Comparison of Obstacle Avoidance Behaviors for a Humanoid Robot in Real and Simulated Environments

RoboCup soccer is still host to frequent collisions between robots that lead to unsavory gameplan that is typically not allowed in human soccer. In the Standard Platform league where ultrasonic sensors are employed, these collisions are forbidden, but there is still leniency in the Humanoid League. Additionally, advanced behaviors, such as dribbling, require robust obstacle avoidance strategies to keep possession of the ball. Challenges in the RoboCup competition seek to demonstrate ways to dribble around static obstacles, but it can be hard to replicate conditions for the sake of iteratively tuning robot behavior. We present a method for evaluating policies in simulation and on the real robot for obstacle avoidance behaviors.