Human-Robot Collaboration by Intention Recognition using Deep LSTM Neural Network

Abstract

When a robot is required to perform specific tasks in human-robot-collaboration scenario, it is necessary for the robot to recognize human intention to more effectively and efficiently assist and interact with human. The relationship of the skeleton-based sequence of human motion provides a possible solution for the robot to recognize human intention. In this paper, we present a deep long short term memory (LSTM) neural network to recognize human intention. Considering the characteristics of deep LSTM neural network, it is a typically multiple stacked LSTM model, which combine the advantages of single LSTM layer and overcome the weakness of learning long-range time dependencies for RNN. The experimental results showed that the deep LSTM network with 2-layers have better prediction performance even only 40% of motion sequences is utilized.