Offloading Floating Car Data through V2V Communications Razvan - PowerPoint PPT Presentation

Offloading Floating Car Data through V2V Communications Razvan Stanica (INSA Lyon), Marco Fiore (IEIIT - CNR), Francesco Malandrino (Politecnico di Torino) 3èmes Journées Nationales des Communications dans les Transports (JNCT) Nevers - 30 May 2013

 Floating Car Data  Mobility Trace  Optimal Gain  Degree-based Mechanism  Degree-based with Confirmation  Reservation-based Mechanism Razvan Stanica INSA Lyon JNCT 2013 1 Offloading Floating Car Data 30.05.2013

Floating Car Data  Up to 100 Electronic Control Units in a car  Large amount of data produced, but not stored  Possible applications:  Fleet management  Remote diagnostic  Customized car insurance  Collaborative urban sensing  Road traffic monitoring  Navigation systems FCD Trace Optimal Gain DB DB-C RB Razvan Stanica INSA Lyon JNCT 2013 2 Offloading Floating Car Data 30.05.2013

Floating Car Data  Up to 100 Electronic Control Units in a car  Large amount of data produced, but not stored  Possible applications:  Fleet management  Remote diagnostic  Customized car insurance  Collaborative urban sensing  Road traffic monitoring  Navigation systems To enable such applications, data needs to be centralized and mined FCD Trace Optimal Gain DB DB-C RB Razvan Stanica INSA Lyon JNCT 2013 2 Offloading Floating Car Data 30.05.2013

Current model  Tom Tom HD Traffic, Meihui TrafficCast , Peugeot Connect Apps … FCD Trace Optimal Gain DB DB-C RB Razvan Stanica INSA Lyon JNCT 2013 3 Offloading Floating Car Data 30.05.2013

Current model  Tom Tom HD Traffic, Meihui TrafficCast , Peugeot Connect Apps …  Very expensive: price limits the frequency of FCD collection  The advantage of the low penetration ratio vs. Market share and application success FCD Trace Optimal Gain DB DB-C RB Razvan Stanica INSA Lyon JNCT 2013 3 Offloading Floating Car Data 30.05.2013

Current model  Tom Tom HD Traffic, Meihui TrafficCast , Peugeot Connect Apps …  Very expensive: price limits the frequency of FCD collection  The advantage of the low penetration ratio vs. Market share and application success FCD can represent a serious challenge for both cellular operators (network capacity) and service providers (collection price) FCD Trace Optimal Gain DB DB-C RB Razvan Stanica INSA Lyon JNCT 2013 3 Offloading Floating Car Data 30.05.2013

Offloading FCD  Use of Vehicle-to-Vehicle communication FCD Trace Optimal Gain DB DB-C RB Razvan Stanica INSA Lyon JNCT 2013 4 Offloading Floating Car Data 30.05.2013

Offloading FCD  Use of Vehicle-to-Vehicle communication  Local gathering and, perhaps, fusion of FCD: reduced network load, reduced cost, more data FCD Trace Optimal Gain DB DB-C RB Razvan Stanica INSA Lyon JNCT 2013 4 Offloading Floating Car Data 30.05.2013

Offloading FCD  Use of Vehicle-to-Vehicle communication  Local gathering and, perhaps, fusion of FCD: reduced network load, reduced cost, more data Addressed questions: What is the possible gain brought by offloading FCD? Is there a simple, practical mechanism for FCD offload? FCD Trace Optimal Gain DB DB-C RB Razvan Stanica INSA Lyon JNCT 2013 4 Offloading Floating Car Data 30.05.2013

Vehicular mobility trace  Largest freely available vehicular mobility dataset  24h on a 400km 2 region in the Köln area: more than 700k trips  Synthetic trace respecting real macroscopic measures  OpenStreetMap, SUMO, TAPAS, Gawron’s relaxation  For more information, see:  http://kolntrace.project.citi-lab.fr/  S. Uppoor, O. Trullols-Cruces, M. Fiore, J.M. Barcelo-Ordinas, Generation and Analysis of a Large-scale Urban Vehicular Mobility Dataset , IEEE TMC 2013 Trace FCD Optimal Gain DB DB-C RB Razvan Stanica INSA Lyon JNCT 2013 5 Offloading Floating Car Data 30.05.2013

V2V connectivity  Each second in the trace results in a connectivity graph  Two vehicles separated by less than R meters are considered connected (this does not imply a unit disk graph radio propagation model)  This allows the detection of important properties: assortative network Trace FCD Optimal Gain DB DB-C RB Razvan Stanica INSA Lyon JNCT 2013 6 Offloading Floating Car Data 30.05.2013

Optimal gain  Gain measured as percentage of vehicles that do not use the cellular uplink  The FCD offloading problem maps to a classical Minimum Dominating Set problem  MDS is NP-hard, but bounded approximation algorithms exist More than 60% of the FCD can be offloaded through V2V communication in more than 70 % of the cases Optimal Gain FCD Trace DB DB-C RB Razvan Stanica INSA Lyon JNCT 2013 7 Offloading Floating Car Data 30.05.2013

Impact of daytime  Up to 90% gain in the time intervals 6:30am-9am and 3:30pm-7pm  This second time interval maps to the daily peak in cellular traffic Optimal Gain FCD Trace DB DB-C RB Razvan Stanica INSA Lyon JNCT 2013 8 Offloading Floating Car Data 30.05.2013

Impact of geographical area  The gain depends on the vehicular density, so it is not evenly distributed  Some areas present a limited gain throughout the entire day, while a 95% gain can be achieved at peak hours in the central region Optimal Gain FCD Trace DB DB-C RB Razvan Stanica INSA Lyon JNCT 2013 9 Offloading Floating Car Data 30.05.2013

Looking for practical solutions  Centralized MDS already NP-hard  Many distributed algorithms proposed for backbone construction in wireless sensor networks  Trade-off between the quality of the MDS approximation and the amount of needed communication  Best algorithms find a dominating set after multiple communication rounds (a different message can be exchanged with each neighbor during a round)  Convergence time too long for a dynamic vehicular network FCD Trace Optimal Gain DB DB-C RB Razvan Stanica INSA Lyon JNCT 2013 10 Offloading Floating Car Data 30.05.2013