Joint Optimization of Resources in Fog-Radio Access Network with Binary Computation Offloading
Algorithm 2
The DDQN algorithm.
Input: Number of UEs , size of tasks and distance between UE and F-AP ;
Output: Suboptimal decision ;
(1)
Initialization: Initialize the parameter of the main network with random weight and the parameter of the target network with random weight and empty the public experience pool;
(2)
Set training interval ;
(3)
for to do
(4)
Reset starting environment information ;
(5)
for to do
(6)
Reset remaining channel resources and remaining fog computation resources ;
(7)
for to do
(8)
The main network generates action with the -greedy method according to ;
(9)
Map to subactions and implement them in the environment;
(10)
Obtain status and reward based on the environmental changing ();
(11)
Mark and store it in the public experience pool;
(12)
if (the public experience pool is full) then
(13)
The target network calculates ;
(14)
Update the main network parameter based on and replace the oldest data with the new one;
(15)
end if
(16)
while (episode mod )do
(17)
Assign the main network parameters to the target network, i.e., ;
(18)
end while
(19)
end for
(20)
end for
(21)
end for
(22)
The target network picks the action with the smallest ) as the suboptimal decision :.
