Vehicular networking is a cornerstone of Cooperative Intelligent Transportation Systems (C-ITSs) and autonomous vehicle management. It is expected to provide significant improvement in terms of transportation safety and efficiency. Cooperative awareness and sensing of the road environment is among the key services supported by vehicular networking. These applications rely on periodical exchange of one-hop broadcast messages.
The effectiveness of cooperative safety applications, built upon message exchange, highly depends on the accuracy and timeliness of status information contained in these messages and shared periodically among vehicles. We aim at defining a message congestion control able to cope with load limit requirements on the wireless channels and with application-relevant metrics characterizing the freshness of the collected data, namely the recently defined Age of Information (AoI).
Despite some heuristic proposals, there is still no satisfactory application-aware congestion control algorithm. In the proposed research we leverage on accurate analytic models that provide insight to identify the best message sending rate in complex vehicular scenarios where hidden vehicle nodes are possible.
The two main innovative results we tackle are therefore as follows. Identification of message congestion control to strike a controllable trade-off between efficient utilization of the wireless channel and timeliness of collected data, measured by the AoI performance (or new freshness metrics, to be identified in the course of the research). A second target is to compare the performance offered by the ETSI ITS G5 standard, based on Carrier Sense Multiple Access MAC protocol according to the amendment IEEE 802.11p, and cellular V2X (C-V2X) based on 5G New Radio. This comparison aims at contributing to the hot debate that has currently broken out in the automotive community about which technology promises better performance and safety levels.
The innovation expected from this research is twofold.
1. Design of congestion control mechanism for the vehicular network channel that accounts for ITS application requirements in terms of information timeliness.
2. Contribute to the comparison of the C-V2X and ITS G5 based approaches and provide input as to their possible co-existence.
As for the first point, the standard Decentralized Congestion Control (DCC) defined by ETIS for the ITS G5 wireless channel, based on CSMA, has been criticized for not being effective in controlling the load on the wireless channel while at the same time supporting update message exchange with AoI and delay requirements.
The cornerstone of DCC is measuring the Channel Busy Ratio (CBR) perceived by a node within its range, i.e., the mean fraction of time that the wireless channel is sensed as busy by the node or is used for transmission by the node. Load control adjusts the message grant rate according to the currently measured CBR.
Achieving a prompt message update delivery and keeping AoI low does not necessarily imply a low load on the wireless channel, even if high congestion levels may be harmful.
This is why a strong interest has arisen to re-designing DCC in an adaptive way to account for the load as well as the freshness of the information in node databases about its neighborhood.
The algorithm presented in Kaul et al. [1] is a first significant attempt along this line.
Such attempts are totally heuristic up to now.
We are developing accurate analytical models of vehicular networks based on CSMA that we plan to use as guideline to gain insight into the design of optimized message congestion control. Leveraging on this understanding, we plan to define a congestion control algorithm that strikes a tunable trade-off between load control on the wireless channel (as assessed by CBR) and freshness of data in node databases (as assessed by AoI). The algorithm should be distributed and only require data that can be collected from neighboring nodes.
As for the second contribution, we plan to consider also C-V2X based on 5G NR, specifically the so called Mode 1 and Mode 2 protocols, which define respectively a centralized and a distributed algorithm for getting radio resource to send one-hop broadcast messages via the cellular network.
Mode 1 adopts a centralized scheduling, with the base station sending in broadcast any message it receives via a random access channel.
Mode 2 defines a distributed semi-persistent randomized scheduling that allows direct communications among vehicular nodes.
There is currently a strong debate in the automotive community as to which should be the way to go. CSMA-based ITS G5 technology is already well established, while the newcomer 5G NR C-V2X is not really mature yet. There is however a hard push of cellular operators to seize at least part of the bandwidth reserved to CSMA-based ITS-G5.
The scientific community can contribute this debate in two ways: (i) by comparing the two technologies in given scenarios (we target a urban scenario) under the profile of both load control and AoI of the collected data; (ii) to propose and evaluate co-existence solutions.
We aim at providing work in the direction of the first target, leaving the second target for future work.
As of now there are only few initial studies that compare the two standards and they only focus on load control.
From very preliminary results that we are obtaining at the time of writing this proposal, it appears that the comparison is by no means obvious when one accounts also for message delivery delays. A good grasp of the fundamental trade-offs and of the most impacting parameters is valuable both from a scientific point of view and from a practical standpoint.
As a matter of fact, exchange of periodic updates in vehicular networks bear many similarities with sensor networks in many other contexts (e.g., meters in the smart greed, LoRaWAN devices sending data to central gateways, swarms of UAVs for rescue and exploration purposes, fleet of robots in industrial and logistic environments). Moreover, we believe that our studies could help shedding more light into the communication protocols useful to support the intense and time-sensitive data exchange required to support autonomous vehicles interactions among themselves and with the surrounding road environment.