The thesis focuses on the provision of cellular vehicle-to-everything (V2X) communications, which have attracted great interest for 5G due to the potential of improving traffic safety and enabling new services related to intelligent transportation systems. These types of services have strict requirements on reliability, access availability, and end-to-end (E2E) latency. V2X requires advanced network management techniques that must be developed based on the characteristics of the networks and traffic requirements. The integration of the Sidelink (SL), which enables the direct communication between vehicles (i.e., vehicle-to-vehicle (V2V)) without passing through the base station into cellular networks is a promising solution for enhancing the performance of V2X in cellular systems. In this thesis, we addressed some of the challenges arising from the integration of V2V communication in cellular systems and validated the potential of this technology by providing appropriate resource management solutions. Our main contributions have been in the context of radio access network slicing, mode selection, and radio resource allocation mechanisms. With regard to the first research direction that focuses on the RAN slicing management, a novel strategy based on offline Q-learning and softmax decision-making has been proposed as an enhanced solution to determine the adequate split of resources between a slice for eMBB communications and a slice for V2X. Then, starting from the outcome of the off-line Q-learning algorithm, a low-complexity heuristic strategy has been proposed to achieve further improvements in the use of resources. The proposed solution has been compared against proportional and fixed reference schemes. The extensive performance assessment have revealed the ability of the proposed algorithms to improve network performance compared to the reference schemes, especially in terms of resource utilization, throughput, latency and outage probability.Regarding the second research direction that focuses on the mode selection, two different mode selection solutions referred to as MSSB and MS-RBRS strategies have been proposed for V2V communication over a cellular network. The MSSB strategy decides when it is appropriate to use one or the other mode, i.e. sidelink or cellular, for the involved vehicles, taking into account the quality of the links between V2V users, the available resources, and the network traffic load situation. Moreover, the MS-RBRS strategy not only selects the appropriate mode of operation but also decides efficiently the amount of resources needed by V2V links in each mode and allows reusing RBs between different SL users while guaranteeing the minimum signal to interference requirements. The conducted simulations have revealed that the MS-RBRS and MSSB strategies are beneficial in terms of throughput, radio resource utilization, outage probability and latency under different offered loads comparing to the reference scheme. Last, we have focused on the resource allocation problem including jointly mode selection and radio resource scheduling. For the mode selection, a novel mode selection has been presented to decide when it is appropriate to select sidelink mode and use a distributed approach for radio resource allocation or cellular mode and use a centralized radio resource allocation. It takes into account three aspects: the quality of the links between V2V users, the available resources, and the latency. As for the radio resource allocation, the proposed approach includes a distributed radio resource allocation for sidelink mode and a centralized radio resource allocation for cellular mode. The proposed strategy supports dynamic assignments by allowing transmission over mini-slots. A simulation-based analysis has shown that the proposed strategies improved the network performance in terms of latency of V2V services, packet success rate and resource utilization under different network loads.