Vision-Based Multi-Task Manipulation for Inexpensive Robots Using End-To-End Learning from Demonstration

TL;DR

This paper introduces a multi-task learning approach for low-cost robots using end-to-end deep learning from demonstration, enabling complex manipulation tasks directly from raw images with improved robustness.

Contribution

It presents a novel recurrent neural network controller that learns multiple manipulation tasks simultaneously from raw images, combining VAE-GAN reconstruction and autoregressive action prediction.

Findings

Successful learning of complex tasks like towel wiping and placement

Weight sharing and reconstruction regularization enhance generalization

Multi-task training improves success rates across tasks

Abstract

We propose a technique for multi-task learning from demonstration that trains the controller of a low-cost robotic arm to accomplish several complex picking and placing tasks, as well as non-prehensile manipulation. The controller is a recurrent neural network using raw images as input and generating robot arm trajectories, with the parameters shared across the tasks. The controller also combines VAE-GAN-based reconstruction with autoregressive multimodal action prediction. Our results demonstrate that it is possible to learn complex manipulation tasks, such as picking up a towel, wiping an object, and depositing the towel to its previous position, entirely from raw images with direct behavior cloning. We show that weight sharing and reconstruction-based regularization substantially improve generalization and robustness, and training on multiple tasks simultaneously increases the success rate on all tasks.