A natural language processing approach for mobile service robot control

Abstract

For the operation of mobile robots, the design of the man-machine interface has a major impact on the acceptance by the user. This is especially valid when dealing with users of minor or no training, which can be found in almost any application area, ranging from transport or construction tasks to assistive technology for elderly and disabled. The following essay introduces a natural language processing software which transforms a natural language sentence into a series of commands for a mobile service robot. Both the language processor and the command set of the controller may be modified by the user to adapt it to different application areas. The language processor is based on a modified finite state transducer network. The grammar/word pattern extracted by the network is processed by a transformation module which converts the contained information into the controller command. A context memory and a question generator allow the processing of incomplete instructions and command modifications. A prototype application is presented which allows severely disabled persons bound to a wheelchair to control the chair and their home environment by natural language.