ROBOLINGO : a CNL for talking to robots

Abstract

The aim of this project is to build a robot with basic physical capabilities and develop software with which it can receive natural language inputs and parse them into machine language commands. This will serve as a proof of concept of the idea that Human-Robot interfacing can be done using human language. As such, research into this area will lead to a more natural interface between humans and robots, using techniques which are currently emerging as popular trends in Artificial Intelligence. The project begins with the construction of a robotic frame. A Controlled Natural Language is then constructed which will be used to give the robot some understanding of commands. This will require (i) a grammar, (ii) some semantic representation of the CNL to control the system. Evaluation for this part of the system is tripartite and proof of the system working would require that all three parts evaluate with some success. First of all, the basic movements and the functions which they give rise to, along with the hardware and limitations of the system are evaluated as without them there can be no motion. Next the controller which interprets the semantic representation and runs a function is to be tested. Finally, the grammar and the lexicon are to be evaluated, where testing is conducted to ensure that the correct data structure is built for each sentence input. The project concludes with possible improvements to the system being identified during evaluation; they are then discussed in the section which follows.