Publications 2017

Ubiquitous computing and The Internet-of-Things (IoT) emerge rapidly in today’s life and evolve to Self-organizing systems (SoS). A unified and scalable information processing and communication methodology is required. In this work mobile agents are used to merge the IoT with Mobile and Cloud environments seamlessly. A portable and scalable Agent Processing Platform (APP) provides an enabling technology that is central for the deployment of Multi-agent Systems (MAS) in strongly heterogeneous networks including the Internet. A large-scale use-case deploying Multi-agent systems in a distributed heterogeneous seismic sensor and geodetic network is used to demonstrate the suitability of the MAS and platform approach. The MAS is used for earthquake monitoring based on a new incremental distributed learning algorithm applied to regions of sensor data from stations of a seismic network with global ensemble voting. This network environment can be extended by ubiquitous sensing devices like smart phones. Different (mobile) agents perform sensor sensing, aggregation, local learning and prediction, global voting and decision making, and the application. The incremental distributed learning algorithm outperforms a prior developed non-incremental algorithm (Distributed Interval Decision Tree learner) and can be efficiently used in low-resource platform networks.

Load-bearing structures are typically designed towards relevant load cases assuming static shape and fixed sets of materials properties decided upon during design and materials selection. Structures that could change local properties in service in response to load change could raise additional weight saving potentials , thus supporting lightweight design and sustainability. Materials with such capabilities must necessarily be composite in the sense of a heterogeneous build-up, exhibiting e. g. an architecture consisting of numerous active cells with sensing, signal and data processing and actuation/stimulation capability. One concern regarding active smart cellular structures is correlated control of cells’ responses, and the underlying informational organization providing robustness and real-time capabilities. We suggest a two-stage approach which combines machine learning with mobile and reactive Multi-agent Systems (MAS). In it, the MAS’ task is to analyze loading situations based on sensor data and negotiate matching spatial redistributions of material properties like elastic modulus to achieve higher-level optimization aims like a minimum of the total strain energy within the structure, or a reduction of peak stress levels. The associated machine learning approach would be employed to recognize loading situations already encountered in the past for which optimized solutions exist and in such cases bypass the MAS system to directly enforce the respective property distribution. In the present study, a proof of concept of the approach is presented which combines finite element method (FEM) and MAS simulation, with the former primarily taking the place of the physical structure. In addition, FEM simulations are used for off-line training of the MAS prior to its deployment in the real or simulated structure. The classification models learned this way represent a starting point which is constantly being updated at run-time during the service life of the structure using incremental learning techniques.

Smart mobile devices are fundamental date sources for crowd activity tracing. Large-scale mobile networks and the Internet-of-Things (IoT) expand and become part of pervasive and ubiquitous computing offering distributed and transparent services. With the IoT, Crowd Sensing is extended by Things Sensing, creating heterogeneous smart environments. A unified and common data processing and communication methodology is required so that the IoT, mobile networks, and Cloud-based environments seamlessly integrate, which can be fulfilled by self-organizing mobile agents, discussed in this work. Currently, portability, resource constraints, security, and scalability of Agent Processing Platforms (APP) are essential issues for the deployment of Multi-agent Systems (MAS) in highly heterogeneous networks. Beside the operational aspects of MAS, an organizational structure is required for the deployment of MAS in crowd sensing and social mining applications. The Planetary Nervous system (Nervousnet) consists of virtual sensors building the core functionality for such applications running on smart phones with a Cloud-like architecture. The virtual sensors enable a holistic composition and modeling approach. Self-organizing and adaptive mobile agents are well known as the core cells of holistic and modular systems. In this work, both concepts are combined. JavaScript agents are introduced as virtual sensors in the Nervousnet environment, evaluated with a simulation of a distributed sensor fusion use-case in a mobile network based on real-world data from Nervousnet, showing the suitability of the hybrid approach, benefiting from local and event-based sensor processing performed by the MAS.

In the past decades there was an exponential growth of computer networks and computing devices, connecting computers with a size in the m3 range. The Internet-of-Things (IoT) emerges connecting everything, demanding for new distributed computing and communication approaches. Currently, the IoT connects devices with a size in the cm3 range. But new technologies enable the integration of computing in materials and technical structures with sensor and actor networks connecting devices in the mm3 range. This work investigates issues in large-scale computer networks related to the deployment of low- and very-low resource miniaturized nodes integrated within materials. These networks operate under harsh conditions with possibility of technical failures requiring robustness. Despite sensor networks used for structural monitoring, self-adaptive materials can profit from self-organizing and autonomous distributed data processing using Multi-agent systems, demonstrated in this paper. Self-adaptive materials are able to adapt the material or mechanical structure properties based on their environmental interaction (load/stress) to minimize the risk of overloading. A structure that could change its local properties in service based on the identified loading situation could thus potentially raise additional weight saving potentials and thus supporting lightweight design, and in consequence, sustainability.

Monitoring of mechanical structures is a Big Data challenge and includes Structural Health Monitoring (SHM) and Non-destructive Testing (NDT). The sensor data produced by common measuring techniques, e.g., guided wave propagation analysis, is characterized by a high dimensionality in the temporal and spatial domain. There are off- and on-line methods applied at maintenance- or run-time, respectively. On-line methods (SHM) usually are constrained by low-resource processing platforms, sensor noise, unreliability, and real-time operation requiring advanced and efficient sensor data processing. Commonly, structural monitoring is a task that maps high-dimensional input data on low-dimensional output data (information, that is feature extraction), e.g., in the simplest case a Boolean output variable “Damaged”. Machine Learning (ML), e.g., supervised learning, can be used to derive such a mapping function. But ML quality and performance depends strongly on the input data size. Therefore, adaptive and reliable input data reduction (that is feature selection) is required at the first layer of an automatic structural monitoring system. Assuming some kind of two-dimensional sensor data (or n-dimensional data in general), image segmentation can be used to identify Regions of Interest (ROI), e.g., of wave propagation fields. Wave propagation in materials underlie reflections that must be distinguished, especially in hybrid materials (e.g., combining metal and fibre-plastic composites) there are complex wave propagation fields. The image segmentation is one of the most crucial part of image processing (Mishra, 2011). Major difficulties in image segmentation are noise and the differing homogeneity (fuzziness and signal gradients) of regions, complicating the definition of suitable threshold conditions for the edge detection or region splitting/clustering. Many traditional image segmentation algorithms are constrained by this issue. Artificial Intelligence can aid to overcome this limitation by using autonomous agents as an adaptive and self-organizing software architecture, presented in this work. Using a collection of co-operating agents decomposes a large and complex problem in smaller and simpler problems with a Divide-and-Conquer approach. Related to the image segmentation scenario, agents are working mostly autonomous (de-coupled) on dynamic bounded data from different regions of an image (i.e., distributed with simulated mobility), adapted to the locality, being reliable and less sensitive to noisy sensor data. In this work, different agent behaviour and segmentation approaches are introduced and evaluated with measured high-dimensional data from piezo-electric acusto-ultrasonic sensors recording wave propagation in plate-like structures. Commonly, SHM deploys only a small set of sensors and actuators at static positions delivering only a few temporal resolved sensor signals (1D), whereas NDT methods additionally can use spatial scanning to create images of wave signals (2D). Both one-dimensional temporal and two-dimensional spatial segmentation is considered to find characteristic ROIs.