Collision Response For Virtual Laparoscopic Surgery
During laparoscopic surgery the surgeon operates inside the abdomen guided by the 2D imagery of the operating area. This image is obtained via a camera inserted through a small incision. Laparoscopic surgery, in contrast to open surgery, results in low pain and recovery time for the patient. However, the surgeons have some difficulty, as the training required for this type of surgery is more intense and rigorous than that for conventional open surgery. Hence, surgical trainers for laparoscopic surgery are important. The main difference in the different types of surgical trainers is the degree to which the actual environment of the surgical procedure is depicted. This thesis describes a virtual reality based surgical trainer and examines the problem of collision response and improvements to the current algorithms for applying physical and visual response in real-time. Mass-spring models are used to simulate the behavior of the various objects, which require extensive numerical methods for implementation. Hence, the computations required to calculate the new parameters of the mass-spring system must be inexpensive. The methods that we describe and have implemented make a few approximations so as to make the system real-time. We classify the response according to the type of collisions so as to get the best possible response after consideration to the various constraints. We also investigate the problem of inter-penetrations of objects due to the lag in discrete-time sampling for collision detection and propose an algorithm to overcome inter-penetrations. We have implemented this fast and stable collision response algorithm.