Interactive directed exploration for mobile robots

Abstract

A mobile robot is often required to function intelligently within a dynamic environment with only partial knowledge of that environment. The robot is usually equipped with an exploration module that allows it to increase its knowledge of its surroundings by exploring unknown regions but it is often difficult for the robot to decide which are the most interesting places to visit first. In our experience the robot spends too much of its time exploring uninteresting regions. Existing robot control systems allow a user to intervene using a GUI program accepting input through a mouse and keyboard. In this thesis we improve on the method of human-robot interaction by building a system for guiding the robot during the exploration phase, using the more natural means of speech and vision. To build a robust system we chose to implement a behavior-based control architecture. The entire system consists of a large number of different behaviors (software modules) running on a number of different computers connected over a LAN. There are distinct behaviors for speech recognition, user finding, speech synthesis, sound localization, safe navigation, map building and high level reasoning. Using this interface, a person can get the robot's attention by calling its name. He can then direct the robot to a location to explore using a combination of verbal commands and hand gestures. We will show how the developed system will provide us with a platform for future research in human-robot interaction and behavior-based robot controllers.