State-of-the-Art approaches like Fledge IOT separate data collection mechanisms into a part that runs close to the machine (south) and a central part (north) that is responsible for data transformation and provisioning. Creating the south services either depends on using a fixed protocol (e.g. eprints.cs.univie.ac.at/5770/1/BPM_2018_paper_33.pdf) or on implementing an API (Fledge IOT, Azure IOT Hub). Besides efficient data collection, the biggest advantage of such approaches is remote management of large edge infrastructures. While each edge node has to have all the libraries and collection logic blocks installed, configuration to use them appropriately, can be deployed from a central configuration management facility to the edge node.
When it comes to the utilization of custom collection or transformation mechanisms, it is either necessary to write own custom collection logic (south), or it is done north. Doing it south, of course has the advantage that through harnessing semantic knowledge of the scenario, data volume and velocity can be drastically reduced without loosing information.
The purpose of this diploma thesis is to explore an alternative approach, where each edge node holds a process engine, that runs custom data collection logic. This would allow to harness the advantages of deploying custom data collection processes with deep semantic knowledge, without actually manually writing code. To evaluate the feasibility of such an approach different key characteristics have to be compared for a wide variety of scenarios: (1) Process Modeling Complexity vs. creation of collection logic blocks, (2) possible data collection frequencies, (3) Manageability and Deployability, ...
The results of this diploma thesis has to include a process engine based solution, a set of scenarios, an automatic evaluation suite, as well as an extensive comparison and discussion against alternative approaches.
Contact: master.i17 [at] in.tum.de