Multi-sensor integration system with fuzzy inference and neural network

It is shown that a sensor integration system (SIS) with multiple sensors can expand the measurable region of an intelligent robotic system with high accuracy and that operators can use the system as easily as a single high-performance sensor system. The authors present an approach to the SIS using the knowledge database of sensors; the proposed SIS allows for the changing/replacing of sensors. The system consists of four subsystems: sensors performing as hardware sensing devices, knowledge database of sensors (KBS), fuzzy inference, and a neural network (NN). The sensor´s measurement value includes a measurement error. This system estimates the measurement error of each sensor using fuzzy inference. Fuzzy rules are developed from the KBS. Measurement values are integrated by the NN. All inferred measurement errors and measurement values are put into the NN. The NN output gives the integrated measurement value of multiple sensors. The proposed system is shown to be effective through a series of experiments