Over the last decade, the advent of vehicular ad-hoc networks (VANETs) fulfilled various functionalities in intelligent transportation systems (ITS). Typical safety applications in ITS include exchange of mutual awareness messages among vehicles, broadcast of warning messages, cooperative driving and so on. The main purposes of these applications are to mitigate the risk of road accidents and improve the traffic efficiency. All of them rely on the technologies of vehicle-to-everything (V2X) communications, in which vehicles can communicate with vehicles, portable devices, as well as infrastructures. However, due to the highly dynamic topology of VANETs, lossy channels and limited bandwidth, the reliability and effectiveness of V2X communications is of great concern in both academic re-search and commercial applications. In this thesis, we aim to address the existing issues of unreliability and inefficiency in safety message dissemination. We propose a novel medium access control (MAC) protocol, enhancing the performance of the network through optimizing the MAC, and bettering resource allocation in (fifth generation) 5G cellular-V2X. A hybrid MAC protocol, namely NC-PNC MAC, for basic safety message (BSM) dissemination is proposed, based on the framework of Dedicated Short-Range Communications (DSRC). In the protocol, BSMs are chiefly broadcast during a centralized session, while a distributed session is reserved for unqualified vehicles that may only apply legacy DSRC MAC. Such characteristic not only can effectively suppress the transmission collisions but possess the compatibility with IEEE 802.11p. In addition, the integration of Physical-Layer Network Coding (PNC) and Random Linear Network Coding (RLNC) further strengthens the reliability and efficiency for BSM dissemination. Comprehensive simulations indicate that, compared with existing schemes, the proposed protocol can significantly improve the packet delivery ratio (PDR) by a range of 20% to 300%. Meanwhile, the normalized throughput of the whole network is substantially boosted by a varying percent between 20% and 160%. Extended studies involved in the hybrid NC-PNC MAC are conducted with the purpose of shedding light on the protocol and enhancing its performance further. The relevant studies include: 1) developing a mathematical model to analyse the subcarrier collision probability in an OFDMA scheme, the communication complexity of the MAC as well as the average PDR performance of the network; 2) identifying the key factors and investigating how they affect the PDR performance and delivery latency; 3) proposing two distinct strategies of performance enhancement based on the prior analysis and investigation. One is to ad-just the parameters of the protocol, while the other is to add extra relay nodes. Under the consideration of the requirement on delivery latency, the PDR can be boosted by 10% to 40% higher at an insignificant cost. Except for V2X in DSRC, V2X communications in 5G cellular networks are becoming a spotlight, and more advanced services are enabled in the 5G networks. To support those services, cutting-edge techniques are employed in both physical layer and MAC layer of the 5G cellular V2X. Among those techniques, resource allocation is an essential part as it determines how effectively the limited resources are utilized. In addition to studies carried in DSRC-based V2X networks, in the thesis, we also conduct research on 5G V2X networks, focusing on dynamic resource allocation in mode 3 and developing three algorithms. The first one is based on random allocation as a benchmark method, in which resources are randomly and exclusively allocated to different users. The second one is based on location-aware resource allocation (LARA) strategy tries to reuse the spectrum according to vehicle’s instant locations. To investigate their performances, we develop a network model for 5G V2X communications in urban areas, and then em-ploy typical V2X services in the analysis. Simulation results reveal that the two algorithms for resource allocation, while able to eliminate or reduce interference at reception, cannot provide guaranteed services, especially for the enhanced V2X services. The third pro-posed algorithm is a self-adaptive fuzzy logic-based resource allocation strategy (named FUZZRA). In FUZZRA, multiple factors are considered, such as data rate, interference level and message priority. By comprehensively processing the factors, the FUZZRA can intelligently dedicate resources to vehicles and other devices. Compared with the two prior proposed schemes, the fuzzy approach can effectively improve the resource utilization and boost the network performance; thereby satisfying the stringent requirements of the services.