1.) Mapping Agent(cube) position into states
WWWWWWWWWWWWOOOOOOOYOOOYOOOYOOOYGGGGGGGGGGGGGGGGWRRRWRRRWRRRWRRRBBBBBBBBBBBBBBBBRRRRYYYYYYYYYYYY frontCounterClockwise(2) 100 WWWWWWWWWWWWWWWWOOOOOOOOOOOOOOOOGGGGGGGGGGGGGGGGRRRRRRRRRRRRRRRRBBBBBBBBBBBBBBBBYYYYYYYYYYYYYYYY WWWWWWWWWWWWRRRROOOWOOOWOOOWOOOWGGGGGGGGGGGGGGGGYRRRYRRRYRRRYRRRBBBBBBBBBBBBBBBBOOOOYYYYYYYYYYYY frontClockwise(1) 100 WWWWWWWWWWWWWWWWOOOOOOOOOOOOOOOOGGGGGGGGGGGGGGGGRRRRRRRRRRRRRRRRBBBBBBBBBBBBBBBBYYYYYYYYYYYYYYYY RRRRWWWWWWWWWWWWWOOOWOOOWOOOWOOOGGGGGGGGGGGGGGGGRRRYRRRYRRRYRRRYBBBBBBBBBBBBBBBBYYYYYYYYYYYYOOOO backCounterClockwise(4) 100 WWWWWWWWWWWWWWWWOOOOOOOOOOOOOOOOGGGGGGGGGGGGGGGGRRRRRRRRRRRRRRRRBBBBBBBBBBBBBBBBYYYYYYYYYYYYYYYY OOOOWWWWWWWWWWWWYOOOYOOOYOOOYOOOGGGGGGGGGGGGGGGGRRRWRRRWRRRWRRRWBBBBBBBBBBBBBBBBYYYYYYYYYYYYRRRR backClockwise(3) 100 WWWWWWWWWWWWWWWWOOOOOOOOOOOOOOOOGGGGGGGGGGGGGGGGRRRRRRRRRRRRRRRRBBBBBBBBBBBBBBBBYYYYYYYYYYYYYYYY WGWWWGWWWGWWWGWWOOOOOOOOOOOOOOOOGYGGGYGGGYGGGYGGRRRRRRRRRRRRRRRRBBWBBBWBBBWBBBWBYBYYYBYYYBYYYBYY innerLeftDown(6) 100 WWWWWWWWWWWWWWWWOOOOOOOOOOOOOOOOGGGGGGGGGGGGGGGGRRRRRRRRRRRRRRRRBBBBBBBBBBBBBBBBYYYYYYYYYYYYYYYY
2.) Adding random scenarios as inputs (I can't fit the file in here its too big, that's why finding a clever way to summarize states is really important)
Syntax: state action reward next-state
Results (virtually it can solve the problem, you just need a huge amount of storage):
As we observed our agent can solve problems but depends with the data given. next we will apply qlearning in a famous game flappy bird.
Lets use q-learning for flappy bird game