Edge Analytics – A gateway-centric software architecture & algorithms for the IoT
Sketch of a dissertation topic, details & relevant literature to be refined
As of 2019-06-03, by Dr. Irene Cramer (INST/ECS5), Bosch Software Innovations GmbH,
Edge computing is gaining more and more popularity in the IoT domain. Most IoT scenarios require a combination of edge and cloud computing as a complementary approach, for example, for high computational processing and storage in the cloud on the one hand, and quick local decisions on the other. However, aspects like latency, data privacy, costs, and autonomy pose challenges to the centralized way of cloud computing and put edge computing in favor: The required processing is placed as close as possible to the data sources, which makes sending data across the Internet superfluous.
While most PaaS and IaaS service providers in the IoT focus on big data approaches (including a Machine Learning (ML) model training in the backend/cloud and a rollout of the resulting ML models from the backend/cloud to the gateway), our approach is to perform a (lightweight) training locally on the gateway (for specific IoT Analytics use cases). This approach is based on two components of the Bosch IoT Suite: (1) Bosch IoT Gateway Software and (2) Bosch IoT Analytics.
Ad (1): The Bosch Iot Gateway Software operates on an edge node of type “gateway”. It is fully hardware independent and has a proven track record of more than 40 types of gateways. It runs on common operating systems such as Linux, Windows, Mac OS, Android, and VxWorks. A prerequisite to run our gateway software is a Java virtual machine. The product is based on OSGi specification which provides modular framework with possibility to dynamically install and update new software.
Ad (2): The Bosch IoT Analytics complements the Bosch IoT Gateway Software by bringing Machine Learning and Data Mining functionalities to the gateway.
Bosch IoT Analytics supports the following use case patterns:
• Predictive Maintenance & Condition Monitoring, i.e. improve maintenance processes, reduce outages and unplanned downtimes, hence reduce related cost
• Quality Signals, i.e. improve product quality, hence reduce the number of claims and/or improve customer satisfaction
• Usage Profiling, i.e. gain knowledge about customer groups and/or improve product market fit
In order to allow a lightweight ML model training, a dedicated software architecture on the gateway and a special set of algorithms are needed. The architecture and the algorithms should be able to work with limited amount of data and limited computational resources as well as satisfy the needs of the above mentioned most relevant Analytics use cases in the IoT.
This dissertation project will study both the architecture and the algorithms and will evaluate them against a number of data sets from real-world IoT projects/device fleets. Any implementation can leverage Bosch IoT Gateway Software and Bosch IoT Analytics and should be integrated with both components as far as possible. In addition, the dissertation should investigate the boundaries of when a completely local approach is feasible (and under what conditions) and deduce an operationalization to support projects with their design decisions.
The applicants must hold a Master's degree in Computer science, Computer Engineering, Applied mathematics or equivalent. English language is a must. They must have knowledge in:
• Principles of Artificial Intelligence, Machine Learning, Statistics & Probability Theory
• Data Management (knowledge on Data Warehouses + Data Architecture/Lambda Architecture, ETL technologies, is a plus)
• Programming skills: Java and Python
• Knowledge of German language is a plus.
· Satyanarayanan M et al. (2015). Edge analytics in the internet of things. IEEE Pervasive Computing 14(2):24-31
· N. Harth, C. Anagnostopoulos, D. Pezaros, Predictive intelligence to the edge: impact on edge analytics, Evolving Systems 9(2), 2017.
· N. Hassan et al. The Role of Edge Computing in Internet of Things, IEEE Communications Magazine PP(99), 2018.
· Shi, W., Cao, J., Zhang, Q., Li, Y., & Xu, L. (2016). Edge computing: Vision and challenges. IEEE Internet of Things Journal, 3(5), 637-646.
· Yu, W., Liang, F., He, X., Hatcher, W. G., Lu, C., Lin, J., & Yang, X. (2018). A survey on the edge computing for the Internet of Things. IEEE Access, 6, 6900-6919.
Please submit a copy of your Master’s certificate, CV and a motivation letter, outlining the match between your professional experience and the research project topic, as well as demonstrating your independent thinking and leadership abilities by
30. November 2019