Distributed-force-feedback-based reflex with online learning for adaptive quadruped motor control