ROBUST FIXED-TIME CONTROL BASED ON REINFORCEMENT LEARNING FOR QUADROTOR SYSTEMS UNDER LUMPED UNCERTAINTIES