Dynamic pivot point technique (DYOP) in narrow-phase collision detection

• Main Author: Hamzah Asyrani Sulaiman
• Format: Thesis
• Language: English
• Published: 2015
• Subjects: TR Photography
• Online Access: https://eprints.ums.edu.my/id/eprint/18899/1/Dynamic%20pivot%20point%20technique.pdf; https://eprints.ums.edu.my/id/eprint/18899/

Contact determination between nearly colliding objects is one of the crucial issues in virtual environment simulation particularly for collision detection system such as in medical simulation, computer games, and engineering visualization. The contact determination technique consists of distance computation, point of contact and depth penetration phase. It requires information of object direction towards other object that moves in opposite direction namely intersection. The main problem in this research is to study a nearly accurate distance approximation computation caused by heavy calculation that deteriorating speed and efficiency of collision detection system. This thesis aims to introduce an agile and new technique for contact determination in narrow phase collision detection that increases speed and improves the efficiency of narrow phase collision detection. The proposed narrow phase collision detection technique in this study consists of distance computation method, point of contact between nearly collided object and depth penetration method namely Dynamic Pivot Point (DyOP). The technique enables numbers of testing to be reduced as compared to the prominent techniques such as Un-Canny and GJK technique. The proposed technique starts by first identifying the nearest triangle of corresponding objects that is going to be intersected with another triangle that are bound with an Axis Aligned Bounding-Box (AABB). Then, we need to find the parallel line distribution created by each axis of AABB for each triangle. Next, an internal intersection using internal AABB by using maximum and minimum parallel line distribution of each object. A pivot point called DyOP will be calculated as a reference point to the contact determination technique. The proposed technique is able to reduce nine vertices testing down to six vertices testing and nine vertex-edge testing's down to only two vertex-edge testing's. Three sets of testing have been conducted to evaluate and verify the proposed technique with two other prominent techniques. The first test was conducted by investigating time execution of ten different triangles with pre-defined size but vary in distance computation phase. Based on the experiments, out of 90 concurrent testing's, the proposed technique managed to achieve 187.3% improvements with the fastest execution time about 55 milliseconds. The subsequent test was conducted with the point of contact phase. Based on the experiments conducted, the proposed technique successfully attained 79.05% improvement with an average of 7.6 milliseconds and almost similar to Uncanny technique for point of contact. The final test to determine depth penetration, the technique managed to achieve 53.63% better improvement and an average of 10.75 milliseconds per test with similar accuracy. The results from the experiments showed that the DyOP technique is efficient to perform contact determination in narrow phase collision detection and increase the speed of distance computation, point of contact and depth penetration by reducing the number of 'testing. These indicated that, the DyOP technique is efficient, accurate and robust not only to the tested environment but other unprepared environment and could be adapted to any desired target area or domain.