The technology evolution of Radio Access Network (RAN) in the context of 5th Generation (5G) is not only guided by improving the network performance but also by the need to transform all the technologies into intelligent dynamic ones. The new 5G is a flexible technology that will be able to satisfy at the same time each user of any type of mobility (static or mobile) or service request (real and non-real time service) without modifying any models or algorithms in networks. All physical use cases will be able to be considered by the network intelligently and resource managed automatically. The objective of this thesis is to analyse and enhance radio performance taking into account vehicular mobility by managing dynamically and intelligently the available resources. To this end, we describe different users mobility models for discrete and continuous modeling. The discrete model using the well-known car following model is well adapted for simulations. The continuous one is useful to derive analytical key performance indicators (KPI). The novelty of the thesis is the analytical formulation of KPIs that take into account the physical mobility in the radio traffic which is not necessary stationary. As an example, the impact of a traffic light on performance indicators in a cell is investigated. It is shown that a periodical physical traffic congestion due to the traffic light deteriorate periodically the cell performance. A first given solution is to improve resource allocation and control in the context of LTE-Advanced heterogeneous network. A small cell is deployed near the traffic light to relieve periodic congestion and QoS degradation. Three resource allocation and control schemes are investigated: a full frequency reuse, a static and a dynamic frequency splitting algorithm that are optimized with respect to a throughput based alpha-fair utility. For sake of financial and energy costs decreasing, another solution is provided using new antenna array technologies in order to manage efficiently heterogeneous, fixed and mobile traffic. A heterogeneous antenna system with different large antenna array technologies is considered to ooad static congestion areas and also the dynamical mobile congestion: Virtual Small Cell (VSC), virtual small cell with Self-Organizing Network (VSC-SON) and beamforming with multilevel global codebook that manages the heterogeneous antenna system at the Base Station (BS). The first two technologies improve the cell performance due to the capability to focus the signal at the traffic concentration. The novel beamforming solution with global codebook can further and significantly improve performance due to the capability to focus the signal along the road and to implicitly balance the traffic between the different antennas. We compare all these technologies and their impact on the network performance. The issue of user selection to allocate a portion (in time or in bandwidth) of the available resource is also analyzed. Moreover the context of resource management and network performance for 5G in high mobility is one of the future challenges. Thanks to the Minimization of Drive Testing (MDT) technology, networks can have Signal to Interference plus Noise Ratio (SINR) information with Geo-Localized Measurements (GLM).We introduce the concept of Forecast Scheduler for users in high mobility. It is assumed that a Radio Environment Map (REM) can provide interpolated SINR values along the user trajectories. Mobile users experience in their trajectories different mean SINR values. In mobile networks, schedulers exploit channel quality variation by giving the signal to the user experiencing best channel conditions while remaining fair. Nevertheless, we cannot record data rates of users with high mobility due to a very small time coherence. The Forecast Scheduling will exploit the SINR variation during users' trajectories.