The surge of interest in sixth-generation (6G) wireless networks is primarily propelled by the emer- gence of data-intensive applications that include virtual, augmented, and mixed reality services, tactile Internet, haptic applications, autonomous systems, and holographic-type communications. This heightened demand is exerting considerable pressure on the existing infrastructure, neces- sitating a departure from traditional ground-based communication paradigms toward innovative frameworks such as integrated terrestrial and non-terrestrial networks (NTNs). These integrated networks encompass terrestrial, aerial, and space layers, thereby augmenting communication ca- pabilities and extending network resources. However, despite the potential advantages in terms of data throughput, coverage, and resilience, the deployment of integrated terrestrial and NTNs introduces novel challenges due to the substantial volume of data and network traffic involved in both inter- and intra-layer communications. What sets 6G networks apart from their predecessors is their resolute commitment to achiev- ing ubiquitous intelligence, wherein native artificial intelligence (AI) assumes a pivotal role in orchestrating wireless networks spanning from the core to the edge and the cloud. Machine learn- ing (ML), a subset of AI, is anticipated to become an indispensable tool in future 6G networks. ML will harness data from all network segments to facilitate intelligent resource management, access control, and multi-layer communications. The vision for 6G networks extends beyond the mere replacement of specific network modules with ML; it envisions each network node en- dowed with intelligence that enables continuous learning from its environment and adaptation to network dynamics. The heterogeneous characteristics and requirements of various nodes within and across layers pose communication management and coordination challenges attributed to the diverse types of data involved. Moreover, conventional AI algorithms require the transmission of raw data generated and stored on local devices to centralized servers for processing. This conventional approach compro- mises user privacy and security while exacerbating network overhead. Furthermore, centralized AI is plagued by prolonged propagation delays, rendering it unsuitable for real-time applications, a concern further magnified within integrated terrestrial and NTNs. In light of the increasing demand for secure AI tools and the enhanced computing and storage capabilities of wireless devices, re- search efforts are shifting from centralized to distributed learning approaches. Deep-reinforcement learning (DRL), a specific variant, is emerging as an enabling technology capable of training wire- less network nodes in a decentralized manner with collaborative processes that do not consume excessive network resources. However, while DRL has garnered considerable attention within the context of wireless networks, its implementation in single and multiple layers within integrated terrestrial and NTNs remains in its nascent stages. Numerous design aspects and challenges as- sociated with inter- and intra-layer communication remain unaddressed, including user selection and scheduling, joint communication and learning, data imbalance, model convergence rate, and resource allocation. In this dissertation, we consider the 6G enablers, i.e., non-terrestrial networks and intelligent networks, with distributed and centralized optimization problems and challenges for both. Firstly satellite systems face a significant challenge in effectively utilizing limited communication re- sources to meet the demands of ground network traffic, characterized by asymmetrical spatial distribution and time-varying characteristics. Moreover, the coverage range and signal transmis- sion distance of low Earth orbit (LEO) satellites are restricted by notable propagation attenuation, molecular absorption, and space losses in sub-terahertz (THz) frequencies. This work introduces a novel approach to maximize LEO satellite coverage by leveraging reconfigurable intelligent sur- faces (RISs) within 6G sub-THz networks. The optimization objectives encompass enhancing the end-to-end data rate, optimizing satellite-remote user equipment (RUE) associations, data packet routing within satellite constellations, RIS phase shift, and ground base station (GBS) transmit power (that is, active beamforming). The formulated joint optimization problem poses significant challenges because of its time-varying environment, nonconvex characteristics, and NP-hard com- plexity. To address these challenges, we propose a block coordinate descent (BCD) algorithm that integrates balanced K-means clustering, multi-agent proximal policy optimization (MAPPO) deep reinforcement learning (DRL), and whale optimization (WOA) techniques. The performance of the proposed approach is demonstrated through comprehensive simulation results, which exhibit its superiority over existing baseline methods in the literature. Secondly, the proliferation of intelligent transportation systems (ITS) has led to increasing demand for diverse network applications is studied. However, conventional terrestrial access net- works (TANs) are inadequate for accommodating various applications for remote ITS nodes, that is, airplanes and ships. On the contrary, satellite access networks (SANs) offer additional sup- port for TANs in terms of coverage flexibility and availability. In this study, we propose a novel approach to ITS data offloading and computation services based on SANs. We use low-Earth or- bit (LEO) and cube satellites (CubeSats) as independent mobile edge computing (MEC) servers that schedule the processing of data generated by ITS nodes. To optimize offloading task selec- tion, computing, and bandwidth resource allocation for different satellite servers, we formulate a joint delay and rental price minimization problem that is mixed-integer non-linear programming (MINLP) and NP-hard. We propose a cooperative multi-agent proximal policy optimization (Co- MAPPO) deep reinforcement learning (DRL) approach with an attention mechanism to deal with intelligent offloading decisions. We also decompose the remaining subproblems into three inde- pendent subproblems for resource allocation and use convex optimization techniques to obtain their optimal closed-form analytical solutions. We perform extensive simulations and compare our proposed approach to baselines, resulting in performance improvements of 9.9%, 5.2%, and 4.2%, respectively. Third, next-generation networks need to meet ubiquitous and high data-rate demand. There- fore, this work considers the throughput and trajectory optimization of terahertz (THz)-enabled unmanned aerial vehicles (UAVs) in sixth-generation (6G) communication networks. In the con- sidered scenario, multiple UAVs must provide on-demand terabits per second (TB/s) services to an urban area along with existing terrestrial networks. However, THz-empowered UAVs pose some new constraints, e.g., dynamic THz-channel conditions for ground users (GUs) association and UAV trajectory optimization to fulfill GU’s throughput demands. Thus, a framework is proposed to address these challenges, where a joint UAVs-GUs association, transmit power, and trajectory optimization problem are studied. The formulated problem is mixed-integer nonlinear program- ming (MINLP), which is NP-hard to solve. Consequently, an iterative algorithm is proposed to solve three sub-problems iteratively, i.e., UAVs-GUs association, transmit power, and trajectory optimization. The simulation results demonstrate that the proposed algorithm increased through- put by up to 10%, 68.9%, and 69.1%, respectively, compared to the baseline algorithms. In brief, this dissertation guarantees the establishment of self-sustaining intelligent networks in non-terrestrial environments for the upcoming generation of wireless networks. This achieve- ment is achieved through the creation of stochastic models and distributed learning approaches to meet the criteria and a specific set of key performance indicators (KPIs) for future wireless networks. The suggested approach is proficient at managing the intricate and ever-changing as- pects of communication, computation, and control within self-sustaining non-terrestrial wireless networks. These networks have the ability to govern and adjust their own operations and services while maintaining their own resources. Ultimately, they autonomously implement decisions aimed at upholding the enduring stability of the complex and dynamic computational, communication, and control environments required for next-generation wireless systems.