A pre-defined influence map was created to map out the region where the character is not suppose to move into. Using such a steering method, we can prevent characters from moving to prohibited areas by simulating collision detectiong without doing the actual expensive calculations.

In the demo program, the repulsion influence maps are pre-defined using an algorithm to mark out regions which the character will be pushed away. This prevents the character from walking through the obstacles. The force exerted by the influence map onto the character is inversely proportional to the distance from the center of the obstacles i.e. if the character gets nearer to the center, it will be repelled more.

The video below shows the algorithm at work. The character is scheduled to walk straight through the obstacle but her path is changed due to the repulsive forces. The white markings on the ground indicates the influence map (which will be hidden in actual games).

Video showing how a character steers round obstacles due to influence maps

An enhancement made to this program is that the influence map will affect the character less as it is more perpendicular to the influence force vectors. What this does is that the character will not be repelled by the walls when it is near the walls but not going against the wall.

Note: Scene setup done by Goh Cheng Teng. Model and animation provided by Dr Golam Ashraf.

Related posts

• Training Neural Networks
• Starry Starry Night Wins Codeathon!
• Starry Starry Night in GCA
• Starry Starry Night
• Hierarchical A* Pathfinding