Intelligent 5G Network Slicing for Vehicular Networks

5G and beyond is envisioned as a flexible and heterogeneous network that supports use cases with a wide range of requirements. This thesis focuses on the optimisation of physical Resource Block (RB) and power allocation among users, bandwidth allocation among Bandwidth Parts (BWPs), and data routing and user association in vehicular networks in a wireless network, considering the concept of network slicing. All the above-given elements involve solving a complex joint optimisation problem with multiple objectives and constraints, including minimising power consumption, maximising bandwidth and energy efficiency, ensuring the quality of service, and maintaining network slicing requirements. This thesis addresses the challenges of solving joint optimisation problems by proposing novel algorithms and techniques that enable efficient and effective optimisation for the above-given objectives. This thesis, therefore, employs three optimisation tools, namely, MILP, heuristics and machine learning. The study considers a city environment with multiple cells that support radio slicing and multi-numerology techniques. Randomly located users generate traffic demands characterised by data rate and delay requirements. Additionally, the vehicles in the network serve as mobile Integrated Access and Backhaul (IAB) nodes, referred to as Vehicular Base Stations (VBSs). In this work, the backhaul network utilises mmWave frequency bands, while the access links operate in sub-6 GHz frequency bands. The developed model can be flexibly adapted to support other frequency distributions between backhaul and access networks. The VBSs establish communication links with nearby BSs and users, facilitating the exchange of data. This combination of mmWave and sub-6 GHz frequency bands allows for efficient and reliable communication between the VBSs and users, enabling the seamless operation of the IAB network for vehicular-based backhaul and access communication. In the MILP optimisation, we consider the weighted objectives of minimising the total transmit power in the downlink direction and co-channel interference (CCI) while maximising the number of simultaneously served users. MILP efficiently allocates RBs and power to the users, bandwidths among BWPs, and makes data routing and user association for the proposed optimisation framework while considering CCI and the many required constraints. MILP allows the simultaneous optimisation of multiple objectives, resulting in an effective and efficient optimisation strategy in the vehicular-based mmWave backhaul network. A multi-objective problem involving multiple dimensions causes an exponentially increasing number of parameters and variables, therefore, prolonged computation times and the need for extensive verification of obtained outputs. Hence, we propose a hybrid optimisation framework designed to address the complex problem of resource allocation and routing in multi-numerology systems with multiple users and sources. By combining heuristic and meta-heuristic approaches, specifically genetic algorithms, we overcome the performance limitations of MILP and provide an efficient, effective, reliable and alternative solution. Furthermore, the same scenario with a low number of users is employed in MILP and heuristic algorithms, and the outputs are compared for verification purposes. In addition to the heuristic algorithms, we also incorporated a Reinforcement Learning algorithm to further reduce decision time for the optimisation problem. We utilised various techniques to optimise power and numerology allocation among users in a pre-allocated BWP environment. With these techniques, we aimed to achieve optimal resource utilisation while meeting the various requirements of users in a 5G and beyond RAN.