Speech recognition in construction equipment : Creating a voice assistant for an autonomous wheel loader

Abstract: This thesis sets out to explore possible applications of speech recognition in construction equipment and autonomous machines. Advancements in autonomous vehicle technology mean that soon, vehicles like wheel loaders will be able perform tasks without human operators. Those vehicles still require a method of interaction with humans and recent improvements in speech recognition mean that it is possible for a natural voice-based interface to be used. The research question of this thesis is the extent to which voice control can replace hand-operated controls in an intelligent autonomous machine. Interviews and observation sessions took place in order to identify the requirements such a speech interface would have to fulfill. Next, a design process took place in order to build a prototype of such system, followed by test sessions to evaluate it. The prototype demonstrated positive attributes, with great learnability and ease of operation, but speech recognition errors meant low performance, and overall user satisfaction.