Unmanned aerial vehicle (UAV) technology has gained a great interest in communication systems due to its ability to host a cellular base station (BS) and thus act as an aerial BS (UAV-BS). The inheritance of mobility in the airspace makes the deployment of UAV-BSs flexible and agile aiming to mainly complement the terrestrial network, extend its coverage, and serve as a capacity injector in high-throughput demand scenarios. Besides, a UAV can also act as an aerial user (UAV-UE) for various use cases, such as aerial data collection and cargo delivery. Such UAV-UE missions need reliable cellular communication links in order to safely operate in beyond visual line-of-sight (BVLoS). Since terrestrial networks were not primarily designed to serve aerial users, due to their down-tilted BS antennas, re-coursing solely to these networks for aerial users’ cellular connectivity might not be a viable approach as a long-term solution. Alternatively, deploying UAV-BSs in this context can substantially improve both aerial and terrestrial users coverage and capacity. One of the challenging issues is how to characterize the UAV-UE performance in integrated aerial/terrestrial networks, called vertical heterogeneous networks (VHetNets). First, we thoroughly study the aerial user’s performance in terms of coverage probability in a VHetNets setup. Under a more realistic system model, we revisit the coverage and throughput performances of an aerial user in VHetNets, considering LoS and non-LoS (NLoS) transmissions and under different spectrum sharing policies among separate aerial and terrestrial networks. Some insights have been concluded on the integration of aerial BSs and UAV-UEs with the existing terrestrial network. Specifically, optimal positioning of UAV-BSs for maximized aerial users coverage was investigated for various aerial users distribution and spectrum allocations. Moreover, visioning that UAV technology will revolutionize the cargo delivery industry, we proposed a new concept of 3D aerial highways, which designs coordinated routes for a massive number of UAVs used mainly for delivery purposes. In this context, multiple network technologies were proposed and discussed to guarantee the cellular connectivity of cargo-UAVs in 3D aerial highways. For the particular case of connectivity supported by terrestrial cellular networks, an optimal energy-efficient and low-handoff trajectory planning for a cargo-UAV mission was proposed, with respect to disconnectivity constraints. Consequently, associated design guidelines and recommendations have been drawn. Leveraging reinforcement learning (RL) tools, we proposed a novel algorithm for path planning and cell association for the cargo-UAV that maximizes its cellular service reliability and minimizes the handoff events. Finally, we introduced a new paradigm, intermittently tethered UAV (iTUAV), as a trade-off between mobility and energy availability for providing cellular connectivity in temporary events.