Feature-level data fusion of robotic multi-sensor gripper using ANN

Several kinds of sensors are installed in the robotic gripper. According to the outputs of multi-sensor, a data fusion technique is utilized to ensure the robot walking or grasping objects safely and reliably. In this paper, sensors of the gripper are introduced, such as force sensors for contact sensing and gripping force control, proximity sensors for collision prevention and position detection, and a displacement sensor for gripper openness control. The experiments of grasping objects with the gripper are presented, including firm grasp, virtual grasp, half grasp, skew grasp, empty grasp and so on. The accurate information of grasping objects with the gripper is obtained using the multi-sensor data fusion technique based on the BP artificial neural network.