Change dot cluster behavior #36
Conversation
|
Ok, this one should be good to go as well. I had to build it on top of #35 therefore I set that as the target branch for now but it should go into the demo branch. |
ll-nick
force-pushed
the
eat_closest_dot_behavior
branch
from
abe8f0c
to
e43d8a2
Compare
Rename clusterFinder to dotClusterFinder
orzechow
force-pushed
the
change_dot_cluster_behavior
branch
from
b100b1b
to
063ae1e
Compare
|
I had to rebase on top of pacman-demo and resolve some minor conflics (introduced by the review changes of the previous PR, I guess) See also 063ae1e for some more optional use in A* – please check, if that makes sense 😁 |
|
Thanks for the rebase, the optional Path stuff (which I meant to add but obviously forgot) and the detailed review. I skimmed through it briefly and I'm afraid I agree with pretty much all of your suggestion 😂 |
|
Closes #26
This should be the final behavior 🥳
I did go for a cluster segmentation algorithm (just a simple breadth first search) as opposed to the quadrant thing we discussed since it seemed more sensible when I started implementing it.
So we now split the dots into clusters (sets of connected dots) using the clustering algorithm in
cluster.hpp. I did move theMazeAdapterandCellclass intomaze.hppand generalized it such that both the clustering algorithm and the A* could utilize these classes. For each cluster, we also compute the center by averaging the x and y coordinates and finding the dot closest to that position.Given the clusters and their respective centers. the new behavior now plans a path from the current position to the center of the closest cluster except for the one that pacman might already be in and commits to it until one of the cluster dots has been reached.
There is one relatively unrelated commit in here:
c27930e Fixes a bug that caused a crash when pacman entered the tunnel while chasing or avoiding a ghost. It was a lot easier to fix on this branch because the
positionConsideringTunnelfunction has been moved out of the A* class.