Sensorized objects used to quantitatively study distal grasping in the African elephant

Abstract

Nature evolved many ways to grasp objects without using hands: elephants, octopuses, and monkeys use highly dexterous appendices. From a roboticist's perspective, the elephant trunk is a fascinating manipulator, which strategies can empower robots' interaction capabilities. However, quantifying prehensile forces in such large animals in a safe, ethical, and reproducible manner is challenging. We developed two sensorized objects to investigate the grasping of an adult African elephant with deliberately occluded vision. A cylinder and a handle provided a distributed force (80 and 6 taxels) and inertial measurements in real-time, resisting dirt and shocks. The animal curled the distal portion of the trunk to grasp the tools. Using force and contact area data of the cylinder revealed the animal's ability to finely modulate pressure. The handle data provided insights into the energy-efficient behavior of the animal, with no significant grasping force changes despite variations imposed on both weight (5-15kg) and initial position of the object.