Towards a Computer Vision Based Approach for Developing Algorithms for Soccer Playing Robots
This chapter discusses multiple linear regression and multilayer perceptron models will be used to predict the distance from the robot to the soccer ball based on a set of engineered features obtained from the image containing the soccer ball. The fields of robotics and machine learning intersect in the pursuit of creating artificial intelligence that can dictate the behavior of a robot autonomously. Using Petri nets to model soccer playing robots is of interest due to RoboCup and the initiative of one day have a team of robots that can compete against humans. The two primary objectives for the robot are to be able to recognize both other robots and the soccer ball as well as predict the distance to the soccer ball based on the image of the soccer ball. Object detection and distance prediction are two separate machine learning problems each requiring their own models to be trained. The TensorFlow object detection.