The BMW Group is currently thinking about evolving their current vehicle-to-backend communication protocol to a new Message Queuing Telemetry Transport Version 5.0 (MQTT) protocol derivative. Part of this new protocol is a gateway application deployed inside the vehicle. Its responsibility is to connect an arbitrary amount of onboard applications with the backend, utilizing only a single mobile connection. Our task was to develop a queueing strategy for this gateway application, governing the outward-facing message queue. We found that the network topology can be abstracted to resemble a delay-tolerant network. For this special subset of networks, queueing strategies based on revenue management have been successfully applied in the past. In this cost-revenue based approach, each queue slot gets assigned a cost and every client holds a specific amount of money. A client is able to acquire a queue slot only if it possesses an amount of money equal or greater than the cost of that slot.
From this theoretical approach, we developed the necessary money and cost functions and simulated both our proposed approach and the drop tail strategy used by the current protocol. To simulate realistic bandwidth limitations, we collected bandwidth data in a test drive and used it in the simulation. When evaluating the results using predefined fairness objectives, our approach performs significantly more fair.
In order to further evaluate the algorithm, we implemented the component and tested both our proposed queuing strategy and the drop tail strategy again. These tests show the same improvements in fairness, proving this new approach to be well suited.
Evaluation and Optimization of Vehicle-to-Backend Communication
Ulm University Ulm UniversityBA Abschlussvortrag, Dennis Hertrich, Ort: Online, Datum: 05.05.2021, Zeit: 10:00 Uhr