Making Computers Play Angry Birds: Two Students Take their Shot


 It sounds quite simple to teach a person how to slingshot digital birds at a pile of evil pigs but when it comes to teaching a computer program to play this video game of angry birds independently, undergraduates Anjali Narayan-Chen, a UW-Madison computer sciences and computer engineering undergraduate and classmate Liqi Xu achieved the feat in August, ranking this single team from United States, third in the recent International Joint Conference on Artificial Intelligence held in Beijing.

The duo did this by playing the game over many times, programming machine learning to let computer take an independent run at the popular video game after categorizing the shots as good or bad and feeding it with lots of examples. About a dozen computers were made to play 724,993 games involving 3,986,260 birds heaved at pigs and after doing away with a deep pool of uncertain outcomes, they fed their software 224,916 “good” shots 168,549 “bad” shots. On each level, the software zooms out to take picture of screen, and overlays the image with a grid, after which it decides what each cell in the grid contains, it chooses a shot then based on its knowledge and perception of all the good and bad shots, relying on its deep library of good targets versus bad.


Leave a Reply