Computational Intelligence for Dynamic Risk Assessment in IoT-connected Autonomous Vehicle Networks

Abstract

Autonomous vehicles, also known as driverless, connected or self-driving cars, are no longer a concept of the more or less distant future, but a transportation factor that, depending on the formal definitions (such as "level-5" autonomy), should reach the full-public availability in the next fifteen years. Encouraged by major vehicle manufacturers who have declared innumerable efforts in research and development, as well as by the spread of vehicles with progressively "partial" or "conditional" self-driving functions, numerous entities have undertaken a wide range of activities to prepare the communications infrastructure and protocols for the full development of these revolutionary vehicles, covering issues related to safety, latency, quality and security of communications. Indeed, such an integration of digital technology can define more efficient traffic flows and stretches in both urban and road areas, revolutionizing some current models and rules. To date, the majority of the functional platforms proposed and some prototypes of autonomous vehicles using recent technology, such as clusters of processors connected to networks with different topologies, i.e., CAN, LIN, FlexRay, Ethernet (IP), DSRC (Dedicated Short Range Communication), with different requirements for continuity and safety, cyclic transmission and redundancy protocols for certain functions, have been designed and tested for certain types of roads and services. Due to the intrinsic security problems related to each technology, sensible to external attacks or faults, the subject of cybersecurity is recognized as an enabling factor both for privacy-sensitive services for passengers and for security-sensitive applications.