Upload New File

CIO_Decision.m

+function action = DDPG_CIO_Decision(k,cn,c_req)
+num_episodes = 100;
+gamma = 0.99;
+actor_lr = 0.001;
+critic_lr = 0.01;
+CIO_min = -6;
+CIO_max = 6;
+actornet = create_actor_network(k);
+criticnet = create_critic_network(k);
+obsInfo = rlNumericSpec([k+1 1]);
+actInfo = rlNumericSpec([k 1],'LowerLimit', 0, 'UpperLimit', 1); 
+actor = rlContinuousDeterministicActor(actornet,obsInfo,actInfo);
+critic = rlQValueFunction(criticnet,obsInfo,actInfo,...
+    ObservationInputNames="netOin", ...
+    ActionInputNames="netAin");
+agentOptions = rlDDPGAgentOptions(...
+    'TargetSmoothFactor', 1e-3, ...
+    'SampleTime', 0.01, ...
+    'ExperienceBufferLength', 1e6, ...
+    'DiscountFactor', 0.99, ...
+    'MiniBatchSize', 64, ...
+    'NumStepsToLookAhead', 1);
+agent = rlDDPGAgent(actor, critic, agentOptions);
+trainOpts = rlTrainingOptions(...
+    MaxEpisodes=100,...
+    MaxStepsPerEpisode=100,...
+    StopTrainingCriteria="AverageReward",...
+    StopTrainingValue=2000);
+replay_buffer = [];
+state = initialize_state(k+1,CIO_min,CIO_max);
+for episode = 1:num_episodes
+        action = select_action(actornet, state, CIO_min, CIO_max);
+        [next_state, reward] = perform_action_prop(state, action,cn,c_req,CIO_min,CIO_max);
+        experience = struct('state', state, 'action', action, 'reward', reward, 'next_state', next_state);
+        replay_buffer = [replay_buffer, experience];
+        criticnet = update_critic(actornet, criticnet, replay_buffer, critic_lr, gamma,CIO_min,CIO_max);
+        actornet = update_actor(actornet, criticnet, replay_buffer, actor_lr, gamma, CIO_min,CIO_max);
+        state = next_state;
+end
+end
\ No newline at end of file