In the event of a sudden network failure in emergency scenarios

the network can be recovered by deploying unmanned aerial vehicles(UAVs) as temporary base stations and edge computing nodes.A single UAV exhibits significant limitations in the processing capacity and coverage range

while multi-UAV collaboration can effectively solve such problems.Aiming at the problem of UAV collaborative computing offloading in emergency scenarios

an air-ground integration architecture is proposed

and a multi-agent deep reinforcement learning(MADRL)algorithm is innovatively designed to jointly optimize offloading decisions

as well as the flight angle

and speed of UAVs

so as to reduce system offloading delay and energy consumption while considering load fairness.The simulation results show that compared with random processing

the proposed algorithm reduces the total system cost by about 77%

and compared with the deep deterministic policy gradient(DDPG) algorithm

it reduces the total system cost by about 46%

which fully demonstrates the superiority of the proposed offloading strategy in extreme scenarios.