sonicrs / rotem_ants Goto Github PK

Input
Configuration space 
split into 10 different areas (graph nodes).
Nodes are connected by edges based on adjacency. 
Not all edges are traversable.
We find similarities between edges.

Similarities
bg(x) = gh(v)
ad(v) = fi(x) = ei(x)
df(v) = ij(v) = de(x)
fg(v) = cb(v)
ac(v) = jh(v)

start = c
end = j, h
Edges carry an initial cost (extract these from experimental data)

Algorithm:
While not at goal:
Find lowest-cost path, and attempt to traverse first edge. 
Edge traversal: local algorithm, that succeeds or doesn’t. 
In case of success: Decrease the cost of this edge, and its similar edges. Walk to the respective node. 
In case of failure: Increase the cost of this edge, and its similar edges. Remain in your node. 

NOTES: 
Randomness of initial cost, or randomness per long term decision.