Obstacle avoidance is a very important part of bot navigation, and it is essential to get it right. Many solutions have been devised in the past to solve this problem, but none provide the simplicity that neural networks offer. Other solutions require the programmer to define and tweak equations to get the desired behaviour. With the neural network approach, high-level parameters can be set in order to implicitly specify the solution. This provides an undeniable flexibility for the system. There is also little need to understand the details of obstacle avoidance, as the system evolves to an automatically near optimal configuration on its own. Finally, the efficiency of this solution implies that it is ideally suited to a game engine, where numerous bots may need to be present.

The next tutorial will dive deeper into specific genetic algorithm issues. This will not directly influence the obstacle avoidance, but will allow us to create near-optimal solutions more consistently and with fewer generations. Other nifty tricks more directly related to obstacle avoidance will be introduced, allowing us to extend the solution a bit more. We will also consider more advanced techniques such as speed control and potentially crouching or jumping. See you then

Remember you can visit the Message Store to discuss this tutorial. Comments are always welcome!. There are already replies in the thread, why not join in?