MULTISENSOR INTEGRATION SYSTEM BASED ON FUZZY INFERENCE AND NEURAL-NETWORK

This paper deals with a multisensor system applied to a carve robot system. This robot system carves products using the surface shape data that are obtained by measurement. Measurement is implemented by sensor array at the tip of the five axis manipulator. The sensor array is carried to the surface of the target by the manipulator. The manipulator approaches the surface by using the sensors' outputs. For precise carving, the system should measure the surface of the work by precise sensors. However, because these sensors do not have long measurement range, it takes time to approach the work. On the other hand, in order to approach the work fast, the system should use long measurement range sensors. For precise carving and fast approach, the system should use both accurate sensors and long measurement range sensors. In order to use these sensors effectively, we use the multisensor integration system based on the neural network and fuzzy inference techniques. Then the system can consider the condition of each sensors for measuring. With consideration of the condition, approach time is improved. The proposed system is shown to be effective through extensive experiments.