Assisted driving encompasses a number of technical challenges, from requiring connectivity with ultra-high reliability and imperceptible delays, and disposing of powerful and flexible (migrable) storage and computing engines for coordinated and distributed road traffic control, to imposing legal privacy-preserving attributes and the possibility of logging traffic events and assisted driving decisions in an auditable way (which is needed, for instance, in case of legal disputes upon traffic accidents). Considering the complexity of making assisted driving control decisions for several coordinated players, and the need to actuate them at sub-second timescales, ECID proposes to leverage on edge/cloud computing and artificial intelligence spread and federated in the context of wireless access network infrastructures. Since a fundamental aspect of assisted driving consists in how to deal with legal aspects and in the auditability of driving decisions, ECID will develop decentralized and secure architectures of distributed ledgers, which offer the capability of logging events and the responsibility of actions in a trustworthy way and with minimum risk of malicious tampering.

The key concept that ECID leverages on consists the multi-access edge computing (MEC), originally introduced by 3GPP in the frame of the standardization of 5G networks and currently object of standardization efforts within ETSI. Moreover, network virtualization, artificial intelligence engines and security/privacy mechanisms of future networks will strongly relay on edge computing capabilities, and MEC features in particular, since the MEC offers an open platform for third party software needed to deploy and offer data services and run intelligent data analysis within the communication/computing network. ECID will use the MEC as an open and distributed platform where to deploy and smartly orchestrate the control functions that will enable intelligent assisted driving in the future. Through the MEC, ECID will instantiate and control assisted driving as a virtualized network function that can be migrated and replicated across the network, and that allows for the federation of multiple instances of the virtual function. Likewise, the MEC and its orchestrator will be used to build a flexible architecture for pushing network services where and when needed, which includes (i) the control of assisted driving tools, (ii) the instrumentation of the network with distributed ledgers and (iii) the support for the optimization of federated learning mechanisms. The adoption of the machine learning paradigm will be key to timely predict road events and connectivity issues, and to make decisions on migration and caching of data, services and virtual function instances.

Therefore, in this project we will tackle three classes of problems: (i) how to intelligently migrate the control of intelligent driving applications through the edge of the communication network, using MEC premises; (ii) how to log intelligent driving decisions in a way that cannot be tampered and can be used to solve legal disputes; and (iii) how to use artificial intelligence in a distributed way in edge computing environments integrated with mobile communication systems.

Research project PID2019-109805RB-I00 funded by: