Sensor integration system for robotic applications

The authors propose an integration method of multiple sensors for application to a curved metal surface cutting system. For precise cutting and fast approaching, the system should use both highly accurate sensors and wide measurement range sensors. In order to use different types of sensors effectively, the authors propose a multi-sensor integration system based on neural network and fuzzy inference techniques. An experiment was performed in which three different types of eddy current sensors were used. This sensor suffers measurement error from the inclination angle between sensors and an object. By considering the inclination angle in this integration system, the approaching time is improved. The proposed system is shown to be effective through extensive experiments.<>