Return to Home ‣ Courses ‣

Collisions Detection

The original question Link

The Code and models used in this project

Introduction

In this project, I use SWIFT++ to detect collision among objects. It's very convient to use, and can detect allcollisions among many objects. Following is something need to note:

Experiments

I use four obj modesl, read their vertices and faces and give them to Add_Convex_Object. These models are:

Below is the running results. The first one record frames, the second one record collision detection times in one second, both with varying different models and objects number.

From the two images, I have several observations:

Observations
①	both the running and collision detection time satisfy a quadratic curve regards to object number
②	rendering is more time consuming than collision detection
③	The complexity of object has little influence on collision detection, since swift++ may use surrounding area to simplify

I change the scale from 1 ,2 ,4 ,8 , 10, and record the detection times and collision times in the following two images.

Observations
①	detection times decrease along with scale increase in a linear manner
②	collision times increase exponentially along with scale increase





Change the algorithm

I adopt the bounding sphere algorithm and generate collision detection times (per second)-- object number and collision times -- object number graph.





This method runs faster than Swift++, but for complex objects this method will be not accurate.

☜ ☞

© 2011–12 Dinghuang Ji