Real-Time Visual Robot Detection and Modeling with Situational Awareness
Visual object recognition and world state modeling is a challenging problem for a mobile robot. This problem is even more difficult on the AIBO robot platform, which must rely only on local sensor information retrieved from a single monocular camera with limited viewing angle. In the RoboCup domain, specifically the 4-legged league, teams of AIBO robots play soccer against one another. Our research focuses on the visual detection and motion modeling of teammate and opponent robots on the playing field in order to build a more accurate model of the game state. Modeling teammate and opponent robots allows for the creation and improvement of a wide range of game behaviors, which our research also focuses on. Some example behaviors include path planning and obstacle avoidance, passing between teammates, teammate coordination, improved defensive and offensive strategy and positioning, avoidance of penalties caused by pushing robots, and searching for and retrieving the ball.