We attempt to have a quadrotor autonomously avoid people while moving through an indoor environment. Our algorithm for detecting people uses a Haar classifier to detect a person's upper body in the image feed from the front camera of the quadrotor. We then localize the person in a 5 x 5 grid based on the location and size of his detected feature. We use this grid to form a Markov decision process (MDP) problem and use reinforcement learning to determine the optimal control policy for the quadrotor so as to avoid colliding with the person.
Link to this comment:
All Comments (0)