When Service-oriented Computing Meets the IoT: A Use-case in the - PowerPoint PPT Presentation

When Service-oriented Computing Meets the IoT: A Use-case in the Context of Urban Mobile Crowdsensing Valérie Issarny, Inria Joint work with G. Bouloukakis, N. Georgantas, F. Sailhan, G. Texier, B. Lefèvre & others… 1 - 28/11/2019

Agenda 01. The promise of mobile crowdsensing but... 02. System architecture for the urban IOT 02. The challenge of sensor accuracy 03. Crowdsensing & The urban IoT networks 04. About the users participation 05. Conclusion 2 - 28/11/2019

01 The promise of mobile crowdsensing... But 3 - 28/11/2019

The power of the crowd 4 - 28/11/2019

The pollution monitoring use case Fixed sensing High cost but Mobile crowdsensing accurate Many but low accuracy Social sensing & diverse Qualitative but subjective 5 - 28/11/2019

Our initial research question Is mobile phone sensing an effective solution to the aggregation of urban knowledge? 6 - 28/11/2019

Our approach: Learning from an urban-scale experiment Ambiciti App informing about individual and collective exposure to urban pollution 7 - 28/11/2019

ambiciti.io The many facets of Ambiciti 8 - 28/11/2019

The many challenges of Ambiciti Accuracy of the measurements Low cost & heterogeneous sensors Context of observations Leveraging the diverse data sources Data assimilation Integrating with the urban IoT networks Gathering measurements from a large crowd Matching Technological and societal innovations And many more…. 9 - 28/11/2019

02 System architecture for the urban IoT 10 - 28/11/2019

A SOC-based fixed & mobile IoT 11 - 28/11/2019

Overcoming the scale 12 - 28/11/2019

Enabling multi-domain/region interactions 13 - 28/11/2019

03 The challenge of sensor accuracy 14 - 28/11/2019

Inaccuracy as the norm 2 calibrated sensors Non-calibrated sensor 15 - 28/11/2019

[R. Ventura et al . JASA 142(5), 2017] Assessing the sensors performance 16 - 28/11/2019

[R. Ventura et al . JASA 142(5), 2017] Phone SPL vs Reference SPL 17 - 28/11/2019

[R. Ventura et al . JASA 142(5), 2017] User-initiated calibration 18 - 28/11/2019

[F. Sailhan , V. Issarny & O Tavares Nascimento, MASS’17] Automated collaborative calibration Calibrate Y through collaboration with X as they meet • The measurements by Y & calibrated X, at time t, can be related as: y(t) = B 0 + B 1 x(t) 19 - 28/11/2019

[F. Sailhan , V. Issarny & O Tavares Nascimento, MASS’17] Multi-party regression 20 - 28/11/2019

[F. Sailhan , V. Issarny & O Tavares Nascimento, MASS’17] Robust regression filtering outliers 21 - 28/11/2019

[F. Sailhan , V. Issarny & O Tavares Nascimento, MASS’17] Collaborative calibration Calibration parameters Connectivity Graph Calibration Parameter Estimator Storage Manager Configuring Spatialised Measurements Connectivity Calibration Measurements Graph Parameters Sensor Manager Communication Service GPS Microphone System Discovery Register Unregister Reachable Sensing Devices 22 - 28/11/2019

[F. Sailhan , V. Issarny & O Tavares Nascimento, MASS’17] Assessing the relevance of a rendez- vous Meeting Meeting 2 3 W 23 W 32 Reference W 26 W 25 W52 Meeting 6 7 5 W W56 67 W65 Reference W 76 23 - 28/11/2019

[F. Sailhan , V. Issarny & O Tavares Nascimento, MASS’17] Evaluation 24 - 28/11/2019

[F. Sailhan , V. Issarny & O Tavares Nascimento, MASS’17] Evaluation Encouraging results in a controlled environment. Ongoing work focused on use in the wild Context-aware collaborative sensing 25 - 28/11/2019

04 Crowdsensing & The urban IoT networks 26 - 28/11/2019

[G. Texier & V. Issarny, LANMAN’18] Combining the IoT infrastructure and crowdsensing to extend the WSN lifetime 27 - 28/11/2019

[G. Texier & V. Issarny, LANMAN’18] The WSN 27 sensors & 1 sink 28 - 28/11/2019

[G. Texier & V. Issarny, LANMAN’18] The WSN leveraging mobile sinks Sensor Sink North path South path Sink SPF: Shortest path tree LB: Load balancing mn5S/N: 5 mobile sinks 29 - 28/11/2019

[G. Texier & V. Issarny, LANMAN’18] The LP formulation Flow conservation Exit at a sink Energy cost incl. routing 30 - 28/11/2019

[G. Texier & V. Issarny, LANMAN’18] The network lifetime Sensor Sink North path South path Sink 31 - 28/11/2019

[G. Texier & V. Issarny, LANMAN’18] Sensor lifetime analysis SPF mn5N 32 - 28/11/2019

05 About the users participation 33 - 28/11/2019

[V. Issarny et al. , Middleware’16] A look at contributed observations (10 months) Filtering Observations Known Bias Criteria Value # % - - 18,047,413 1279.0 Paris Loc accuracy < 100 meters 1,411,174 100.0 896,917 Location <30 meters 63.5 293,253 acc. Noise level <25 & >95 555,377 39.3 260,221 Outdoor Clustering 73,841 5.2 34,351 Speed < 7km/h 62,290 4.4 26,830 Proximity No 36,768 2.6 14,503 34 - 28/11/2019

[B Lefevre & V. Issarny, Smartcomp’18] The diverse user perspectives 35 - 28/11/2019

[B Lefevre & V. Issarny, Smartcomp’18] Take away & next step User-centered plastic interfaces Must allow the adaptation of the user interface by and for the user Effective usage environment & subjective aims Privacy Guaranteeing privacy remains an open research question 36 - 28/11/2019

06 Conclusion 37 - 28/11/2019

Back to our initial research question Is mobile phone sensing an effective solution to the aggregation of urban knowledge? Yes but… 38 - 28/11/2019

Thank you!! To know more valerie-issarny.me project.inria.fr/siliconvalley mimove.inria.fr ambiciti.io 39 - 28/11/2019