A Network Fault Prediction-Based Service Migration Approach for Unstable Mobile Edge Environment

Algorithm 1

: Service migration based on network fault prediction and DQN.

Input:
State set S, Action set A, discounting factor , explore probability
Output:
Migration strategy .
1:	Initialize the Experience Pool with a capacity of M
2:	Initialize the evaluation network neuron weight vector
3:	Initialize the target network neuron weight vector ,the rest of the parameters are the same as the evaluation network
4:	for episode = 1, 2…do
5:	Initialize user location and the location of edge server ,initialize the first state
6:	fort = 1, 2…do
7:	Predict the faulty node f and add it to the set of faulty nodes
8:	Randomly choose action with probability
9:	Or choose the action
10:	perform action ,calculate the penalty value p, reward value and the next moment state
11:	Put the sample into the experience pool
12:	Randomly select a small batch of samples from EP
13:	if if episode terminates at step t + 1 then
14:	set
15:	else
16:	set
17:	end if
18:	Train the network according to the loss function
19:	Set every x steps
20:	end for
21:	end for