Parameters	Meaning
n	Entire system time steps
nd	Number of connected Internet of Things devices to the fog node in each fragment
nb	Number of neighboring fragments
nQ	Learning time steps
ɛ	Probability of random selection
α	Learning coefficient
γ	Constant discount factor
p	Predefined device priorities
zm	Determines the situation of the fragment
zn	Average of zm of all neighbors
Thl	Threshold length
Thold	Last threshold level
Thnew	New selected threshold level
Threshold	Contains Thnew based on zn, zm, and Thold and is continuously updated while learning the best threshold level
NTh	Number of Threshold elements
Thc	Correct threshold levels, based on the internal and external situations, as the target threshold matrix
RTh	Received reward of learning the best threshold level through reinforced learning
QTh	Represents the Q-value of the selected threshold level
punish	Number of feasible helpers with lower priority than the emergency device priority
countpun	Total feasible helpers with lower priority than the selected helpers for all needy devices in time step n
penalty	Number of all devices with lower priority than the current needy device priority
blame	Punishment when choosing a neighboring device when feasible devices have lower priority than the threshold
flag	flag(i) = 0 indicates that device i is in a normal situation, and flag(i) = 1 indicates an emergency
track	track(i) = 1 indicates that device i has received extra bandwidth
fnort	fnort(i) = 0 indicates that needy device i has not received extra bandwidth
a	Selected action (helper)
RnQ	Received reward for the fog node
Qnew	Represents the Q-value of the selected device k for helping device i
SBM	Successful bandwidth management