Refine
Year of publication
- 2020 (29) (remove)
Document Type
- Master's Thesis (29) (remove)
Language
- English (29) (remove)
Has Fulltext
- yes (29)
Is part of the Bibliography
- no (29)
Keywords
- International business (2)
- ROS (2)
- A priori & a posteriori segmentation (1)
- AI on the Edge (1)
- AI use cases (1)
- Acculturation (1)
- Acquisitions (1)
- Artificial intelligence (1)
- Austrian labour market (1)
- Automated optimisation (1)
The classification of waste with neural networks is already a topic in some scientific papers. An application in the embedded systems area with current AI processors to accelerate the inference has not yet been discussed. In this master work a prototype is created which classifies waste objects and automatically opens the appropriate container for the object. The area of application is in the public space.
For the classification a dataset with 25,681 images and 11 classes is created to re-train the Convolution Neuronal Networks EfficientNet-B0, MobileNet-v2 and NASNet-mobile. These Convolution Neuronal Networks run on the current Edge \acrshort{ai} processors from Google, Intel and Nvidia and are compared for performance, consumption and accuracy.
The master thesis evaluates the result of these comparisons and shows the advantages and disadvantages of the respective processors and the CNNs. For the prototype, the most suitable combination of hardware and AI architecture is used and exhibited at the university fair KasetFair2020. An opinion survey on the application of the machine is conducted.
Integration of an industrial robot manipulator in ROS to enhance its spatial perception capabilities
(2020)
Robots without any external sensors are not capable of sensing their environment, often leading to damaging collisions. These collisions could potentially be avoided if the robot had a way to sense its environment in the first place. This thesis attempts to tackle this problem by equipping such a robot with extra sensor hardware for perceiving environmental objects. The robot used within this thesis is a KUKA LBR iiwa 7 R800. The goal is a robot capable of moving in an unseen environment without colliding with obstacles nearby.
The research covers different sensor options, robots in cramped areas as well as algorithms and simulation topics. Software platforms and libraries used for the implementation are briefly introduced.
Multiple infrared sensors are directly installed onto the robot manipulator. The extra sensors and the robot are integrated into the ROS middleware to create an application capable of sensing the robots’ environment and plan collision-free paths accordingly.
The experiments show, that the low amount of available sensor data can not map the robots’ environment with enough detail. Additional problems, such as sensor noise corrupting parts of the generated map or the robot recognizing itself as an obstacle, lead to a negative result in total. In future work, the choice of sensors shall be reconsidered and tested upfront via simulation software.
An implementation approach of the gap navigation tree using the TurtleBot 3 Burger and ROS Kinetic
(2020)
The creation of a spatial model of the environment is an important task to allow the planning of routes through the environment. Depending on the number of sensor inputs different ways of creating a spatial environment model are possible. This thesis introduces an implementation approach of the Gap Navigation Tree which is aimed for usage with robots that have a limited amount of sensors. The Gap Navigation Tree is a tree structure based on depth discontinuities constructed from the data of a laser scanner. Using the simulated TurtleBot 3 Burger and ROS kinetic a framework is created that implements the theory of the Gap Navigation Tree. The framework is structured in a way that allows using different robots with different sensor types by separating the detection of depth discontinuities from the building and updating of the Gap Navigation Tree.
The humidification dehumidification (HDH) cycle is a process for thermal water treatment. Many studies were carried out investigating operation of an HDH cycle with water and seawater as working liquid. Currently research into other areas of application is limited. Exchanging the working liquid in the humidifier from seawater to a water oil emulsion and investigating its behavioural changes is the basis for the expansion into applications such as bilge water treatment. This master’s thesis covers analysis of the behaviour of an HDH cycle operated with a water oil emulsion. The main elements are (1) proof of concept for operation of the HDH cycle with a water oil emulsion, (2) comparison of measurements and thermodynamic calculations, (3) investigation of the impact of operating parameters and (4) optical analysis of the bubbly flow in water and oil.
Operation of the HDH cycle using water oil emulsion was shown to be feasible with a small change to the setup previously used for investigations with seawater as working liquid. To keep the emulsion from separating into its individual parts, constant movement of the working liquid needs to be ensured. For this a magnetic stirrer was introduced into the bubble column humidifier (BCH) used. In a batch process an oil concentration of >97 % was reached without visible traces of oil in the produced condensate.
Comparison of the measured and thermodynamically evaluated productivity shows that measured productivity is higher. The proposed explanation for this is supersaturation of air at the BCH exit. Further investigation into this phenomenon is needed to confirm this hypothesis.
Influential parameters investigated are (1) liquid temperature, (2) superficial air velocity and (3) sieve plate orifice diameter. Increase of liquid temperature results in an exponential increase in productivity. At superficial air velocities up to 3 cm/s productivity increases with superficial air velocity. For superficial air velocities higher than 3 cm/s productivity plateaus. At low superficial air velocity, an increase of sieve plate orifice diameter results in increasing productivity. Further increase of the sieve plate orifice diameter inverses this phenomenon.
Bubbly flow in water and oil is influenced by the different viscosities of the liquids. Water creates small bubbles of similar size at low superficial air velocities. At superficial air velocities >2 cm/s turbulences start to increase and finely dispersed bubbles are present in the water. Bubbly flow in oil creates larger bubbles at all superficial air velocities. The airflow transitions to plug flow at velocities of 3 cm/s and above.
Result from this master’s thesis can be used for as a basis to broaden the understanding of the HDH cycle and find new areas of applications.
Moving from one country to another, from one cultural context to a different one comes with many challenges and problems. The expatriate adjustment process, in general, has been evaluated extensively in the literature. Little is known if the knowledge in the literature is also valid for the situation of expatriates in rural Vorarlberg. In this paper was examined, which are the most common problems for highly skilled immigrants that are moving to Vorarlberg. In a mixed-method approach, information was gathered with an online questionnaire whose results served as a basis for a series of semi-structured interviews. In addition, an expert talk with a local relocation consultant was conducted. It was found that by far, the most severe difficulty is based on the domestic language situation. An expatriate needs to talk and understand German, but the local language is an Alemannic subsection of the German language that is not easy to understand. Additional difficulties that cause culture shock are limited opening hours, mobility troubles, and several others. The awareness about the composing of these problems might help to find the appropriate measures to support expatriates to come in the future.
Although pilot projects are an accepted means of entry into prospects, research on the object of startups selling SaaS and use pilots to enter and to further scale within their prospect’s organization is limited. The reader can expect a collection of key practices of SaaS startups in the field of Decision Support Software. These combine the main sales-oriented elements within pilot projects that are reflected on by Customer Success Management, Change Management as well as cultural dimensions. Explorative interviews, mainly with stakeholders in Decision Support Software startups, were conducted to further gain an understanding of the research object. Results indicate that pilots are strategically used in the sales of such startups to simultaneously deal with their customer’s uncertainties and as a means for the startups to get commitment and increase their value proposition through the additional service that they offer in order to acquire an internal support basis. Customer Success Management as well as Change Management are furthermore advantageous in quickly achieving measurable results that leverage buyers and seller’s justification for further sales.
The purpose of this work is to explore implicit schemes underlying the market segmentation analysis process. Boosting transparency for and in the new discipline of healthcare marketing, the work offers a toolbox of both primary and secondary methods to identify the accurate target market. This is crucial, since resource allocation in B2C segmentation and targeting is still often misleading. An Austrian, internationally present niche player serves as a research object to turn theoretical insights into practical verification. Data for the thesis are collected through company-internal data analysis and desk research, grounded in a multi-method approach with primary and secondary research. On the one hand, the work assesses the most effective segmentation and attractiveness/knock-out criteria according to scientific sources. Delving into the topic of a priori and a posteriori segmentation, an overview of suitable techniques is going to be offered. On the other hand, the thesis illustrates how the accurate target segment in the healthcare industry can be evaluated and determined through companyinternal consumer and market data.
Primary research on demographics (age, gender), psychographics (preferred channels), behavioral criteria (new/existing, CLC) and product categories is found to be particularly meaningful for the healthcare player. Results vary between countries, which is why an international-marketing strategy instead of a domestic-marketing approach is advisable.
Secondary research shows that socio-demographic and behavioral criteria are most used as a priori criteria, whereas a posteriori segmentation is promising to reveal psychographic clusters. One of the author’s recommendations is to niche down accurate market segments such as LOHAS or “best agers” by refining psychographics/socio-demographics with behavioral segmentation through “occasions” (e.g. back pain, depression, injuries). Novel approaches such as outcome-based segmentation or emphasizing “promoters” are discussed too.
The findings pave marketing managers the way for identifying the accurate target segments in the B2C health industry, selecting accurate methods grounded in profound scientific research and with concepts suitable for SMEs. The thesis proves that marketing segmentation is no longer a “nice-to-have” but a “must” in the health(care) industry.
Graphite substrates underwent two methods of creating doped silicon carbide films via carbothermal reduction; the first method being liquid-phase processing, or dip-coating, and the second gas-phase processing, otherwise referred to as the solid-vapour reaction. The dip-coating procedure resulted in flaky coatings, while the solid-vapour reaction resulted in polycrystalline films with columnar growth that displayed promising morphological and electrical properties. The films were tested on their performance as semiconductor diodes, and proved that carbothermal reduction in the gas phase is a promising technique for creating polycrystalline silicon carbide films for the application of light-emitting diodes.
Many test drives are carried out in the automotive environment. During these test drives many signals are recorded. The task of the test engineers is to find certain patterns (e.g. an emergency stop) in these long time series. Finding these interesting patterns is currently done with rule based processing. This procedure is very time consuming and requires a test engineer with expertise. In this thesis it is examined if the emerging field of machine learning can be used to support the engineers in this task. Active Learning, a subarea of machine learning, is used to train a classifier during the labeling process. Thereby it proposes similar windows to the already labeled ones. This saves the annotator time for searching or formulating rules for the problem. A data generator is worked out to replace the missing labeled data for tests. The custom performance measure “proportion of seen samples” is developed to make the success measurable. A modular software architecture is designed. With that, several combinations of Time Series Classification algorithms and query strategies are compared on artificial data. The results are verified on real datasets, which are open source available. The best performing, but computational intensive solution is an adapted RandOm Convolutional KErnel Transform (ROCKET). The custom query strategy “certainty sampling” shows the best results for highly imbalanced datasets.
This master thesis investigates a Computational Intelligence-based method for solving PDEs. The proposed strategy formulates the residual of a PDE as a fitness function. The solution is approximated by a finite sum of Gauss kernels. An appropriate optimisation technique, in this case JADE, is deployed that searches for the best fitting parameters for these kernels. This field is fairly young, a comprehensive literature research reveals several past papers that investigate similar techniques.
To evaluate the performance of the solver, a comprehensive testbed is defined. It consists of 11 different Poisson equations. The solving time, the memory consumption and the approximation quality are compared to the state of the art open-source Finite Element solver NGSolve. The first experiment tests a serial JADE. The results are not as good as comparable work in the literature. Further, a strange behaviour is observed, where the fitness and the quality do not match. The second experiment implements a parallel JADE, which allows to make use of parallel hardware. This significantly speeds up the solving time. The third experiment implements a parallel JADE with adaptive kernels. It starts with one kernel and introduce more kernels along the solving process. A significant improvement is observed on one PDE, that is purposely built to be solvable. On all other testbed PDEs the quality-difference is not conclusive. The last experiment investigates the discrepancy between the fitness and the quality. Therefore, a new kernel is defined. This kernel inherits all features of the Gauss kernel and extends it with a sine function. As a result, the observed inconsistency between fitness and quality is mitigated.
The thesis closes with a proposal for further investigations. The concepts here should be reconsidered by using better performing optimisation algorithms from the literature, like CMA-ES. Beyond that, an adaptive scheme for the collocation points could be tested. Finally, the fitness function should be further examined.