The implications of natural language structure on the control of intelligent robotic systems

Abstract Control theory focuses on the means by which arbitrary controlling signals adjust and maintain performances of arbitrary systems. Any developments, then, that would qualitatively change the variety of controlling signals available or the mechanics for using these signals to effect control, would be of great interest to the field. This paper describes the prototype of a natural language control system that can make just such changes, a system that accepts ordinary English sentences as control input for a robotic manipulator and transforms the input into performance. The development of this system has proven to be a difficult enterprise, to some extent due to the differences between human and robotic hardware: natural language has evolved to support communication among beings of a hardware design different than that of robots. It has required the specification of a new model of natural language, named Residential Grammar, that embodies concise descriptions of syntactic, semantic, and pragmatic know...