Implementing reliable communications is a major challenge for rail and road communications networks. Current V2X (Vehicle-To-Everything) technologies come up against major limitations. These include the existence of white zones and incomplete coverage by cellular networks (4G/5G). This implies the joint use of multiple access technologies at the level of each vehicle and a real-time and transparent transition between these radio access technologies. The definition of optimal handover mechanisms based on Artificial Intelligence (AI) tools is part of the answer to this problem. However, the definition of a high-performance network manager, guaranteeing low latencies, is also an important challenge. Therefore, this paper proposes a novel Software-Defined Network (SDN) controller designed to optimise dynamic handovers between wireless interfaces based on real-time network conditions. It also considers Machine Learning (ML) to determine performance parameter requirements. This approach improves the reliability, flexibility and adaptability of in-vehicle communication systems, helping to increase the efficiency of Cooperative Intelligent Transportation Systems.