Multi-objective Vision to Redesign Wireless Network Architectures

Abstract

Future wireless systems are expected to provide seamless, reliable, and wide-area connectivity across environments where conventional terrestrial architectures are no longer sufficient. Existing networks are largely built on two limiting assumptions that communication infrastructure remains confined to the ground, and that the wireless propagation environment is an uncontrollable medium. As connectivity demands expand across remote, obstructed, and heterogeneous scenarios, these assumptions become structural bottlenecks. This thesis argues that meeting such demands requires not incremental improvement, but a fundamental redesign of wireless network architecture through enabling technologies that expand both infrastructure placement and propagation control. Within this redesign, NTN and RIS emerge as two complementary architectural enablers. The former extends communication infrastructure into the non-terrestrial domain, while the latter transforms the radio environment into a programmable system component. Their integration creates a new architectural paradigm, i.e., non-terrestrial RIS, in which network geometry and signal propagation can be jointly controlled. Among the possible realizations of non-terrestrial RIS, RIS-enabled HAPS, so-called HAPS-RIS, offers a particularly compelling balance of wide coverage, high probability of LoS connectivity, lower latency than satellite-based solutions, greater endurance and payload capacity than UAV-based alternatives, and more flexible deployment than terrestrial installations. Integrating such enabling technologies does not simplify wireless network design; rather, it expands the architectural design space by introducing new controllable dimensions in mobility, coverage, infrastructure hierarchy, and resource allocation. Consequently, the design problem becomes inherently multi-objective, since these dimensions generate competing objectives that cannot be captured through conventional single-objective formulations. On this basis, the thesis proposes a multi-objective vision for redesigning wireless network architectures, using HAPS-RIS as a representative architectural platform. The multi-objective vision is developed through three main technical contributions. (i) An aerodynamic HAPS-RIS architecture is introduced for connecting unconnected ground stations under time-varying channels. By geometrically modeling the predictable mobility of the platform, a multi-objective framework is established to jointly maximize cascaded channel gain, eliminate Doppler spread, and control delay spread, leading to a closed-form Pareto-optimal RIS phase shift design. (ii) An aerostatic HAPS-RIS assisted UAV network architecture is proposed to address the scarcity of UAV infrastructure in remote coverage scenarios. A hierarchical bi-level framework is proposed where the leader problem maximizes the number of users covered by HAPS-RIS while the follower problem minimizes the number of UAVs, subject to QoS and feasibility constraints. The hierarchical optimization problem is solved via the proposed DyRaZARC and KADUS techniques. The results reveal that passive HAPS-RIS capability can substitute for active UAV deployment, exposing a direct trade-off between RIS resources and the number of required UAVs. (iii) A unified joint multi-objective framework is proposed for fully integrated HAPS-RIS and UAV networks to maximize the covered users by HAPS-RIS, minimize the required number of UAVs, and minimize the total average UAV path loss, simultaneously, subject to QoS and feasibility constraints. A central technical achievement lies in transforming an intractable large-scale optimization problem into a scalable and implementable solution. The deployment of UAVs is shown to admit an equivalence to K-means clustering, which is proved via the proposed theoretical link between classical optimization and unsupervised learning. RIS phase shifts admit a practical closed-form design, and high-dimensional combinatorial association variables collapse into intuitive geometric and bandwidth-level control parameters, proved via the proposed mapping technique. Finally, the transformed optimization problem is solved via the proposed dynamic Pareto optimization framework which exposes the intrinsic priority structure of integrated aerial networks. The simulations uncover clear regime-dependent behaviors, indispensable trades-offs, benchmarking, and practical design guidelines. HAPS-RIS alone sufficed for low-rate services, whereas UAV assistance becomes progressively necessary as rate requirements increase. Remarkably, conventional architectures naturally emerges as special and degenerate cases of the unified framework, through a single bandwidth-portioning factor. Taken together, the strongest thesis-level insight is therefore that redesigning wireless network architecture via enabling technologies requires a multi-objective vision of managing trade-offs within an expanded design space.