Multi-sensor/knowledge fusion

Abstract

This paper presents a new approach to solve the sensor fusion problem. This new approach is called multi-sensor/knowledge fusion (MSKF) which takes into account not only the fusion of information between multiple sensory systems but also the fusion between sensory information with system's prior knowledge of its environment. This MSKF architecture is a modification of the real-time control system (RCS) developed by Albus (1997). This RCS was designed to be a reference control architecture model for creating intelligent systems. The RCS itself uses a hierarchical control strategy with each layer of the hierarchy decomposes into plans for parallel and sequential sub-tasks. However, the sensory processing module in the RCS was re-designed in the MSKF architecture for the use of multiple types of sensors. The test of this new control architecture is carried out on a LEGO toy kit assembly task.