Petri net modeling of perceptual decision making algorithms for better ball control and navigation

Abstract

RoboCup poses the unique challenge of programming a robot to autonomously navigate an environment with both dynamic and static objects. Robot teams must coordinate with one another to pass and kick the ball into the goal, all while avoiding collisions with opponent robots and staying within the bounds. The two algorithms proposed allow a robot to navigate a soccer field with dynamic objects and pass a soccer ball to teammates. High-level Petri nets, a graphical and mathematical modeling tool, are used to model and simulate these algorithms. Because of the complexity of soccer gameplay, regular Petri nets do not always have the capability to model these algorithms well. The Stochastic Petri net (SPN) incorporates a factor of randomness and can model a memoryless algorithm due to its reachability graph being isomorphic to a continuous time Markov chain. The Colored Petri net (CPN) includes programming functionalities that allow for the inclusion of discrete logic in modeling. These models were used to simulate algorithms proposed to improve the gameplay of the NAO robots.