In this paper, a trajectory optimization algorithm for multiple unmanned aerial vehicles (UAVs) is investigated for free space optic (FSO) based wireless communication networks, which is composed of multiple UAVs and multiple ground terminals (GTs). Considering the FSO channel model and line of sight (LoS) probability, the altitude of the UAVs for FSO connection between each GT and UAV is decided by the predetermined radius of the communication area of the GT. A multi-UAV trajectory optimization (MUTO) scheme is proposed to maximize the service time. In the MUTO scheme, first, the network is divided into multiple sectors using graph partitioning, and assigned to each UAV, and then the traveling salesman problem (TSP) is used to determine the order of the GTs that the UAV passes through. The number of UAVs that maximizes the time for the UAV to stay within the communication area of the GT is derived. The UAV trajectories are determined by sequentially minimizing the energy consumed when moving between GTs and maximizing the service time by applying successive convex approximation. In addition, to assist operations that use multiple UAVs, two collision avoidance schemes are applied to the MUTO scheme. First is the additional constraints (AC) method, which uses a geographic formula to control the distance between the UAVs, and second is the initial delay (ID) method, which deliberately adds a delay to the UAV trajectories to provide sufficient UAV spacing. The simulation results show that under low density operation conditions, MUTO-ID and MUTO-AC are sufficient to avoid all UAV collisions, and in overly dense UAV operations with very close UAV trajectories, MUTO-ID can reduce collisions by approximately 80%, MUTO-AC can reduce collisions by approximately 85%, and MUTO-ACID can reduce collisions by approximately 95%, when compared to the MUTO scheme with no collision avoidance support applied.

With the wide spread of various smart devices and the proliferation of internet of things (IoT) sensors, the amount of traffic on mobile networks is rapidly increasing, and applications with extreme requirements are increasing. Network function virtualization (NFV) and mobile edge computing (MEC) are emerging as core technologies to satisfy users' real-time service demands. Adapting NFV technology to MECs allows the ability to assign cloud-computing capabilities near the base stations (BSs) of radio access networks (RANs), resulting in extremely fast service access to user equipment (UE). However, placement of virtualized network functions (VNF) within the edge network need to consider the location and the requirements of the user which change in real-time. There has been almost no consideration in the existing research on VNF resource allocation (VNF-RA) based on these aspects. Therefore, in this paper, a VNF resource allocation scheme based on context-aware grouping (VNF-RACAG) technology is proposed that enables groups (based on the geographic context of users, such as location and velocity) to compute the optimal number of clusters to minimize the end-to-end delay of network services. Then, a graph partitioning algorithm is used to minimize user movement between clusters, optimizing the data rate that users lose due to VNF migration.

Network slicing is one of the key technologies for 5G mobile communications. Depending on the requirements of network services, the traffic flow to the core cloud may be a problem. Therefore, network slicing in mobile edge computing (MEC) which cloud computing functionalities can be allocated on the base station near the users becomes important. By adapting network function virtualization (NFV) technology to the MEC network, this paper provides a service function chaining (SFC) slicing scheme that uses the popularity of network services to decide the number of replicas of network services to minimize service time.