Fully autonomous intersections, where traffic lights are not needed, are a long-term goal in urban traffic planning. In these types of intersections, the priority rules and the order in which vehicles can cross the intersection are determined by wireless communication. Eliminating the human factor can lead to fewer incidents; however, to achieve this, the system must operate perfectly. In the case of autonomous intersections, communication between vehicles and infrastructure must be flawless, as must the execution of the desired maneuvers (for example, avoiding emergency situations). The aim of this research is to analyze autonomous vehicles, intersections, control methods, and how failures in different components of the specified infrastructure can lead to safety and security issues. Firstly, it is necessary to categorize the different types and elements of autonomous intersections and control methods. This includes analyzing the concepts and components of centralized and decentralized intersections, the components of self-driving vehicles, communication protocols, and the topics of safety, security, and cybersecurity in autonomous systems. Potential failures are identified using the Failure Mode and Effect Analysis (FMEA) method, which categorizes and evaluates the impact of different failures in the intersection system. This analysis provides an overview of potential failures and pinpoints components that pose the highest risk. In addition, this paper conducts a simulation-based analysis to assess the severity of some critical failures. Providing comprehensive solutions to such failures is beyond the scope of this work.

As connectivity of road traffic grows, so does the concern for its reliability and security. On the other hand, real-world testing of large-scale vehicular ad-hoc networks would be extremely costly. This paper presents a mixed-reality proof-of-concept for evaluating the communication of real vehicles using standard V2X messages while simulating the rest of the traffic. Detailed simulation of large-scale communication networks is also a computationally complex task. To be able to achieve the desired real-time simulation performance for mixed-reality testing, a so-called mesoscopic communication node was introduced, fusing the communication of the simulated vehicles into a single entity. This single entity realizes a sensor spoofing on other communicating vehicles. To verify the feasibility of the proposed approach, real-worlds tests were carried out involving two cars equipped with V2X devices and a third device transmitting standard messages from the simulation to the real vehicles in an "informed" denial-of-service-like way. Test results show that the mesoscopic simulation can retain real-time performance while replicating the communication of over 100 vehicles. Real vehicles can extract these virtual messages and packet drops remain in a plausible range.