No. You could do this with any model that has feedback built in. I don't think you'd even need something as complicated as Q-learning but any type of reinforcement "learning" method would work. You're basically teaching a PID that has a clock.