AI at work in Safety-critical Remote Operations

PhD candidate: Karthick Deivasagayam
Supervisor: Eric Monteiro
Co-supervisor: Vidar Hepsø
Co-supervisor: Elena Parmiggiani

In the energy industry (such as wind, oil and gas, utility), safety-critical remote operations are increasingly conducted from remote locations such as control rooms. Remote operations rely on real-time sensor data feeds that combine AI-based (meta)data visualization and detection systems to help operators identify the relevant information and make efficient decisions. Due to safety and reporting concerns, it is important that human operators remain in control. To ensure safe and efficient decision-making, there is a need to develop a better, in-depth understanding of these complex human-AI hybrids, as well as explore improved approaches to sensemaking and training based on e.g. Extended Reality (XR)-based serious games that help prevent the “data drowning” problem. The goal of Kathrick's PhD project is to contribute investigating how AI-infused remote operations can be implemented to improve decision-making processes in organizations. The research design is inspired by Design Science Research, in which empirical data collection will be alternated with prototyping and experiments to test the developed hypotheses. Kathrick will be involved in the Digital Enterprise strategic research area at the faculty of Information Technology and Electrical Engineering, and in the BRU21 multidisciplinary programme which will coordinate design and access to empirical cases and partners from the energy industry.

Developing Recommendations for Trustworthy Methods in AI Applications

PhD candidate: Lena Jedamski
Supervisor: Kerstin Bach
Co-supervisor: Andreas Hafver
Co-supervisor: Inga Strumke

Building trustworthy AI systems is a cornerstone to apply AI technologies in practice and therefore we need to explore methods, build tools, and incorporate different perspectives when developing novel AI applications. Based on the definition of the High Level Expert Group of the European Union, trustworthy AI should be lawful, ethical, and robust. While guidelines exist, research on implementation methodologies are currently under development and this research will contribute to develop principles for creating trustworthy AI applications. Together with partners in the NorwAI SFI Lena will work on the creation for guidelines for a sustainable and beneficial use of AI, explore privacy-preserving technologies and create explainable, interpretable and transparent prototypes to be tested in industrial settings.

BigData and AI for Future Mobility Solutions

PhD candidate: Oluwaleke Umar Yusuf
Supervisor: Adil Rasheed
Co-supervisor: Frank Lindseth
Co-supervisor:

The aim of the PhD will be to develop and exploit big data and artificial intelligence-driven digital twin of urban mobility infrastructures to solve challenges in achieving a carbon-neutral mobility future. A digital twin is defined as a virtual representation of a physical asset or process enabled through data and simulators for real-time prediction, optimization, monitoring, control, and improved decision making. The PhD will work towards developing enabling technologies to instill physical realism in such a digital twin. The enabling technologies will consist of data acquisition, pre-processing, fusion, and postprocessing techniques using an array of physics-based, data-driven, and hybrid models. In addition, the work will also involve the development of tools for communicating insights in a way that facilitates informed public opinion building and decision-making.

Statistical Machine Learning in Distributed Acoustic Sensing to Automate Event Detections

PhD candidate: Khanh Truong
Supervisor: Jo Eidsvik
Co-supervisor: Robin André Rørstadbotnen
Co-supervisor: Jan Petter Morten

Khanh will be conducting statistical research on spatial and spatio-temporal processes that can be used for extracting relevant information from various types of geophysical data. The study focus of his PhD project is on the development of new statistical methodologies for analysing geophysical data. Such data include seismic and electromagnetic measurements, as well as newer data types such as fiber-optical sensing data. The project will also focus on using statistical methodologies for research developments in the centre. Core components of his PhD project include Bayesian hierarchical modeling and methods for conditioning in such models. He will aim for a coherent interpretable modeling approach honoring the large-size spatio-temporal data. Khanh will work on the statistical foundations for such frameworks.

Hybrid Machine Learning Approaches for Explainable Reservoir Inflow Forecasting

PhD candidate: Rahmathulla Madathi Parambath
Supervisor: Jayaprakash Rajasekharan
Co-supervisor: Hossein Farahmand
Co-supervisor:

In Norway, hydropower is the primary source of electricity, and it accounts for 96% of electricity generation. Due to the long-term reservoir storage constraints in hydropower dominated systems, the resource scheduling should be done for a sufficiently long-time period and with an appropriate representation of uncertainties (such as inflow to reservoirs, wind power generation, demand etc.). Existing hydropower scheduling models face many shortcomings when dealing with such uncertainties. The IntHydro project aims to develop and demonstrate a new hydropower scheduling tool based on machine learning techniques, which manages the hydropower plants more efficiently and effectively through optimizing water resource management and multi-dispatch between hydropower and variable renewable energy sources. In this context, the main task of Rahmath in the doctoral work will be two-fold. First, to integrate machine learning techniques in different stages of hydropower scheduling models and define comprehensive coupling principles between the strategic and operational modelling. Second, to develop a prototype for fundamental hydropower scheduling that allows modelling of RES on a detailed time scale.

Natural Language Processing: Personalized Content Summarization

PhD candidate: Vandana Yadav
Supervisor: Jon Atle Gulla
Co-supervisor: Lemei Zhang
Co-supervisor: Özlem Özgöbek

Content summarization uses machine learning to extract the main ideas of individual texts or collections of text. It may also be used to explain recommendations and provide explanatory features in artificial intelligence in general. Central to the technique is the use of machine learning and language models to interpret text and contextualize content with respect to users or tasks at hand. Summarization is often combined with other NLP techniques to provide personalized summaries or build conversational systems that make use of external sources. Vandana will develop new summarization techniques that are incorporated into generative conversational systems and personalized with respect to users and contexts. The research will address both English and Scandinavian languages. The expected results include industrial prototypes in collaboration with industry partners as well as high-quality publications in top NLP venues.

Predictive digital twin for offshore wind farms

PhD candidate: Evi Elisa Ambarita
Supervisor: Agus Hasan
Co-supervisor:  Anniken Susanne T. Karlsen
Co-supervisor: Francesco Scibilia
Co-supervisor: Frank Lindseth
Co-supervisor: Steffan Sørenes

The idea of this project is to leverage the use of AAS for digital twin development of offshore wind farm. Evi will investigate current applications of AAS in other industries that could be relevant for wind farms, possibility to transfer AAS learning and solutions from other industries. In general, this task should look at the opportunities of using the AAS framework in the offshore wind industry and highlight possible challenges for implementation, which include architecture design, data requirements, predictive analytics, information modeling, and simulation and visualization solutions for asset management and predictive maintenance. Furthermore, Evi will implement prototypes of simulation and visualization solutions based on Digital twin (AAS-based). Use cases will be provided from the Hywind Tampen wind farm operated by Equinor.

Autonomy and Simulation for a Sustainable and Carbon-Neutral Mobility Future

PhD candidate: Florian Wintel
Supervisor: Frank Lindseth
Co-supervisor:  Gabriel Kiss
Co-supervisor: Adil Rasheed

Florian's PhD project will investigate both modular (mapping and localization, perception and prediction, planning and control) and end-to-end (imitation and reinforcement learning) approaches to autonomous driving in a Nordic environment. Training and validation of autonomous agents (from shuttle busses to last-mile delivery robots) will be done in simulated environments (e.g. CARLA, NVIDIA DRIVE Sim) of a HD-map / Digital Twin representation of the area in question, as well as in the real-world environment using our in-house full-scale research platform for autonomous driving. Agents should learn to co-exist with humans in a real-world mobility setting. Current traffic patterns will be visualized and future “what-if” scenarios will be simulated (e.g. NVIDIA Omniverse) before physically constructing the optimal solution for a given area.

From Challenges to Solutions: Enhancing Active Travel in Hilly Cities through City Bike and Bus Integration in Trondheim

PhD candidate: Muhammad Tsaqif Wismadi
Supervisor: Yngve Karl Frøyen
Co-supervisor: Kelly Pitera
Co-supervisor: Adil Rasheed

Tsaqif's PhD research centres on "Enhancing Active Travel in Hilly Cities," with a primary focus on integrating city bike networks and public transport systems to tackle mobility challenges unique to hilly urban environments. Utilizing advanced spatial analysis and machine learning models, his work explores the factors influencing city bike ridership and develops strategies to optimize the placement of bike stations and bus stops for improved connectivity. By modelling and forecasting travel behaviour, the project aims to promote sustainable mobility by enhancing the accessibility and integration of public transit and active travel modes like cycling. Using Trondheim as a case study, this research provides data-driven solutions to support the development of more efficient and inclusive urban transport systems.

Measureing impact of sustainable transport initiatives on interactions of multimodal road users: Simulation-based modelling with empirical data considering digital twins

PhD candidate: Zakiya Aryana Pramestri
Supervisor: Trude Tørset
Co-supervisor: Kelly Pitera
Co-supervisor: Yngve Karl Frøyen
Co-supervisor: Kathrine Strømmen

Zakiya's research is part of a larger interdisciplinary project and lab, Mobility Elgeseter which aims to explore innovative and sustainable mobility solutions in an urban, campus environment. Transport models are used to replicate various aspects of the transport system, and to predict and evaluate the results of changes to the system. Transport models are used for analyzing strategic decisions about the transport system, either to achieve goals or to estimate impacts of suggested initiatives. The objective of the PhD projects is to improve the state of the art of urban transport models by capitalizing on increased digitization and big data. The research will explore how sources of dynamic data and passive data can be utilized to further develop and improve transport modelling. In addition to addressing increased digitization, the projects will aim to further developed urban transport models to address the shifting focus from car passenger transport to active transport and public, and the need to integrate freight transport with passenger transport within models. In addition to the theoretical research within the projects, there will also be a strong focus on the applied aspects of this research. This includes the development of transport models for the Elgeseter case study area which can be used to understand both the current mobility picture and future mobility scenarios. Finally, the projects will consider the integration of such models into an interactive platform for modeling and visualization in the 3D virtual environment of a digital twin, and consider how model inputs, results, and uncertainties can be better communicated to decision-makers.

Automated AI/Computer Vision Driven Generation of Digital Twins for Sustainable Mobility Infrastructure

PhD candidate: Sachin Verma
Supervisor: Gabriel Kiss
Co-supervisor: Frank Lindseth
Co-supervisor: Hongchao Fan

The main aim of Sachin's project is to build a static digital twin of the MobilityLab’s focus area that will act as a baseline for all further simulations and will be extended to be able to receive dynamic information from IoT sensors. An accurate digital representation of the focus area will be created, based on existing geometric information and high-resolution aerial photographs. Raw data is already available for the project either as public or previously acquired datasets, additionally during the project more raw-data will be acquired with an in-house full scale research platforms (Kia e-Niro with an NVIDIA DriveWorks stack as well as dedicated high-resolution surveying drones). The final aim of his project is full automation of the conversion process from raw data to a usable digital twin representation (i.e. AI based methods and handling of Big Data arriving in real-time from the research platforms). Furthermore, improving the model via automatic texturing and interactive editing will also be investigated (e.g. Instant NeRF or Block NeRF). The intended platform for the development of the digital twin will be NVIDIA Omniverse. models into an interactive platform for modeling and visualization in the 3D virtual environment of a digital twin, and consider how model inputs, results, and uncertainties can be better communicated to decision-makers.

Modeling of Integrated multimodal urban mobility: Integrating mass transit with on demand mobility options

PhD candidate: Zelalem Birhanu Biramo
Supervisor: Trude Tørset
Co-supervisor: Kelly Ann Pitera
Co-supervisor: Yngve Karl Frøyen
Co-supervisor: Kathrine Strømmen

Zelalem's research is part of a larger interdisciplinary project and lab, Mobility Elgeseter which aims to explore innovative and sustainable mobility solutions in an urban, campus environment. Transport models are used to replicate various aspects of the transport system, and to predict and evaluate the results of changes to the system. Transport models are used for analyzing strategic decisions about the transport system, either to achieve goals or to estimate impacts of suggested initiatives. The objective of the PhD projects is to improve the state of the art of urban transport models by capitalizing on increased digitization and big data. The research will explore how sources of dynamic data and passive data can be utilized to further develop and improve transport modelling. In addition to addressing increased digitization, the projects will aim to further developed urban transport models to address the shifting focus from car passenger transport to active transport and public, and the need to integrate freight transport with passenger transport within models. In addition to the theoretical research within the projects, there will also be a strong focus on the applied aspects of this research. This includes the development of transport models for the Elgeseter case study area which can be used to understand both the current mobility picture and future mobility scenarios. Finally, the projects will consider the integration of such models into an interactive platform for modeling and visualization in the 3D virtual environment of a digital twin, and consider how model inputs, results, and uncertainties can be better communicated to decision-makers.

Area of Radio Channel Measurements and Modeling in Maritime Scenarios

PhD candidate: Giacomo Melloni
Supervisor: Torbjörn Ekman
Co-supervisor: Kimmo Juhani Kansanen
Co-supervisor:

Giacomo will work on challenging research problems related to radio channel measurements and modeling, within the general area of radio-system design. The activities include Massive MIMO radio channel measurements in the Trondheim fjord, accompanied by channel modeling. With insights gained from measurements, the main objective of the research is to identify current solutions' weaknesses and suggest improvements, design concepts, and deployment strategies for large antenna maritime communication systems, in collaboration with SINTEF Digital and Telia. As part of the research activities, he will collaborate actively with some of the partners. Also, he will be affiliated with the IoT@NTNU and with the Norwegian Open AI lab, and will be encouraged to have a research stay with one of the collaborating universities outside Norway during the research period.

Goods transport including applications of dynamic/big data

PhD candidate: Irene Hofmann
Supervisor: Trude Tørset
Co-supervisor: Yngve Karl Frøyen
Co-supervisor: Eirin Olaussen Ryeng

Irene's positions are part of a larger interdisciplinary project and lab, Mobility Elgeseter (link) which aims to explore innovative and sustainable mobility solutions in an urban, campus environment. Transport models are used to replicate various aspects of the transport system, and to predict and evaluate the results of changes to the system. Transport models are used for analyzing strategic decisions about the transport system, either to achieve goals or to estimate impacts of suggested initiatives. The objective of the Irene's project is to improve the state of the art of urban transport models by capitalizing on increased digitization and big data. The project will explore how sources of dynamic data and passive data can be utilized to further develop and improve transport modelling. In addition to addressing increased digitization, the project will aim to further developed urban transport models to address the shifting focus from car passenger transport to active transport and public, and the need to integrate freight transport with passenger transport within models. In addition to the theoretical research within the projects, there will also be a strong focus on the applied aspects of this research. This includes the development of transport models for the Elgeseter case study area which can be used to understand both the current mobility picture and future mobility scenarios. Finally, the project will consider the integration of such models into an interactive platform for modeling and visualization in the 3D virtual environment of a digital twin, and consider how model inputs, results, and uncertainties can be better communicated to decision-makers.

Machine Learning for Maritime Communication System Performance Prediction

PhD candidate: Manju James
Supervisor: Kimmo Juhani Kansanen
Co-supervisor: Torbjörn Ekman
Co-supervisor:

The radio propagation channel behavior, including its statistical variation in time and space, depends on the location of transmitter and receiver, leading to location-dependent achievable data rates and reliability. The maritime channel, in addition, will depend on weather conditions that may change the propagation environment (sea) geometry. Manju's work studies the use of data-driven, machine learning trained spatial models that can be used to predict the achievable rates and reliabilities of the maritime channel, and the related radio system. The data sources foreseen are the radio system channel quality indicators at the physical layer, the location information, and eventually other data sources that can be used to improve the prediction accuracy (e.g. weather information).

Ship-Shore Radar Network

PhD candidate: Lukas Herrmann
Supervisor: Egil Eide
Co-supervisor:  Edmund Førland Brekke
Co-supervisor:  Andreas Brandsæter

The key ambition of the project will be to develop an advanced real-time network of shore based and ship-based maritime radars to assist monitoring and guidance of autonomous ships. The Trondheim fjord was appointed as the world’s first accredited test area for autonomous ships in 1996. As a part of building this research facility, NTNU and SINTEF will establish a network of shore-based maritime radars to be applied both for monitoring and guidance of autonomous vessels and for research application within the fields of radar scattering of sea surface, radar detection and tracking, and guidance/control. The radars will be connected in a real-time network using 5G or similar broadband radio systems. In addition, the network architecture will allow for real-time interchange of radar data between ships and ship-shore. All data will be jointly processed and stored in a shore-based server. Resulting target tracks will be distributed to all ships in the area to enhance the situation awareness for both manned and unmanned vessels.

The research and project work will put emphasis on developing efficient signal processing and real-time communication architecture for exchange of data between radar sites and users. Activities and goals under the PhD-work will be:
• System studies and system architecture of the radar data processing chain and radar data distribution system with low latency.
• Developing advanced algorithms for radar detection, sea and rain clutter filtering, and target tracking based on low level data fusion from a multitude of ship-borne and shore-based radars.
• Design and specifications of the access and network protocols, in terms of satellite link characterization for availability, capacity and reliability.
• Experimental testing and verification of the network with real vessels in the fjord. The goal will be to demonstrate the applicability of a complete radar network both for maritime operations and for general research work.

Ultra-Low Power Integrated Circuit Design for Self-powered Internet of Things

PhD candidate: Muhammad Umer Khalid
Supervisor: Snorre Aunet
Co-supervisor:  Trond Ytterdal
Co-supervisor:  Magnus Själander
Co-supervisor:  Rakesh Kumar

We need ultra-low power IoT devices that can harvest their energy from their surroundings and operate without a battery to accelerate large-scale deployment of IoT devices and the advent of a sustainable smart world (smart health, cities, transportation, etc.). Subthreshold circuits have been shown to achieve power and energy savings of up to orders of magnitude as compared to conventional of the shelf components, but extreme vulnerabilities to process-, voltage-, and temperature variations on chip still hinder widespread adoption in industry. The low supply volage at which subthreshold logic operates results in low switching speeds, which sometimes cause limitations. It is therefore foreseen that IoT devices require both subthreshold as well as conventional logic to meet both their energy and performance requirements. The objective of this PhD project is to investigate design principles for ULP devices. Umer will investigate Ultra-Low Power Internet of Things (ULP-IoT) devices that can collect sensor input and perform computational tasks, monitoring energy levels and waking up additional parts of the system when, e.g., more performance or wireless communication is required, and enough energy is available to the system.

Acoustics of fiber optic cable sensing for monitoring

PhD candidate: Josephine Nell Schulze
Supervisor: John Robert Potter
Co-supervisor: Pierluigi Salvo Rossi
Co-supervisor:

Josephine will explore novel applications of fibre optic cable sensing for environmental monitoring. This includes using DAS for passive detection of ships and marine mammals for deconfliction using AIS, and perhaps also exploring the potential for AUVs to communicate over long distances while submerged using DAS as a one-way acoustic channel to shore.

Edge Intelligence for Dynamic Digital Twins

PhD candidate: Omkar Bhoite
Supervisor: Kimmo Juhani Kansanen
Co-supervisor: Stefan Werner
Co-supervisor:

A digital twin is defined as a virtual representation of a physical asset or process enabled through data and simulations for real-time prediction, optimization, monitoring, control, and improved decision making. For real-time operation, the twin requires continuous updating with new measurements from its observation platforms. However, different sensors produce significantly different amounts of data, with different time-criticality and importance for the twin. Omkar willT focus on the coupling between the twin and the different kinds of sensors, and will work with pre-processing, compression, optimization and communication strategies to support accurate and timely twin operation while transmitting least data possible. Machine learning and AI-based strategies that learn and adapt to the characteristics of the sensors, and the requirements of the twin will be of special interest.

Lattice-based cryptography and homomorphic computation

PhD candidate: Cristian Alonso Baeza Miranda
Supervisor: Anamaria Costache
Co-supervisor:  Danilo Gligoroski
Co-supervisor:  Tjerand Silde
Co-supervisor:  Jiaxin Pan

Cristian will work on cryptographic algorithms and protocols in the general areas of privacy-preserving computation, post-quantum cryptography or lightweight cryptography.

Digital Twin Security Models and Mechanisms

PhD candidate: Jessica Barbosa Heluany
Supervisor: Vasileios Gkioulos
Co-supervisor:  Sokratis Katsikas
Co-supervisor:  Siv Hilde Houmb

Digital twins have evolved from passive monitoring and state estimation systems to integrated sociotechnical mechanisms that are essential for strategic modeling and planning, as well as operational real-time monitoring and control of cyber-physical systems. In order to support these functionalities, the operation of digital twins relies heavily on maintaining fidelity and synchronization with the production systems to which they are targeted. This is particularly important in safety-critical systems, from power station generators to manufacturing systems or production facilities in the oil and gas industry.

Secure infrastructure for cyber-physical ranges

PhD candidate: Vyron Kampourakis
Supervisor: Vasileios Gkioulos
Co-supervisor: Sokratis Katsikas
Co-supervisor: Habtamu Abie

Cyber Physical Systems constitute the core of Critical Infrastructures, yet their architectural and operational characteristics are not thoroughly captured by contemporary cyber ranges, which are commonly narrow in scope or purposefully aligned with subsections of specific target systems. The anticipated use of the physical reference environments investigated, modelled, and integrated in this work will be twofold, namely as demonstrators for education and dissemination, but also as testbeds for activities related to research and training.

Dynamic Interdependency Models for Critical Infrastructures

PhD candidate: Yana Bilous
Supervisor: Stephen Dirk Bjørn Wolthusen
Co-supervisor: Vasileios Gkioulos
Co-supervisor: Nils Kalstad

Graph and combinatorial models representing dependencies and interdependencies and flows of resources have been studied intensively also with a view to representing susceptibility of represented structures to attacks. However, we are interested to study highly sparse dynamic graphs and structures within these as well as efficient algorithms that allow the determination particularly of local properties. This is motivated by the application domains of modern infrastructure networks whose topology will usually be dynamic and may only be partially known; this would render insights into properties that would e.g. indicate vulnerabilities obsolete or wrong. The ability to obtain timely insights also for large networks would hence allow targeted responses and resilience mechanisms that are widely applicable.

Digital Twin and AI/ML Supported Threat Modeling Framework for Cyber-Attack Prediction and Projection

PhD candidate: Gizem Erceylan
Supervisor: Vasileios Gkioulos
Co-supervisor: Sokratis Katsikas
Co-supervisor: Sandeep Pirbhulal

Digital twins have evolved from passive monitoring and state estimation systems to integrated sociotechnical mechanisms that are essential for strategic modeling and planning, as well as operational real-time monitoring and control of cyber-physical systems. In order to support these functionalities, the operation of digital twins relies heavily on maintaining fidelity and synchronization with the production systems to which they are targeted. This is particularly important in safety-critical systems, from power station generators to manufacturing systems or production facilities in the oil and gas industry.

Secure, Human-Centered XR Experiences in Critical Sectors

PhD candidate: Camille Rose Sivelle
Supervisor: Katrien De Moor
Co-supervisor: Bian Yang
Co-supervisor: David Palma

Efforts to strengthen cybersecurity and resilience of critical sectors such as healthcare, smart districts, manufacturing,… go hand in hand with technological innovations and novel approaches and measures. For instance, the advanced use of AI and big data analytics leads to increased intelligence capabilities (e.g., for video-based or digital traces-based surveillance, remote monitoring). In addition, technologies enabling remote presence like Extended Reality (XR) are seen as a key enabler allowing to reduce physical human involvement, for instance in dangerous or life-threatening environments. XR technologies also pave the way for novel, immersive experiences in critical sectors (e.g., remote caregiving, …). While these innovations and technologies create new opportunities and enable a range of cybersecurity measures, they also often have an invisible flip side: their ethical implications (e.g., related to trust, transparency, control, privacy, digital inequality, bias, …) are to date still under-adressed.