0 5000 10000 15000 20000 17743 9662 5322 9745 5127 3432 ...

3 downloads 29 Views 600KB Size Report
Design and Development of a Software Controller for the Ms. Pac-Man Game. Objective/Background. Objective/Background. Objective/Background. Objective/  ...
Design and Development of a Software Controller for the Ms. Pac-Man Game Student: Wo Yuling

Programme: BEngCE BIM BEngCEBIM

Supervisor: Dr. Sung Albert C W

Objective/Background •Artificial Intelligence (AI) in Video Games •Ms.Pac-man vs Ghosts League (an international competition) •Design intelligent program for Ms Pac-man •Simulate real-world problems with simpler rules Very Dangerous State: •Out of power pill. •Main concern is survival. •Use find-safe-path function.

No

Methodology • Route-finding strategy • State Classifications • Java Implementations • Performance Evaluation

Dangerous State: •Still has power pill •Temporary run away •Find neighbour that farthest from nearest Ghost •Adopt run-away function

Power pill still available? (or Ghost in edible state)

Yes

Yes

Edible Ghost State •Pac-man chase ghost •Adopt Eat-Ghost function •May be interrupted •Priority lower than Dangerous

Ghost within alert distance? Yes No

Edible ghost available? No

Yes

Ambush State •Wait at the neighbour of power pill •Until Ghost get close •Eat pill after ghost get close enough

Close to a power pill? Safe state •None of the above is fulfilled •Eat the nearest pill •Least priority •Rarely happen

No

Results • A ‘smart smart’’ pac-man called “MyPacman MyPacman”” • Outperforms two benchmarks (against 3 ghost controllers) • AI techniques can be deployed to real life 20000

17743

15000 10000

9745 5419

5000 0 MyPacman findSafePath NearestPillPacman

9662 5322

5127 2246

3432 4172

Random Ghost

Aggresive ghost

Legacy ghost

17743 9745 5419

9662 5127 2246

5322 3432 4172

MyPacman findSafePath NearestPillPacman