How to teach a bot to navigate in the game world using the minimap?

Question

Neural networks

Good afternoon,
I am developing bots - clickers for games for windows. Clickers use the keyboard and mouse to send commands, such as moving a character in a 3D game world. Almost every walker has a minimap, with which you can determine whether the character is in a certain coordinate of the game world. Advise how to teach a bot to move independently between locations using a neural network? We will assume that the full map of the area is not known to us in advance, and during training, the bot is controlled by a person.

Answer 1

1. If the training will take place online (a person plays, and the bot is simultaneously trained on the data received), then training will probably take a very long time for the bot to learn at least something.
2. I think that the algorithm can strongly depend on the nature of the map - either it is a labyrinth, or it is an open area with areas prohibited for movement.
3. In general, the task of finding a path is a fairly simple problem. If the whole map is known, then it is solved without any neural networks - for example, A*. Maybe there are some variants of algorithms for partially observable environments.

Answer 2

Advise how to teach a bot to move independently between locations using a neural network?

I advise you to study ANN (especially since you are developing bots).