An Intelligent Offloading System Based on Multiagent Reinforcement Learning

Algorithm 1

Loser Kuhn–Munkres Algorithm.

	Input: calculate all vehicle-RSU pair matches using KM algorithm and the weight of vehicles pairing with RSUs’ weight; record the vehicle nodes that are not matched as the loser set.
Output: updated pairs match
(1)	Collect matched orders
(2)	Initialize a directory challenge, arrays attacker, and challenger
(3)	Travel across all the weights between Loser i and RSU j:
(4)	If weight [Loser [i]] [j] > weight [match [j]] [j], then
(5)	Challenge [Loser [i]] [j] = weight [Loser [i]] [j]]−weight [match [j]] [j]
(6)	If Loser [i] is not in attacker, then
(7)	Attacker push (Loser [i])
(8)	End if
(9)	If j is not in attacker, then
(10)	Challenger push (Loser [i])
(11)	End if
(12)	End if
(13)	Initialize a two-dimension array chaWeight
(14)	Check if challenger j is in the challenge list of attacker i:
(15)	If challenger [j] is not in Challenge [attacker [i]], then
(16)	Challenge [attacker [i]] [challenger [j]] = 0
(17)	End if
(18)	chaWeight [i][j] = Challenge [attacker [i]] [challenger [j]]
(19)	Calculate attacker-challenger pairs chaMatch with chaWeight by using KM algorithm
(20)	For i = 1:challenger, do
(21)	Match [challenger [i]] = attacker [chaMatch [i]]
(22)	End for