The FiRÆ project aims to free mobile robots from their shackles - namely, their power cords. We do this because the traditional power-cord-and-battery configuration will always be the limiting factor for the utility of mobile robots, both in commercial applications and especially in industry, where automation efforts are stifled by tethered or static power delivery.

To address this, we are developing a framework for adaptive and expandable wireless charging. Our goal is to create a physically flexible, modular wireless charging mesh that integrates effortlessly onto any existing surface, providing power to multiple robots in motion.

The project presents several key challenges: miniaturizing power-delivery circuitry, developing close-to-metal power control algorithms, and designing networks of flexible charging coils.

Beyond mobile robots, the potential applications of this project are far-reaching. Imagine electrified roads that continuously power electric vehicles, mobile medical equipment in hospitals that autonomously moves to where it's needed, and ultimately, a future where any surface can become a power-delivery platform.

A skyrmion is a topological soliton that appears in some magnetic materials. It is small in size, highly stable, and follows a vortex-like configuration of magnetic moments, known as spins, which can be stabilized through magnetic interactions. These configurations can be generated, transferred, and eliminated, and opening doors for potential applications in spintronic devices and data storage.

Recent literature presents that many research studies have focused on materials for better control while paying less attention to skyrmion measurements. Studies have also highlighted that skyrmions oscillate at THz frequencies. Consequently, the OpSky project aims to develop an optical sensor using THz technology to measure skyrmions, offering advantages such as flexibility, non-contact measurement, and ultrafast measurement ability.

With the rapid growth of the Internet of Things (IoT), vast amounts of time series data are continuously generated by different devices. As more devices are deployed, managing and efficiently transmitting this data becomes increasingly challenging. Traditional methods, which rely on data compression to reduce transmission loads, focus on preserving the original data for reconstruction. However, this approach often does not align with modern IoT applications, where transmitting task-relevant information is prioritized over full data recovery, requiring innovative strategies.

This project addresses these challenges by developing a system that intelligently identifies and transmits only the most relevant information for specific tasks, a concept known as goal-oriented semantic communication. By focusing on essential data, especially within time series formats, this approach enhances transmission efficiency and minimizes unnecessary data flow. Additionally, by reducing the transmission of sensitive raw data from devices, this method strengthens data privacy and security. By integrating machine learning for more effective semantic extraction, this project aims to build a secure and adaptable IoT framework that can meet the complex communication demands of modern applications.

My PhD project centers on advancing machine learning techniques to enhance early diagnosis and monitoring of Parkinson's disease (PD) by analyzing sleep patterns using a minimally invasive device called MiniPSG. This innovative device allows for continuous, at-home sleep monitoring without the need for direct supervision by healthcare professionals and thereby allowing for large-scale data collection.

The project specifically aims to identify and track subtle changes in sleep behavior that are indicative of early stages of PD, particularly in patients who have not yet developed noticeable symptoms. By leveraging large datasets of longitudinal sleep recordings, I aim to develop algorithms that can reliably detect these early markers and monitor disease progression over time.

This research holds the potential to significantly improve early diagnosis and intervention strategies, offering a more accessible and cost-effective way to manage PD. Ultimately, this work could pave the way for earlier and more precise treatment, potentially delaying or even preventing the onset of debilitating symptoms.

The project is carried out in close collaboration with the Department of Nuclear Medicine & PET Center at Aarhus University Hospital.

All systems which contain software are built on top of layers of abstraction. The abstraction layers can vary in level, but they are always present. Well-designed abstractions are essential for the performance and portability of any system. The capabilities of modern hardware increase every year, both regarding computation speed and the available features of the chips, and the full utilization of these features is vital for increasing the performance and capabilities of systems. This continuous improvement in hardware capabilities requires that the hardware abstraction layer (HAL) is designed with this in mind. This PhD project aims to research a method for extending and enhancing HALs in a way that enables composing the existing hardware peripherals into new hardware accelerators. The proposed method for accomplishing this feat is to create a domain specific language (DSL) for this purpose. This approach allows us to ensure correct usage of hardware peripherals as well as providing certain safety and correctness guarantees by using rigorous techniques. Incorporating this functionality as an extension of the HAL allows the composed accelerators to be ported across different platforms and reused in multiple projects reducing development time while increasing testability, safety, correctness and robustness. 

This project focuses on developing cost-effective and efficient silicon solar cells by replacing silver (Ag) contacts with copper (Cu) through an innovative process called electroplating. By combining this process with black silicon (Si) technology and advanced passivation techniques, the project aims to enhance silicon solar cell efficiency while reducing manufacturing costs. The goal is to produce high-efficiency silicon solar cells that are more environmentally friendly by eliminating the use of lead and reducing the carbon footprint of the production process. This new approach also aims to achieve better light absorption and lower reflectance, making the silicon solar cells more suitable for large-scale production and applications like building-integrated photovoltaics (BIPV). The project combines expertise in solar energy systems, semiconductor physics, and electroplating to innovate in the field of photovoltaic technology.

This project aims to create a comprehensive classification system called a Taxonomy, to organize
the field of “effort estimation” in software engineering, which is the process of predicting how
much time and resources are needed to complete a software project or task.

By developing a gold standard taxonomy for Software Effort Estimation, we will create a clear and
organized way to categorize all the key factors involved in this area. This helps everyone in the
field use the same terms, understand complex relationships more easily, and identify gaps in
research or practices that need improvement.

During the course of this PhD we will study diverse visualizations techniques to allow for an
immersive understanding of taxonomies. Additionally, we will use artificial intelligence and natural
language processing to analyze and organize information from past studies on software effort
estimation. This AI-driven approach will provide clearer insights and better decision-making
support for both researchers and practitioners. The focus is to make the results easy to
understand and apply to real-world software projects.

This project aims to develop novel techniques for distributed AI learning and inference in resource-constrained environments, and in edge-cloud computing continuum. Given the limitations of IoT devices and embedded systems in terms of memory, processing capacity and energy, we will design methods to intelligently split AI tasks across multiple devices. The goal is to improve the inference and learning speed, scalability, and energy efficiency of AIoT systems, while maintaining accuracy.

We will optimize resource usage while meeting accuracy and latency requirements, enabling AI models to function efficiently even in dynamic computation and communication environment. The developed methods will be applied to real-life use cases in industry. The outcomes of this project will make advanced AI more feasible for real-world applications, especially on low-power devices in distributed environments.

This project aims to develop new Artificial Intelligence (AI) models to allow their integration in environments with constraints over computational resources or time.

In particular, continual AI models offer a more efficient way to process sequence, with little to no loss in performance in comparison to their non-continual counterparts. Moreover, there are other methods to create even more light models, such as dimensionality reduction or low-rank approximation. All of this could be complemented with other techniques to improve performance with no increase in computational cost, such as making a clever processing of the data available.

The developments of the project will allow a reduction of the energy footprint of AI models and to make them a feasible option for a wider range of cases for Artificial Intelligence of Things.

Today digital development is becoming increasingly more important in industrial applications. However, the systems engineering process of companies —especially those with a long history— may not originally have been designed with this tendency in mind and may be inherently suboptimal as a result. Additionally, many such companies are hesitant to adopt new technologies or radically change their existing processes due to a variety of reasons. A hot topic such as artificial intelligence (AI) is a prime example of this. AI-based tools boast significant potential for increasing productivity and streamlining processes across various domains. Nevertheless, concerns related to preserving intellectual property, data privacy, and legal implications, among other factors, hinder the adoption of such tools.

This project explores various techniques for improving the systems engineering process. Specifically, the goal is to increase traceability, transparency and possibly automation by linking various internal data models, tools and design steps using a Digital Thread. By establishing a communication flow between traditionally siloed elements, the hypothesis is that real-time insights from the data sources can enhance the decision-making process and the collaboration between teams.

The project aims to develop advanced deep learning models to analyse and predict the behaviour of individual investors in financial markets. This project addresses the limitations of conventional (typically statistical) financial models, which often lack empirical observations from real-world markets. A data-driven approach will be employed, utilising various deep learning techniques both independently and in combination to analyse large-scale financial data, with a focus on investor behaviour and market interactions.   

One of the objectives of the project is to integrate data from diverse sources to produce more accurate predictions and insights into investor behaviour and its evolution over time. The project is funded by the Research Council of Denmark. 

During their lifetime pumps experience wear and tear that can impact their performance.
Detecting these issues in a timely manner is crucial to avoid steep maintenance costs. Additionally,
understanding the impact on performance parameters such as flow and pressure is important for
efficient pump control. Presently, these conditions are either not monitored at all or rely on
expensive external sensors.

To address this issue, this project aims to investigate the feasibility of using low-cost current
sensors directly installed on the motor controller to complement or even replace the data
collected by existing sensors. By analyzing the data captured by these current sensors during
pump operation, the project seeks to determine if any signals related to conditions or
performance parameters can be estimated.

My PhD project aims to improve the efficiency and sustainability of electronic devices by focusing on Wide Bandgap semiconductors, which are materials that allow devices to operate faster and use less energy than traditional materials. These semiconductors are already helping to make power electronics more efficient, but there is still room to optimize them further.

To achieve this, I am using Quantum Machine Learning, an innovative method that can help fine-tune the way these materials operate. Specifically, I am working to reduce the size of components like capacitors and inductors, which are used to store energy in power circuits. By making these components smaller, we can build more compact, energy-efficient devices.

Another key goal of this project is to remove the need for additional sensors that monitor voltage and current in electronic systems. This will make devices simpler, cheaper, and more reliable. I will test these improvements on a wireless power charger to demonstrate how much energy, space, and cost can be saved by these methods.

“DONUT” is a project funded by the European Union. Its primary objective is to enhance the education and training of doctoral candidates, particularly in the fields of Brain-Computer Interface (BCI) and Electroencephalography (EEG) technologies.

The project's mission is to establish a multidisciplinary and inter-sectoral Doctoral Network (DN) for talented researchers, specifically doctoral candidates (DCs), to equip them for research in EEG-based BCI applications. These BCI tools can serve various purposes, including communication, signal analysis, applications in healthcare (rehabilitation, neural prosthetics, diagnostics), industrial applications, and even entertainment (Virtual Reality, biometrics).

Aarhus University will develop a generic and low-power solution for continuous impedance measurements in wearable EEG devices for user-friendly everyday use. As a part of this development, an integrated circuit suitable for high-impedance measurements must be designed.

In diseases causing blindness by photoreceptor degeneration, such as retinitis pigmentosa and age-related macular degeneration, the neuronal circuit in the eye remains intact and functional. Two approaches to restore visual function are photovoltaic (PV) retinal implants and optogenetics (OG).
The PV implants are micro-scale solar cell units which when illuminated produce a small current. The neuroretina consists of neuron layers in the retina that propagate the signal from the photoreceptors to the visual cortex in the brain. When implanted in close proximity to the neuroretina the currents of the PV implant activate the neurons locally, thereby mimicking the photoreceptor signals, which then can be perceived as an image in the brain.

In OG viruses are used to deliver DNA encoding light-sensitive proteins to the neuroretina, often targeting the retinal ganglion cells. The neurons can then be stimulated directly by illumination as the light-sensitive proteins induce their activation.
Both PV implants and OG have been tested separately, but in this novel project we aim to combine them to restore visual function further than the two approaches can discretely. To achieve this aim micro solar cells and viruses are designed and produced and tested individually in ex vivo experiments with pig eyes. Subsequently, animal studies will be performed to assess the combination treatment in vivo in a disease model.
Furthermore, the project aims to develop a controlled drug delivery method for the optogenetics component that ensures safety, quantity, and area of delivery.

The significance of improving the sustainability of buildings has been increasing rapidly over the recent years. While it is important to analyze the technical aspects that might be related to the impacts of buildings and neighborhoods on our environment, it is crucial to consider the social criteria in the context of sustainable building environment.

In this project, we investigate the importance of social criteria in the building environment.

In addition, we develop and implement a decision support software solution in order to provide an assessment methodology. This methodology can verify the satisfaction of social aspects that directly affect the well-being of the buildings’ occupants such as indoor air quality, visual and thermal comfort.  

This research will be carried out on actual buildings that are part of the newly renovated AU living lab, as an integration with the EU-funded PROBONO project, which aims to provide strategic planning tools from technical, environmental, economic and social perspectives to create recommendations and standardization actions for the European construction industry.

To design, implement and evaluate secure data sharing and controlled, privacy-preserving processing of IoT data using the cloud-to-thing IoTalentum infrastructure, trusted execution environments (e.g., Intel SGX), local computing resources, and judicious allocation of data. To investigate compression techniques that allow for processing directly on compressed data as a way to reduce data traffic and memory usage, both in local, MEC, and cloud devices in order to address the explosion in generated data worldwide.

The number of hearing-impaired individuals is growing rapidly on a global scale, due to the increase of the elderly population. Hearing impairment can make it difficult to communicate with the world around us, and it can be so difficult that certain social situations are avoided, which can lead to isolation, cognitive decline and even depression. Therefore, it is no great surprise that how well a person can understand speech (speech intelligibility) has been a high priority when developing and fitting hearing aids. The golden standard for determining speech intelligibility is a behavioral test, where the hearing-impaired individual is introduced to speech elements and repeats back what is heard. This is not only a highly time-consuming process, but also a very subjective procedure, that has shown not to always correspond to the users experience in their daily lives.  

Can we measure speech intelligibility using electroencephalography?
This project aims to investigate whether it is possible to measure how well speech is understood using electroencephalography (EEG). If this is possible the immediate benefit will be obtaining a new objective measure to evaluate whether for example a hearing aid improvement is enhancing speech intelligibility in a lab setting. The next natural step will be to investigate whether it is possible to measure speech intelligibility using ear-EEG. Ear-EEG is a device used for recording electrophysiological signals using electrodes placed inside the ear. The benefit of using the ear-EEG is that it is possible to measure EEG in the natural environment of the user, in an unobtrusive and mobile manner. This hopefully resulting in a better speech intelligibility for the end user, and thereby closing the communication gap caused by their hearing impairment.

Physicians are typically concerned with sounds originating from the body, which are traditionally listened to through an acoustic stethoscope. However, interpretation of sounds through such instruments is subject to the training of the individual physician. To overcome this challenge and further advance the method, an electronic version of the stethoscope has been invented, allowing for more objective interpretation of such sounds.

Ear-EEG is a technique specifically designed to monitor brain activity during everyday activities. Through an ear-fitted device, physiological signals reflecting the subject’s brain activity are measured.

This project aims to extend the ear-EEG platform with a body-coupled microphone, i.e. an electronic stethoscope in the ear, enabling recording in real-life environments with a discrete and unobtrusive wearable device. Furthermore, we seek to explore methods for joint processing of ear-EEG signals and sounds picked up from the body-coupled microphone . This would allow for a broader understanding of the current state of the body and is highly applicable in both research and in medical devices for health monitoring.

This project is carried out at the newly founded Center for Ear-EEG