Calibration of Virtual Contact Forces with Real Contact Scenarios

Haptics means sensing and manipulation through touch. The aim of haptic rendering is to make able the user to touch, feel and manipulate the objects through a haptic interface which is a user interface that allows a user to receive haptic feedback via movement of a probe. Haptic rendering algorithms solve two major problems which are collision detection and collision response. When the user manipulates the probe, the new position is set to the haptic probe and collisions of the virtual objects are detected by the haptic rendering algorithms. When the collision is detected, haptic rendering algorithms generate virtual forces-torques that might have a meaning, but it is not the real one. The purpose of this master thesis is to minimize the di erence between the real and virtual forces-torques. In order to achieve this goal, the initial step is literature research on the benchmarking of haptic rendering algorithms. This thesis compares Voxmap-Pointshell Algorithm (VPS), The Inner Sphere Tree Algorithm (IST), RAPID, PQP, V-COLLIDE and PolyDepth algorithms according to the benchmarking suite presented by the German Aerospace Center (DLR). This benchmarking suite presents a set of testing scenarios and compares the performance and quality of haptic rendering algorithms using synthetically generated paths. The results of benchmarking show the position of VPS algorithm in the haptic rendering algorithms. This result is important because the virtual forces-torques computed by the VPS algorithm are compared with the real ones recorded by force-torque sensors. For that two real scenarios are de ned and experiments are repeated in reality and virtual reality. During the experiments in reality, a physical probe which is moving at the surface of a physical object never penetrates into the object. However, the virtual haptic probe should penetrate the surface of a virtual object to compute a force for the VPS algorithm. Because of this reason, even if a user tries to simulate the same virtual probe movements with the real probe, a virtual trajectory can not be identical with a recorded trajectory. This thesis presents an o¬ine calibration method which is trying to nd the virtual inner trajectory equivalent to the real one. O¬ine calibration method results in the same force-torque output for both real and virtual scenarios