Intelligent agent deep reinforcement learning method based on partner network

The invention discloses an agent deep reinforcement learning method based on a partner network, and the method comprises the steps: constructing guidance signal functions in different reward environments according to the control state of an agent, generating an experience tuple needed by the partner network, and storing the experience tuple to experience playback of the partner network; constructing a partner network model by adopting a multi-layer long-short-term memory network, respectively calculating guidance signals corresponding to the intelligent agent according to historical control states and selectable actions, and updating partner network parameters; performing attenuation processing on the guidance signal corresponding to the intelligent agent; and constructing a deep Q network reinforcement learning model based on a multi-baseline model, fusing the attenuated guidance signal and a set decision signal, training the deep Q network reinforcement learning model, generating an experience tuple require