Publication – Studying mobility patterns at urban intersections to form vehicle clusters for service deployment

Outline the paper

Many vehicles spend a significant amount of time in urban traffic congestion. Due to the evolution of autonomous vehicles, driver assistance systems, and in-vehicle entertainment, these vehicles have plentiful computational and communication capacity. How can we deploy data collection and processing tasks on these (slowly) moving vehicles to productively use any spare resources? To answer this question, we study the efficient placement of distributed services on a moving vehicle cluster.

This paper focuses on the concept of scaling and placement of distributed services on vehicle clusters, harnessing the knowledge of mobility patterns. The novelty of the work is in considering the mobility pattern of urban road traffic and utilizing moving vehicles as a potential site for deploying services. The services are made adaptable for the dynamic vehicular environment and can be scaled dynamically, based on the resource and mobility state of the multi-hop cluster. We have introduced a flow model for the traffic, depicted predictability in vehicular flow, and estimated communication capacity using real vehicular traffic data. We also introduced a detailed mathematical model for the mobility-aware scaling of distributed services based on resource-rich and resource-poor as well as stable and unstable cluster states.

Who will it help?

Vehicles follow predictable mobility patterns through different days of the week. For example, there is a huge influx of vehicles from residential areas in a city to the city centre and business parks. These patterns can easily be identified during peak-traffic hours. Some intersections close to airports have different mobility patterns in comparison. They have steady vehicular flows during the early morning as well as later in the evening. The vehicular flow at such intersections increases significantly during holidays. We use real vehicular density data from Dublin and California transport datasets to study the predictability of vehicular flows at urban sections of the cities. We then use these mobility patterns to utilize vehicles as infrastructure for service deployment.

What is the future of this research?

As part of the future work, a decentralized, mobility-aware task offloading algorithm will be introduced that solves the optimisation problem in real-time. To make the service model more practical, we will introduce a distributed service reconfiguration scheme to send collected data or service states back to the vehicle cluster. We aim to use hyper-parameter optimization techniques to decide the number of task instances to be deployed in real-time, for satisfying the service placement requirements. We will focus on the replacement of concurrent, data-dependent tasks as part of the failure recovery scheme.

Such applications can be used for surveying purposes for any kind of industry that requires large scale data collection. Our application not just collects data, but also processes data to improve the quality of data that is collected. For example, moving vehicles can study the density of people in city centres and their interaction with different businesses.

This publication has emanated from research conducted with the financial support of the Science Foundation Ireland (SFI) funded CONNECT project.

Publication Title: Scaling and Placing Distributed Services on Vehicle Clusters in Urban Environments
Authors: Kanika Sharma, Bernard Butler, Brendan Jennings
Publication Date: 10 May 2022
Name of Conference: IEEE Transactions on Network and Service Management
Link to publication: https://ieeexplore.ieee.org/document/9772343