Multi-sensory integration as a result of perception

Conference · Tue Sep 13 04:00:00 EDT 1988

OSTI ID:6792110

Stansfield, S A

For the most part, research into multi-sensor data fusion has concentrated on two areas: sensing and sensor modeling, and the creation of some central, usually geometric, representation of the world. From our perspective, this approach is narrow and local. The perceptual system will in fact need to integrate sensed data in many different ways/endash/homogeneously, heterogeneously, supportively, and directly. Integration of sensory information takes place at several different levels and utilizes several different mechanisms. A first level of integration takes place as sensor primitives are extracted from the world and combined to form more complex features. As these features are identified, they may be used to invoke motor functions which extract other, related, features from the object being perceived. This use of one set of sensory inputs to guide the extraction of others may be thought of as the second level of integration carried out by the perceptual system. Finally, at the highest level, multiple, disparate sensory features are aggregated into a heterogeneous central representation of the object as a whole. How the information contained within this aggregrate is used by the system is a function of both the available information and the task at hand. Thus, at this level, integration may be thought of as a top-down, knowledge-driven process. This paper explores how this model of multi-sensory integration might be implemented and utilized within a robotic system equipped with visual and tactile sensors. 21. refs., 5 figs.

• Research Organization: Sandia National Labs., Albuquerque, NM (USA)
• DOE Contract Number: AC04-76DP00789
• OSTI ID: 6792110
• Report Number(s): SAND-88-2657; CONF-8810164-1; ON: DE88017343
• Country of Publication: United States
• Language: English

Related Subjects

99 GENERAL AND MISCELLANEOUS
990210* -- Supercomputers-- (1987-1989)
ARTIFICIAL INTELLIGENCE
AUTOMATION
CONTROL SYSTEMS
KNOWLEDGE BASE
REMOTE SENSING
ROBOTS