A human machine interface framework for autonomous vehicle control

The recent development of autonomous vehicles has attracted much attention. The technologies for persons without technical knowledge to handle autonomous vehicles easily are highly demanded. We are developing an intuitive multimodal interface system using speech, gesture, and eye-gaze recognition. We designed a multimodal understanding component and dialog control component of the interface system separately using finite state transducers. As for implementation, conventional dialogue systems are controlled by a finite state transducer with user actions as inputs and system actions as outputs. Our multimodal understanding and dialog control component can be seen as a cascade of two separate transducers. Cascaded transducers can be composed to one transducer. Our system can operate a virtual car in an autonomous car simulator.