Rigid and Deformable Collision Handling for a Haptic Neurosurgery Simulator.

Abstract

Simulation has been widely used for training and rehearsing difficult or unusual actions in several fields such as aviation and the military. However, the simulators available in some disciplines do not fulfil the requirements of reliability and accuracy that users demand. This happens in neurosurgery. In order to overcome these difficulties, this thesis presents a multimodal Neurosurgery Simulator focused on patient-specific surgical learning and training. One of the aspects that most influences the behavioural reality of a simulator is the way in which the scene objects interfere. For that reason, detecting collisions and giving them a feasible response is particularly important. This work presents the collision handling methods for rigid and deformable volumetric objects and their haptic response to be integrated into the Neurosurgery Simulator. With the aim of evaluating our methods in terms of continuity and stability, the present document analyses the time consumption of the collision handling algorithms and the stability of the force parameters they return. Real-time virtual reality simulators require accuracy but are also time dependent. Thus, their computational cost is a vital aspect. This thesis also proposes a methodology to optimize the time consumption of collision detection algorithms that are based on the uniform spatial partition technique. It is validated experimentally and compared to other approaches. Additionally, the optimization is applied to our deformable collision detection method in order to improve its performance.