Abstract:
Over the past few years, with the rapid increase in the number of natural disasters, the need to
provide smart emergency wireless communication services to collect victim`s information
has become crucial. Unmanned aerial Vehicles (UAVs) have gained much attention as
promising candidates due to their unprecedented capabilities and broad flexibility. However,
the limited UAVs battery and the limited UE`s energy, due to the breakdown of ground
power system after disasters affected the mission, which necessitates an energy-efficient
emergency wireless access procedure to insure the network lifetime. In this thesis, we
propose multi-UAVs assisted emergency wireless access to collect disaster time information,
where the UAVs fly safely to the victim`s UEs and control the UEs transmit power during the
communication access. Oure goal is minimize the total energy consumption of both the
UAVs and the UEs, which is needed to accomplish the emergency wireless access mission.
We formulate this problem into three sub-problems of single UAV navigation, UE power
control, and multi-UAV scheduling, and model each part as a finite-horizon Markov Decision
Process (MDP). We deploy deep reinforcement learning (RL)-based frameworks to solve
each part. Specifically, we use the deep deterministic policy gradient (DDPG) method to
control the UE’s transmit power during wireless access and to generate the best trajectory for
the UAVs in an obstacle-constrained environment to reach the UEs. To schedule activity
plans for each UAVs to visit the UEs, we propose a multi-agent deep Q-learning (DQL)
approach by taking the energy consumption of the UAVs on each path into account. Our
simulations show that the UAVs can find a safe and optimal path for each of their trips.
Continuous power control of the UE achieves better performance than the fixed power and
fixed rate approaches in terms of the UE’s energy consumption and the wireless access
completion time. In addition, compared to the two commonly used baselines framework, our
multi-UAVs scheduling framework achieves better results in the simulated scenario.
