The invention discloses a joint optimization method and
system for a two-layer UAV-MEC network, and the method comprises the steps: collecting a historical request, and predicting a content request, and obtaining the popularity of the content; on a long-time scale, an upper-layer unmanned aerial vehicle selects a cache and a plan according to the
content popularity, and puts the cache and the plan into an experience
pool; in a short time scale, the lower-layer unmanned aerial vehicle performs local cache adjustment and unloading selection according to the
content popularity, and puts the content into an experience
pool; cVaR is introduced, time
delay and
energy consumption are combined, rewards are formed, and a unified target is constructed; and according to the experience
pool and the unified target, carrying out centralized training and updating a multi-agent strategy, and issuing the multi-agent strategy to each unmanned aerial vehicle for execution. The
system comprises an upper-layer unmanned aerial vehicle serving as an air small
cloud server and a lower-layer unmanned aerial vehicle serving as an air
base station. According to the method, the average time
delay, the
total energy consumption and the
tail time
delay risk of the
system are minimized through collaborative
decision making of two time scales. The method can be widely applied to the field of
mobile edge computing.