For semi-autonomous and fully autonomous driving, vehicles will also rely on information about other road users and conditions beyond their own field of vision or the V2V range, and much of this information needs to be specifically prepared or verified in advance. Data exchange follows the concept of vehicle-to-cloud-to-vehicle (V2C2V) communication, in which vehicles (V) establish a direct connection to the cloud (C) using the traditional mobile network.
Unlike V2V communication, the licensed spectrum of mobile network operators (MNO) is used. The new 5G network takes a central role for this: with its ability to route security-critical data traffic in a separate network slice, network quality can be actively managed (quality of service as opposed to best effort). Moreover, with the new generation of mobile phones, available bandwidth and network capacity will also significantly increase and reduce latency as well.
For security-critical applications, however, low latencies of approximately 1ms are only achieved if multi-access edge computing (MEC) is used in addition to 5G. An entirely new infrastructure market is emerging: 5G NewRadio will be supplemented by an edge cloud that can be provided by mobile carriers as well as the major cloud providers (Amazon, Microsoft, IBM, Google, etc.). At present, the standardization of the required interfaces is being expedited by the MEC Industry Specification Group.
In addition to MNOs and MEC providers, application developers also play a pivotal role in completing the infrastructure. Exemplifying this there are two areas of cloud-based applications for autonomous driving:
1. High-resolution real-time navigation services that reflect dynamic changes in the traffic situation in almost real-time: high-precision maps (HD maps) are continuously augmented, validated, and updated by information from many road users in a crowdsourcing process. Instead of using the central backend of the respective map providers (e.g., Here, TomTom, Navmii) for data processing, a segment of these activities can be outsourced as an application to an edge cloud.
The application receives information from its assigned region and updates regional map materials. Lastly, the map update is geo based, i.e., only the information required for the planned route is returned to the vehicle. Because of the decentralized edge structure, reduced latency and an optimized load distribution for the application provider is achieved in comparison to the previous central backend approach.
2. Intersection movement assistance(IMA): IMA applications control, by way of example, the right of way at complex intersections and depend on information being exchanged with maximum reliability and within a few milliseconds. On the edge cloud, IMA applications are hosted by third-party providers (e.g., Siemens, Savari) having a regionally limited, exclusive area of responsibility. With each new piece of information, the IMA application determines the vehicle sequence and sends it to all vehicles in the vicinity. The stream distributed via the 5G network in broadcasting mode enables network operators to achieve significantly higher network efficiency than the previous point-to-point distribution through the LTE network. What remains challenging is the fact that it will be a long time before all road users can be safely identified and factored in - particularly when vehicles or other road users without connectivity are involved.
In both examples, questions about roaming scenarios and interoperability are still unanswered, because all subscribers - regardless of their own network subscription - require access to the edge cloud, even abroad. To this end, an edge cloud must either be able to communicate with a foreign network edge cloud or enable communication for vehicles in foreign networks via local breakout procedures. However, such scenarios have not yet been defined or developed through common standardization groups. All three major MNOs in Germany are conducting trials to gain initial experience, and are cooperating with automobile manufacturers and their integrators (e.g., NuTonomy) as well as with suppliers (e.g., Continental).