Disclosed is a method for using deep
reinforcement learning to optimize dependent task offloading in
the Internet of Vehicles. The steps of the present invention comprise: S1, constructing a vehicle
system network; S2, constructing a task model for an application; S3, constructing a task
load model: on the basis of the vehicle network
system and the task model, calculating time
delay,
power consumption, and paid incentive rewards for each of the following three offloading
modes: local computing, offloading to a nearby vehicle, and offloading to a nearby roadside unit; S4, determining task priorities: first, on the basis of the allocation of predecessor nodes of sub-tasks in combination with a dynamic network environment, determining the priority of each sub-task, and then carrying out scheduling on the basis of a multi-
queue algorithm; and S5, searching for an optimal offloading strategy by means of deep
reinforcement learning. The method of the present invention does not require excessive prior knowledge, has good
reusability in similar application scenarios, and has high practical value.