Refine
Year of publication
- 2022 (77) (remove)
Document type
- Conference Proceeding (77) (remove)
Keywords
- Poster (6)
- Blockchain (2)
- CNN (2)
- CPS (2)
- Convolutional neural network (2)
- Emotion recognition (2)
- ML (2)
- MLOps (2)
- Machine learning (2)
- Measurement environment (2)
Elevators contribute significantly to the electricity consumption of residential buildings, office buildings and commercial enterprises. In this paper, the electricity consumption is investigated using an elevator system and its individual operating states as example. In addition to analyzing and allocating the energy demand, this work examines how the individual operating states can be determined solely on the basis of the power consumption of the elevator. The knowledge gained from this, such as the usage behavior, the travel profile, or load, is determined independently of the elevator control system. A subsequent installation on any system can be easily realized. In this work, the investigated elevator requires a substantial part of the total annual power consumption in standby (>90 %). This shows an enormous potential for energy savings. The individual elevator states, as well as the load, can be detected very well on the basis of the measured total power consumption. The work thus shows exemplary the potential of an intelligent measurement system for the state detection of elevator systems.
In several domains of product design - like medical device design -
risk related use scenarios have to be analyzed in an early stage of design process.
Virtual reviews with users make it possible to get early insights in use problems.
Also, situations that are difficult to imitate in reality can be modeled and simulated
in virtual reality without risking the health of user. Therefore, virtual usability
tests are a promising approach which allow testing different scenarios under
controlled conditions. We chose a sample scenario from medical device design
and compare and evaluate different technical systems which can be used for virtual
usability tests. Aim is to derive guidance for virtual usability tests including
systems that can be used for specific conditions and the qualitative and quantitative
data, which can be collected with these systems. A formative test is performed
to evaluate and compare different systems. Also, a summative test is performed
to evaluate the selected systems. Results show that virtual usability tests
make it possible to test scenarios with users in an early stage and thus to encounter
possible interaction problems. But there are also many additional and new
things to consider compared to normal usability tests, such as checking motion
sickness, maintaining presence and the extensive operation of the technology.
Finally, a proposed method for virtual usability testing is described which also
comprises our equipment recommendations for virtual usability tests.
Operations within a Cyber Physical System (CPS) environment are naturally diverse and the resulting data sets include complex relations between sensors of the shopfloor devices setup, their configuration respectively. As Machine Learn- ing (ML) can increase the success of industrial plants in a variety of cases, like smart controlling, intrusion detection or predictive maintenance, clarifying responsibilities and operations for the whole lifecycle supports evaluating the potentially feasible scenarios. In this work, the need for highly configurable and flexible modules is demonstrated by depicting the complex possibilities of extending simple Machine Learning Operations (MLOps) pipelines with additional data sources, e.g., sensors. In addition to the particular modules core functionality, arbitrary evaluation logic or data structure specific anomaly detection can be integrated into the pipeline. With the creation of audit-trails for all operational modules, automated reports can be generated for increasing the accountability of the different physical devices and the data related processing. The concept is evaluated in the context of the project Collaborative Smart Contracting Platform for digital value-added Networks (KOSMoS), where a sensor is part of an ML pipeline and audit trails are realized using Blockchain (BC) technology.
Evolutionary strategy is increasingly used for optimization in various machine learning problems. It can scale very well, even to high dimensional problems, and its ability to globally self optimize in flexible ways provides new and exciting opportunities when combined with more recent machine learning methods. This paper describes a novel approach for the optimization of models with a data driven evolutionary strategy. The optimization can directly be applied as a preprocessing step and is therefore independent of the machine learning algorithm used. The experimental analysis of six different use cases show that, on average, better results are attained than without evolutionary strategy. Furthermore it is shown, that the best individual models are also achieved with the help of evolutionary strategy. The six different use cases were of different complexity which reinforces the idea that the approach is universal and not depending on specific use cases.
A simulation-based approach for measuring the performance of human-AI collaborative decisions
(2022)
Despite the widespread adoption of artificial intelligence and machine learning for decision-making in organizations, a wealth of research shows that situations that involve open-ended decisions (novel contexts without predefined rules) will continue to require humans in the loop. However, such collaboration that blends formal machine and bounded human rationality also amplify the risk of what is known as local rationality, which is when rational decisions in a local setting lead to globally dysfunctional behavior. Therefore, it becomes crucial, especially in a data-abundant environment that characterizes algorithmic decision-making, to devise means to assess performance holistically, not just for decision fragments. There is currently a lack of quantitative models that address this issue.
The importance of machine learning (ML) has been increasing dramatically for years. From assistance systems to production optimisation to healthcare support, almost every area of daily life and industry is coming into contact with machine learning. Besides all the benefits ML brings, the lack of transparency and difficulty in creating traceability pose major risks. While solutions exist to make the training of machine learning models more transparent, traceability is still a major challenge. Ensuring the identity of a model is another challenge, as unnoticed modification of a model is also a danger when using ML. This paper proposes to create an ML Birth Certificate and ML Family Tree secured by blockchain technology. Important information about training and changes to the model through retraining can be stored in a blockchain and accessed by any user to create more security and traceability about an ML model.
In diesem Vortrag werde ich auf die Programmierumgebungen (ROS) und Schnittstellen (keras/Tensorflow) eingehen, die es ermöglichen Roboter mit Hilfe von maschinellem Lernen zu trainieren. Dabei werde ich insbesondere die Möglichkeiten vorstellen, wie man einen Roboter in der Simulation (gazebo) trainieren kann, um die trainierten Modelle auf echte Roboter zu übertragen. Anhand von praktischen Beispielen mit mobilen Robotern und Greifarmen werden die Konzepte des Reinforcement Learnings, Active Learnings, Transfer Learnings und der Objekterkennung demonstriert. Das Testszenario besteht aus einem Holz-Labyrinth und einem Turtlebot Roboter, der mit Laser Range Scanner und einer 2D-Kamera ausgestattet ist. Dabei soll der Roboter lernen, autonom den Weg zur angegebenen Zielposition zu planen ohne dabei gegen ein Hindernis zu fahren. Es wird hierbei untersucht in wie weit die trainierten Modelle in leicht abgeänderten Szenarien funktionsfähig bleiben.