the promise of enabling an agent to learn to predict the effects of its actions, which could provide flexible predictive models for a wide range of tasks and environments, without detailed human supervision. We develop a method for combining deep action-conditioned video prediction models with model-predictive control that uses entirely unlabeled training data. Our approach does not require a calibrated camera, an instrumented training set-up, nor precise sensing and actuation. Our results show that our method enables a real robot to perform nonprehensile manipulation -- pushing objects -- and can handle novel objects not seen during training.