220 likes | 241 Views
This paper explores the development of a Vehicular Data Cloud Platform that integrates IoT and cloud computing to provide intelligent transportation services. Detailed discussion on related works in IoT and Cloud, proposed platform architecture, and case study on Intelligent Parking Cloud Service. The study also delves into Vehicular Maintenance Data Mining and the challenges faced in scalability, performance, and security for such services.
E N D
Developing Vehicular Data CloudServices in the IoT Environment Wu He, Gongjun Yan, and Li Da Xu, Senior Member, IEEE Presented by Jonathan Lobo
Motivation • Modern vehicles already equipped with lots of sensors and communication devices • IoT and Cloud computing provide an opportunity to address transportation issues • Intelligent Transportation Systems (ITSs)
Intelligent Transportation System • A vehicular data platform using IoT and the cloud “where transportation-related information, such as traffic control and management, car location tracking and monitoring, road condition, car warranty, and maintenance information, can be intelligently connected and made available to drivers, automakers, part-manufacturers, vehicle quality controllers, safety authorities, and regional transportation division”
Related Work - IoT • Vehicular ad-hoc networks (VANET) • Support V2V and V2I communication • Integrate different communication technologies and sensor networks • Driver safety, traffic monitoring, roadside assistance • iDrive (BMW) • Informatics system using sensors to track vehicle location, road condition, and provide directions • Intelligent Internet of Vehicles Management System (IIOVMS) • Collect traffic information in real-time
Related Work - Cloud • Vehicular cloud service platforms • Integrate existing vehicular networks, sensors, on-board vehicular devices • Service Oriented Architecture (SOA) • Vehicular devices exchange services and information to collaborate in real-time • ITS-Cloud architecture • 3 layers: cloud, communication, end-user • Integrates in-vehicle CPS, V2V, V2I
Proposed Vehicular Data Cloud Platform • Goal: provide secure, on-demand vehicular services to customers • Conventional cloud • Data processing, high-level traffic administration applications • Temporary cloud • Formed on demand • Vehicles provide under-utilized computing power, networking, storage • Support for dynamic applications such as traffic monitoring, smart parking
Intelligent Parking Cloud Service • Goal : Collect and analyze geographic location, parking availability, parking space reservations, and traffic information to make finding parking spots easier $345 in wasted time, fuel, and emissions
Intelligent Parking Cloud Service • Vehicle has transceiver • Before arriving, reserve an open spot • When a car enters the parking lot, the entrance booth will validate the reservation and direct driver to the reserved parking slot • Parking lot has wi-fi network, infrared devices, and parking belts • Validate whether a car has parked
Intelligent Parking Cloud Service • Wireless transceiver tower in parking lot broadcasts parking lot information • Roadside transceivers display the information
Parking Service Models • Predict revenue / spot availability using birth-death stochastic process • Birth = vehicle entering parking lot • Death = vehicle leaving parking lot Birth rate Death rate Number of spots occupied at time t
Parking Service Models Occupied spots at time 0 Probability of a car parking event at time t when there are j cars trying to park Expected number of parked cars at time t
Vehicular Maintenance Data Mining • Motivation • Maintenance is frustrating for customers • Car manufacturers, parts designers can also learn from data • Goal • Data mine all vehicular maintenance data documents and classify by issue • Use data to detect dangerous road situations, issue warning messages, prevent accidents, assess vehicles’ performance
Naïve Bayes Classifier Documents Classes Classifier Training Data Joint probability
Naïve Bayes Classifier Maximum a Posteriori (MAP) Estimation
Logistic Regression Model Class Probability Model Parameters
Challenges • Scalability • Must be energy efficient, handle dynamically changing number of vehicles, spikes in traffic at different times, emergency situations • Performance, reliability, quality of service • Vehicles are moving so communication may be unreliable • Different cloud data centers to optimize response time • Lack of standardization • Coordination between stakeholders • Lack of clear business model • Security and privacy • Lack of established infrastructure for authentication and authorization
Conclusion • Architecture is great, but in order to be useful, services need to be developed and deployed • Integrating data from vehicular devices and the road infrastructure will allow innovation in the automobile industry