Heterogeneous sensors cooperation for an advanced perception system

A third generation robot evolving in a changing world has to model the external events in a way that allows it to operate safely in all circumstances. Hybrid models are well adapted to partially known environments characterized by smooth variations interlaced with almost discontinuous context jumps. An approach is suggested, as far as the external world perception is concerned, where a rough sensor (the hit-or-miss type and dedicated to the context discrimination) operating in parallel with a conventional sensor mounted on the robot is used to enhance the overall state estimation. The design may be optimized. Using the method, context identification and trajectory planning can be achieved for a minimum computational cost increase.<>