A team of engineers was recently successful in developing an artificial Ms. Pac Man player making use of a completely new approach towards computation of real-time gaming strategy. This new artificial character has been successful in attaining the highest score with respect to a computerised game. It happens to be one of the most popular arcade games, Ms. Pac-Mac need to keep away all ghost enemies while she is collecting different items and makes her way through obstacle-populated maze. It is one of the favorites among computer scientists and engineers who have been competing for quite some time to program the best mechanized player.
The best score at yearly Ms. Pac-Man Screen Capture Competition has been at 36, 280. However, a trio of researchers that included a mechanical and aerospace engineering professor from Cornell, Silvia Ferrari, was successful in generating lab score of 43, 720. The high score could be achieved with the use of decision-tree approach under which artificial player moves are extracted from a complete maze of dynamic equations and geometry that finds out the ghosts movements with a high accuracy of 94.6 percent. As the game moves ahead. The decision tree keeps getting an upgrade in real-time.
Ferrari says, “The novelty of our method is in how the decision tree is generated, combining both geometric elements of the maze with information-gathering objectives.” She noticed that any kind of information in this respect pertains to fruits collected by Ms. Pac-Man for bonus points. Her team is one of the first ones to model a game’s components mathematically. All previous artificial players till date were developed only with the help of model-free methods which didn’t prove to be that efficient. This new approach is great and will definitely play a significant role in future developments.
Filed Under: News