RBF Network Based Feature-Level Data Fusion for Robotic Multi-sensor Gripper

There are many kinds of sensors equipped on the robot gripper, for example, force sensor, proximity sensor, displacement sensor and etc. They can all be utilized to determine the state of connection. However, due to measurement error, uncertain work environment, no data of any certain sensor is sufficient to determine the state of connection. In order to grasp objects safely and reliably, the information fusion should be carried out for the output data of multi-sensors. In this paper, we present a novel technique to do the feature-level data fusion by RBF network. According to the result of fusion, control system can fast and accurately determine the state of connection between gripper and objects in real-time.