Responsive Drone Autopilot System for Uncertain Natural Language Commands

Abstract

The purpose of this investigation is to realize how possible is it to control a drone with human's English natural language commands, even the command contains some quantitative uncertain words. Basically, the system interprets the commands with uncertain words, into a machine understandable format. According to the literature review, human-robot association does not go with uncertain words and it is a considerable gap. This proposed system has a client side mobile application to input voice. To build a machine understandable command, it refers to an ontology based knowledge store. The knowledge store gains information and expands its knowledge by crawling websites. Once the user commands, the drone to make a movement, the system will analyze the instruction to check if there is uncertainty and if it points to a surrounding object that is visible to the drone's camera. For extracting the uncertainty it uses NLTK along with a specific grammar rule and that methodology is succeeded according to the results. After collecting that information the system can navigate through the user's instruction. This research helps non-skilled drone pilots to have a smooth flying experience, and also other researchers to discover about natural language processing with robotics.