Performance Analysis Of UAV Enabled Wireless Communication In 5G And Beyond Networks

Unmanned Aerial Vehicles(UAVs)are small automated aircraft,which are currently a significant element of the mobile network due to their expanding use in a variety of applications.The availability of bandwidth,cost-efficiency,and adaptability allows integrating Fifth Generation(5G)and UAV technologies which is a potential solution for enabling smooth communication in applications like rescue operations,and network reconstruction during emergencies.UAVs can also be utilized as airborne base stations,edge servers,and relay nodes in mobile networks to guarantee that communication services are provided in the intended area.UAVs offer the advantages of mobility and Line of Sight(LoS).To collect data from ground nodes,UAVs can be set up as Access Points(APs).The 5G network has also seen the emergence of Multi-Access Edge Computing(MEC)technology,which aims to improve the Quality of Experience(QoE)for users with varying Quality of Service(QoS)demands.Data collection and service provided by UAVs in an emergency or a MEC environment encounter a variety of design challenges that need to be addressed,such as optimal positioning,path and network planning,obstacle detection,collision avoidance,and recognition of specific events in an emergency.Due to the lack of dynamic control and the external influence of weather effects,the overall performance,such as QoS,reliability,and energy efficiency,is affected.This research provides an approach named Dynamic Positioning and Energy Efficient Path Planning(Dynamic-UAV)to enhance the performance metrics through weather prediction,event recognition,clustering,positioning,and path planning in a disaster-effected 5G-assisted multiUAVs environment.Dynamic-UAV is a collective approach consisting of four different algorithms to improve the network performance in terms of QoS,reliability,and energy efficiency.For UAVs in MEC,path planning based on QoS requirements of Ground Users(GUs)is performed by using Deep Reinforcement Learning(DRL).The significant contributions of this dissertation are as follows:1)In the Dynamic-UAV approach,Light Weight Gated Recurrent Unit(LGRU)and Densitybased OPTICS Clustering(DBOC)are designed to maximize the throughput and PDR and minimize the delay and energy consumption caused by the lack of weather prediction,event recognition,and clustering of IoT devices in a weather effected scenario in a 5G-assisted multiUAVs environment.LGRU classifies the scenario as disaster and non-disaster zones and understands the environmental factors.LGRU considers different weather parameters as historical information and predicts the weather effects.LGRU offers more advantages over existing classification algorithms because it has a gating mechanism for selective updating,is faster to train,and more memory-efficient,and is robust to noisy inputs.DBOC is a method employed for clustering IoT devices in regions detected by LGRU to improve communication throughput.DBOC has advantages over k-means and other existing clustering algorithms in UAV enable networks because it does not require specifying the number of clusters,can handle non-linear shapes,allows for hierarchical clustering,and is optimal in terms of energy and communication.Simulation results demonstrate the effectiveness of these techniques in achieving an average throughput of 1.59 bit/s,PDR of 0.88,delay of 18.5 ms,and average energy consumption of 7.84 kJ.2)The Dynamic-UAV approach is further enhanced by performing positioning and path planning to address the issues of the determination of the exact number of UAVs for a mission,coverage,and energy efficiency.By using the Dynamic Positioning-based Soft Actor-Critic(DPS AC)technique,the number of UAVs needed for a given mission is calculated and then placed in an appropriate position to enhance the system performance.DPSAC is more efficient than existing learning techniques for optimizing UAV positioning in wireless networks due to its ability to handle continuous action and state spaces,adapt to changing environments,and handle highdimensional state spaces.The Shuffled Shepherd Optimization with Dynamic-Window Method(SSO-DWM)is employed to avoid obstacle detection and execute path planning for UAVs by considering multiple parameters.SSO-DWM efficiently handles the dynamic of the environment and can be adopted to changing environments with multiple objectives compare to the existing path planning techniques which have a single objective and only focus on the shortest path.The proposed method is compared to existing methods for validation in terms of QoS,reliability,and energy efficiency.The findings show that the Dynamic-UAV obtained the highest possible coverage probability of 0.82,the maximum possible number of collected packets(7109-5875),the maximum possible energy efficiency of 1.544,and the least possible delay of 16.4 ms.3)In the UAV-enabled MEC network,the issues of trajectory planning and mobility management are handled,considering the QoS needs of the GUs.A DRL-based DQN model has been created for optimal trajectory planning to avoid collision between UAVs and meet QoS requirements from GUs in an energy-efficient manner.Therefore,this dissertation also proposes a well-designed Artificial intelligence(AI)and MEC-enabled framework for a UAV-enabled future network.It uses a powerful Deep Reinforcement Learning(DRL)algorithm to plan and optimize trajectories in real-time.Moreover,it provides QoS-aware service provisioning.DRL-based DQNs offer several advantages over tabular Q-learning in UAV path planning.These advantages include efficient handling of large state and action spaces,the ability to directly learn the complexity of the environment,and the ability to yield better rewards in terms of energy-efficient paths,as well as ensuring QoS in service provisioning.The findings show that the DRL technique for energyefficient and QoS-aware trajectory planning is superior to the baseline models because it achieved almost 99%QoS fulfillment for ten GUs.