Mechanical Engineering - Theses
Permanent URI for this collection
Search Results
1 - 1 of 1
-
ItemRobotic Haptic Object Identification through Grasping( 2023-05)Robotic haptic object identification is the process of identifying objects out of a given object set using a robotic hand equipped with tactile and finger-joint displacement sensors. Efficiency and accuracy, two essential evaluation metrics in haptic object identification, are the focus of this thesis. In terms of identification efficiency, when the robotic hand is smaller than an object, multiple grasps are required to capture the whole information of the object. However, from the practical consideration for robotic haptic object identification, it is always preferred to have the least number of grasps to identify an object. Regarding identification accuracy, when taking measurements by grasping the object, the uncertainties in the pose of the object relative to the hand will affect the identification accuracy. Each tactile sensor can capture contact within their specific local areas, thus, any change in the positions where objects make contact in relation to the robotic hand will significantly affect the tactile measurements. This thesis, therefore, aims to address the issues proposed above. The contributions of the thesis are: 1) An information gain-based method is proposed to improve the efficiency of haptic object identification by determining where to grasp to obtain the most distinguishing information about the object, thereby minimising the number of grasps needed; 2) A statistical method based on the Beta mixture model is proposed to improve the accuracy of haptic object identification by systematically characterising the uncertainties in the haptic measurements caused by the deviation in the relative pose between the object and the end-effector.