MACHINE CONTROL BASED ON AUTOMATED LEARNING OF SUBORDINATE CONTROL SKILLS

摘要

For controlling a machine (100) according to multiple control objectives, a device (200) provides multiple subordinate control skills (221, 222, 223, 224, 225), each of the control skills being assigned to a different one of the multiple control objectives. Further, the device (200) provides multiple learning processes (211, 212, 213, 214, 215), in particular reinforcement learning (RL) processes. Each of the learning processes (211, 212, 213, 214, 215) is assigned to a different one of the multiple control objectives and being configured to optimize the corresponding subordinate control skill (221, 222, 223, 224, 225) based on input data received from the machine (100). Further, the device (200) is configured to determine a superordinate control skill (240) based on the subordinate control skills (221, 222, 223, 224, 225) and to control the machine (100) based on the superordinate control skill.