Multiagent Reinforcement Learning for Task Offloading of Space/Aerial-Assisted Edge Computing

Table 1

Notation.

	Number of GUs
	Number of UAVs
	Number of initial offloaded tasks
	Number of time slot needed to finish offloading
	Offloading indication matrix of GUs at time slot
	Computation Capacity of UAVs
	Computation Capacity of Satellite
	Computation Capacity of GUs
	Data size (in bits)
	Computing workload (in CPU cycles per bit)
	Computational cost of all SAGIN components
	Queuing time
	Remained computation tasks at time
	Local execution energy consumption
	Transmission power
	Bandwidth
	Noise power
	Channel gain of GU to satellite
	Path loss of GU to UAV
	Transmission rate of GU to UAV or satellite
	Transmission delay
	Communication-related energy consumption
	Discount factor
	Soft update factor
	Actor online with parameter and input
	Actor target with parameter and input
	Critic online with parameter and
	Critic target with and
	Minibatch size
	Replay buffer
	Action random noise