Deep deterministic policy gradient actor-critic algorithm with supervision controller

[Emercy44]

The deep deterministic policy gradient actor-critic algorithm with supervision controller is proposed. The latent(indirect) supervision can guarantee the stable convergence of the control actor network. The conventional controller doesn’t need excellent control performance. The main role of the conventional controller is to supply helpful control experience for the actor network. Then the actor network is trained by those experience.
Deep deterministic policy gradient actor-critic algorithm with supervision controller.docx