The Design and Implementation of a Voice-Programmable Robotic Assistant

Abstract

A recent surge in interest in open-software libraries such in areas of computer vision, voice recognition, and robotics operating system has empowered individuals and laboratories with powerful and accessible tools. Alfred is a pythonic voice-programmable application that implements some of the interesting packages openly available to researchers in Robotics and Intelligent Systems. This thesis presents the design of a unifying platform that glues face recognition, voice recognition, simultaneous mapping and localization, and chatbots in an operating system that is driven by voice driven operating system. Alfred runs on open-sourced Turtlebot robotics kit, which includes a Xbox Kinect v1.0 and a Kobuki base for range finding and mobile capabilities. From a set of predefined voice commands, Alfred allows the user to task the Turtlebot to travel to predefined locations, to look for faces a webcam, and to converse with a psychotherapy chatbot. For more complex tasks, Alfred has its own simple programming language that uses decision graphs to preform actions and examine outcomes. This thesis is not only a demonstration of interesting open-source packages, but also an example of a useful frameworks that integrate wildly different packages written in different languages created by different authors.