AI Bootcamp

Learning of AI with a scholar project named AIBootcamp, creating an adaptative algorithm to navigate Npcs on an hexagonal environment.  This project was a 3 months project, firstly solo, then duo,  and at the end a group of 4 .

This project highlighted many of the game-AI fields, such as path finding (A*), behaviour trees, finite state machines, and exploration using utility AI in order to control movement and actions of Npcs. 
Alongside the AI Boot Camp, we learned other game-AI related algorithms, such as Reciprocal Velocity Obstacle, Crowd Behaviour, GOAP/HTN, and many others.

The AI Boot Camp documentation : 
https://sites.google.com/view/aibootcamp/home

The goal is to move the NPC toward green tiles on a map. If all Npcs are on a green tile, it leads to the success of the map. The algorithm goal is to succeed all map from 1 to 80. 
 However, there are multiple obstacles that appear in each levels : 
- The pawn's vision are limited
- The calculation time cannot exceed 15ms, even when there are a lot of pawns on a huge grid
- Some tile are not walkable
- There are obstacles along the way (walls, glass, doors)
- Some of the doors are hidden inside a wall, and some other can only be opened by triggering a pressure plate
- ... and in each iteration, there were new rules added to the game, making it harder to navigate