Problem with picking up “Mystery-Boxes” At first before the Agent became aware of there being “Mystery-Boxes” that exists, he would just drive casually around the map. Now with adding 3 rays to have the AI to look for boxes, if he scans boxes or a box he will pick randomly which one to grab as you can see in the Right Image, and if the agent misses the “Mystery-Box” he chose then boo-hoo, move on.

Calculating which path to take more often Before the agent could be aware of how pathways work he would just randomly choose which one to choose and we instead want the agent to calculate which one is a better path and why it is a better path. Different things the agent considers within each path is how long each path is, if they have a “Mystery-Box”, and the average speed each path is to take. calculating each of those components were simple but then when it came to calculating the total cost of each path, that’s where I began to struggle a bit; The average speed points was simple to scale because its an average ratio value, the “Mystery-Boxes” was also simple I just added a certain amount of points to the path if it had “Mystery-Boxes”, but then calculating points based on distance took me ages to figure out due to me knowing that if a path is shorter then it will gain more points than a path that has a longer distance. I ended up fixing the distance issue with lerp to a ratio of 0-1 but based on the shortest value and the max value and the path gains points based on if its shorter or longer. Now the agent has a probability of each path to get picked from their points combined. Something that I did add to make this more realistic of a choice is if a path has a “Mystery-Box” and the agent already has an item then he will less bother with those paths. in the Left Image there’s an example of the agent taking the red path and the brown arrows on the image show where the “Mystery-Boxes” are, but the reason he took that path is because he didn’t have an item yet so it increased the chances to take a path with “Mystery-Boxes”.

Thank You, Published By: Christian A. Camargo-Cortez Posted on: 4/12/2024