Commanding Mobile Robot Movement based on Natural Language Processing with RNN Encoder-Decoder

This work utilizes the potential of NLP and machine learning for the challenging task of human-machine communication. A task of robot movement is selected as the context of the research work where the goal is to create a software system that receives natural language input movement command from human and produces the set of precise trajectory information for the robot to perform. The proposed system consists of Pre-processing function, Command classification, Parameter classification, Post-processing function where RNN Encoder-Decoder is used for the implementation of the classification process. The system was trained using a dataset of 1,600 unique entries. The experiment results show that the average accuracy in case of single movement command is 79.23% whereas the average accuracy in case of multiple command in one sentence is 73.65%