Our research is located at the interdisciplinary point of intersection between Internet of Things (IoT), i.e., sensors and digital production technology and process-oriented information systems (process science). Our main research activity focuses on methods, principles, and techniques grounded in artificial intelligence for the formal specification, verification, monitoring, mining, and execution of process-oriented and event-driven information systems. For the specification and analysis of information systems we are considering descriptive, predictive as well as prescriptive analysis methods. We are studying the integration of several different models and systems, on the one hand to capture the system dynamics, and on the other hand to account for the underlying data and events. Furthermore, our research aims at providing methods, models, and guidelines that support organizations to effectively exploit the value propositions of IoT for improving business processes.
The world is increasingly linked through a large number of connected devices, typically embedded in electrical/electronical components and equipped with sensors and actuators, that enable sensing, (re-)acting, collecting and exchanging data via various communication networks including the Internet of Things (IoT). As such, it enables continuous monitoring of phenomena based on sensing devices (wearable devices, beacons, smartphones, machine sensors, etc.) as well as analytics opportunities in smart environments (smart homes, connected cars, smart logistics, Industry 4.0, etc.). Event processing focuses on capturing and processing events with minimum latency, i.e., near real-time, for detecting changes or trends indicating opportunities or problems. In the context of dynamic systems like process-oriented information systems, events may represent state changes of objects. Complex event processing (CEP) comprises a set of techniques for making sense of the behavior of a monitored system by deriving higher level knowledge from lower level system events in a timely and online fashion. We have investigated the feasibility of applying the concept of event processing to process-oriented information systems in order to allow for closed-loop monitoring and control of processes within IoT environments. We developed an architecture for integrating information systems and event processing systems in a closed monitoring and control loop through the exchange of (complex) events. The proposed techniques have been implemented and extensively evaluated on several real world case studies in digital production environments.
The second stream of research in this field can be summarized under the term physical analytics. It aims at optimizing spatially distributed processing steps in the manufacturing sector. The goal is to develop directly applicable and AI-based tools and methods for physical production processes with high variant diversity and flexibility. Recently, we were devising predictive techniques for the derivation of forecasts, i.e., predictive maintenance techniques in the context of distributed, event-driven production information systems by applying different classification and regression methods consisting of several uncorrelated decision trees and different settings of artificial neural networks.
The application of measurement and analysis techniques on event data from interaction systems has been proposed under the terms process mining and process analytics. Process mining is an approach at the intersection of model-driven engineering and data science, whose purpose is to analyse the event data generated through the execution of processes to obtain insights on how processes are executed in reality, and enable continuous improvement based on facts. Our contributions in this area can be grouped along three directions. First of all, while a large share of process mining focuses on automated discovery of imperative process models from event data, we contributed to create the sub-field of declarative process discovery, whose main goal is to extract rules from event data. The proposed techniques have been tested on several real world case studies. Second, we study how to predict processes based on deep learning models. In addition, we focused on the problem of data availability and preparation for process mining, by adopting (i) techniques from image recognition like small sample learning, and (ii) clustering methods to pre-process event logs.
We developed wearable user interfaces that allow humans to be notified in real-time at any location in case new tasks occur. In many situations humans must be able to directly influence data of IoT objects, e.g., to control industrial machinery or to manipulate certain device parameters from arbitrary places. We implemented an approach towards a framework for IoT data interaction by means of wearable process management. Additionally, humans can actively influence environmental data, e.g., production parameters, in real-time from arbitrary locations. This approach is based on speech recognition fueled by research in neural networks. We rely on an end-to-end (E2E) model approach that runs entirely on a smart device.
Our latest line of research aims at providing methods, models, and guidelines that support organizations to effectively exploit the value propositions of IoT for improving business processes. The objective is therefore a holistic consideration of IoT-based BPI including major existing challenges that prevent organizations from performing beneficial IoT projects. The topic of IoT-based BPI encompasses and connects the research areas of IoT and Business Process Management (BPM) respectively BPI.
At first, the identification of problems and challenges, as well as the investigation of potential opportunities for specific processes has been conducted. With regard to IoT technology, it must be clearly elaborated to what extend IoT can be exploited for BPI and which value propositions, fitting the respective processes, can be anticipated by decision makers. This core principle can be denoted as identifying possible IoT-based BPI propositions. Second, after having identified potential value propositions, the decision on appropriate IoT-based BPIs is required. Organizations need to be supported in the selection phase for specific IoT technologies and applications that fit the anticipated BPI goals and the underlying process details. Also, organizations need to have detailed knowledge about their internal capabilities to successfully realize these projects. As these capabilities can change over time, a continuous maturity assessment has been developed. This principle can be summarized as investigating advanced methods that support the selection of IoT technologies and applications.