Osmotic MindSphere: Multi Resolution Air Quality Modelling across Cloud and Edge

Posted on

< Project Overview >

Introduction

This research looks to develop a prototype transformative ICT platform (based on IoT sensing, cloud computing, and edge computing approaches).This project looks to fuse knowledge from academic research in IoT application orchestration and real-world deployment of IoT devices to extend an existing IoT platform, i.e. The MindSphere platform provided by our industry partner, Siemens. Extending an existing ICT platform, provided by one of the lead players in the IoT market, for a test deployment (facilitated by Newcastle University) by a public body (Newcastle City Council), establishes the means of exchanging cutting-edge research knowledge and experience with industrial partners and a ready route to market. The involvement of Newcastle City Council in this project will not only ensure excellent societal benefits (i.e. real-time AQ monitoring and modelling) but also transfer of our innovative knowledge into a software product for the public good.

Project aims

In this project, we aim to overcome deep learning model deployment challenges. Deep learning models are used in many areas, such as air quality monitoring and prediction. However, these models are always trained offline and open-sourced by machine learning researchers. During the training phase, the deployment constraints such as computing resources, network conditions may not be considered. Therefore, these models may not perform very well in a real-world deployment.

To overcome these barriers, we developed an Osmotic platform that deploys deep learning models automatically in various computing resources, including IoT platform (e.g., MindSphere), edge computing platform (Raspberry Pi, Nvida Jeston nano), and cloud computing platform (e.g., Amazon EC2).

Additionally, to solve the model updating issue, the Osmotic platform is able to retrain a model automatically with new data, while redeploying the retrained model on the available resources.

This platform offers a fast solution to university researchers and company developers to test their machine learning models on different resources. Our developed real-time air quality monitoring system can also be transferred to Newcastle city council to ensure societal benefits.

What was done?

Cooperation with companies – we had  meetings with Siemens MindSphere and Software AG to exchange knowledge and experience. The discussion gave us many good ideas for designing the Osmotic platform.

Cooperation with air quality experts – we had discussions with colleagues from the  department of Earth and Environmental Sciences at Manchester University. They offered us a deep learning based air quality model.

Results

We developed the Osmotic platform that offers rich features to deploy machine learning models.

We developed a software that utilises the deep learning technology to do real-time air quality reduction.

We developed an IoTSim-OSMOSIS simulation platform that helps researchers and practitioners in testing their edge-cloud resource and data management policies.

Deliverables and other tangible outputs

Publications:
D.N Jha, Y. Li, Z.Wen, G. Morgan, P. P Jayaramn, A. Y. Zomaya and R. Ranjan, “GeoDeploy: A User-Centric Cost-Efficient Geo-Distributed Web-Applications Deployment”, IEEE Transactions on Computers (Under review)

U. Demirbaga, Z. Wen, A. Noor, K. Mitra, S. Garg, A. Y. Zomaya and R. Ranjan, “AutoDiagn: An Automated Real-time Diagnosis Framework for Big Data Systems”, IEEE Transactions on Computers (Under review)

B. Qian, Z. Wen J. Tang, Y. Yuan, A. Y. Zomaya and R. Ranjan, “OsmoticGate: Adaptive Edge-based Real-time Video Analytics for the Internet of Things”, IEEE Jouranl on Selected Areas in Communications (Under review)

Z. Wen, R. Yang, B. Qian, Z. Wang, J. Xu, A. Y. Zomaya and R. Ranjan, “Janus: Latency-Aware Traffic Scheduling for Data Streaming in Edge Computing”, 2021 IEEE 41th International Conference on Distributed Computing Systems (ICDCS’21) (Under review)

K. Fizza, A. Banerjee, K. Mitra, P. P. Jayaraman, R. Ranjan, P. Patel, D. Georgakopoulos,”QoE in IoT: A Vision, Survey and Future Directions”, Discover Internet of Things (Accepted).

Software: Osmotic platform: https://github.com/loftytopping/Prophet_forecasting_AQ

Demo: we are preparing a demo to be released at the end of February

K. Alwasel, D.N. Jha, F. Habeeb, U. Demirbaga, O. Rana,T. Baker, S. Dustdar, M. Villari, P. James, and R. Ranjan, “IoTSim-Osmosis: A Framework for Modelling & Simulating IoT Applications Over an Edge-Cloud Continuum,” Journal of Systems Architecture. (accepted December 2020).

Impact

The outcomes of this project are expected to create both industrial and academic impact. We plan to deploy the Osmotic platform in the Urban Observatory testbed for conducting real-time air quality modelling. Newcastle City Council will be a direct beneficiary of this deployment. Finally, IoTSim-OSMOSIS is released as an open source project for the academic and industrial community.

We expect this tool to become a de facto platform for conducting performance benchmarking research in edge-cloud settings.

Next steps

Develop REF 2027 case study around Osmotic and IoTSim-OSMOSIS software platforms.

Lessons learned

The Osmotic framework is used in other projects for machine learning model deployment. Also, the air quality monitoring prototype has inspired us to generate more ideas to attract more grants.

What has Pitch-In done for you?

This funding has given my team a unique opportunity to solve a real-world problem faced by the city council. By using this funding, we were able to develop an air quality monitoring software tool that can work in the edge-cloud continuum.

Project lead:

Professor Ravij Ranjan, Newcastle University

Project partners:

  • Siemens
  • Newcastle City Council