Perception and Strategy Learning in Robot Soccer

Soccer robots need to reason efficiently and make real-time decisions in an uncertain game. However, with more objects of interest being changed to white and games being pushed towards variable lighting conditions, advanced vision algorithms are crucial for robot’s ability to perceive new surroundings in real-time. Upon accurately perceiving a dynamic game environment, intelligent role switching, team positioning, and path planning methods are also key to winning the game. We explored some novel techniques in lighting-invariant perception and high-level team strategies which allowed our team UPennalizers to stay competitive in the Standard Platform League (SPL).