| Topic title |
Possible scientific supervisors |
Source of funding |
|
Dynamic Forest Carbon Sequestration Assessment
|
prof. dr. Rytis Maskeliūnas |
state-funded |
Research Topic Summary.
Recent developments in the European Union’s climate policy have placed significant emphasis on improving the accuracy of carbon accounting in line with the ambitious carbon neutrality goals under the European Green Deal. Forests, as vital carbon sinks, play an important role in this strategy, especially within the Land Use, Land-Use Change, and Forestry (LULUCF) regulations. However, the challenge remains in calculating forest carbon sequestration with sufficient precision to meet the evolving requirements of EU policies and the upcoming taxation, which call for near-real-time, transparent, and scalable methodologies. Traditional static models for carbon stock estimation struggle with the complexity and variability of forest ecosystems and climate change impacts. AI with its capacity to handle vast datasets and capture non-linear relationships, offers a promising solution to overcome these limitations.
This thesis will aim to develop state-of-the-art AI driven techniques to transform forest carbon sequestration assessments, using dynamic modeling approaches that integrate remote sensing data, climate variables, and forest management practices. The to be developed predictive models should be capable of processing high-resolution satellite imagery, LiDAR data, and environmental variables to improve the accuracy of carbon flux measurements across diverse European forest ecosystems. These models should not only estimate current carbon stocks but also predict future sequestration potential based on scenarios of land-use change, deforestation, and forest regrowth. The use of explainable AI is required to make these models interpretable for policymakers and aligning them with the stringent transparency and accountability criteria set by the EU.
|
|
Development and research of a hybrid optimisation method for robot control using an AI platform
|
prof. dr. Renaldas Urniežius |
state-funded |
Research Topic Summary.
If you're passionate about advancing AI platforms and creating hybrid models for robotics and automation, we invite you to begin your PhD journey with our team and contribute to pioneering research at the cutting edge of technology.
|
|
Model and Research for Assessing Cyber Resilience in Digital Ecosystems
|
prof. dr. Šarūnas Grigaliūnas |
state-funded |
Research Topic Summary.
The topic of this dissertation focuses on advancing cyber resilience strategies to protect digital ecosystems from increasingly complex cyber threats. Considering the ongoing digitalization and interconnectedness of organizations, businesses, and government structures, this research aims to develop and apply advanced resilience strategies that enable more effective resistance and recovery from cyber incidents. The research outcomes will contribute to creating a safer digital ecosystem landscape and help ensure the long-term security and continuity of digital infrastructures.
|
|
Machine learning and information integration research for disease prediction and health care
|
prof. dr. Robertas Alzbutas |
state-funded |
Research Topic Summary.
The objective of the research is the creation of methodology and test calculations based on machine learning and information integration for the most accurate disease prediction and health care.
Tasks:
1. Review and compare possibilities of information integration methods and algorithms as well as the application of related software and machine learning methods intended for health care.
2. Define information integration and smart systems accuracy criteria and their evaluation procedures in the context of data relevant to health care.
3. Develop and demonstrate the information integration tools whose usage increases the accuracy and/or reduces the risk of incorrect decision-making.
4. Perform the pilot studies of smart systems intended for health care and create a methodology for the effective application of these systems.
For further information please contact the supervisor of the topic.
|
|
Smart Forest Growth Monitoring using Federated Learning
|
prof. dr. Rytis Maskeliūnas |
state-funded |
Research Topic Summary.
As forests become increasingly impacted by climate change and anthropogenic activities, real-time monitoring of forest ecosystems has emerged as a critical tool for managing forest health and preventing diseases. We now have IoT sensors, capable of monitoring key tree parameters such as soil moisture, tree sap flow, air temperature, humidity, and surrounding environmental factors.
This thesis will focus on developing a robust AI-based framework for processing and analyzing data streams from IoT-enabled smart forest monitoring systems. The study will integrate data from sensors tracking soil conditions, tree physiology, meteorological data, and environmental parameters, all to create a holistic view of forest ecosystem dynamics. Thesis encourages to use unsupervised learning to forecast critical forest events. For example, early detection models can be built to predict forest growth decline based on a combination of drought conditions, temperature anomalies, and soil moisture levels, while disease outbreak models can analyze changes in tree sap flow and leaf moisture content as precursors to potential pathogen spread.
One of the key objectives of this research will be the creation of a scalable system that not only forecasts events but provides actionable, XAI insights for forest managers. The models should exploit subtle patterns and correlations between environmental changes and forest health, enabling real-time alerts and decision-making support for proactive intervention.
Moreover, this study will investigate how federated learning techniques can be employed to enhance model training while ensuring data privacy and security, especially when IoT data comes from multiple sources and regions, allowing AI models to potentially learn from distributed data sets across various EU forest ecosystems, improving the generalization and accuracy of predictions while respecting local data governance laws imposed by standing EU acts.
|
|
Integrating Explainability in Vision-Language Models: Methods and Applications for Multimodal AI Transparency with Application in Furniture Manufacturing and Surface Defect Analysis
|
doc. dr. Arūnas Lipnickas |
state-funded |
Research Topic Summary.
Artificial Intelligence (AI) and machine learning have transformed industries by enabling process automation and improving decision-making. In the furniture manufacturing sector, advanced AI models have shown great potential in quality control, especially in surface defect analysis. Using vision-language models, which integrate visual and textual data, manufacturers can automate the inspection process and detect defects like scratches, dents, and stains on wooden surfaces. However, the "black-box" nature of these models poses a challenge: the lack of transparency and explainability.
Modern vision-language models such as CLIP (Contrastive Language-Image Pre-training) and BLIP (Bootstrapped Language-Image Pretraining) have demonstrated strong capabilities in jointly understanding visual and textual inputs. In industrial applications, these models can be employed for defect localization and classification.
Interactive AI systems allow users to query the model for explanations or request additional information about detected defects. This approach leverages the capabilities of multimodal models to enhance user understanding and decision-making. Currently, there is no standard metric for evaluating the quality of explanations provided by vision-language models. What constitutes a "good" explanation can be subjective and depends on the user's expertise and the application context. Developing objective metrics to assess the clarity, usefulness, and fidelity of explanations will be a key challenge.
Integrating explainability into vision-language models is essential for their adoption in critical applications, such as surface defect analysis in furniture manufacturing. By utilizing advanced explainability methods, we can bridge the gap between complex AI models and users, fostering trust and ensuring more transparent decision-making.
|
| Secure Service Composition and Orchestration in Fog Computing |
prof. dr. Algimantas Venčkauskas |
state-funded |
|
Investigation of a self-driving robotic system for manipulating objects moving along a conveyor
|
prof. dr. Vidas Raudonis |
state-funded |
Research Topic Summary.
This research focuses on the integration of robotics and automation in industrial settings, specifically the development of self-driving robotic systems designed to manipulate objects moving along conveyor belts. The primary objective is to enhance efficiency and productivity within industrial processes by automating repetitive tasks and reducing labor costs. The research will explore the technological advancements and challenges associated with developing such systems, including real-time object detection and tracking, motion planning and control, and safe human-robot interaction. By successfully implementing these self-driving robotic systems, industries can streamline their operations, improve product quality, and achieve a higher level of operational efficiency.
|
|
Signal processing and machine learning for monitoring and risk analysis of battery heterogenic ageing
|
prof. dr. Robertas Alzbutas |
state-funded |
Research Topic Summary.
More accurate estimation of battery State of Charge (SOC) and State of Health (SOH), as well as optimization of charging cycles and overall battery usage, could benefit from decision support tools that better (faster and more precise) utilize and translate the information held in signals coming from Battery-Management System (BMS).
The objective of the research is to develop a method and system, used to support BMS, able to autonomously monitor battery SOH and estimate the risk of different types of malfunctions and which will provide easy-to-interpret recommendations and decisions for battery usage.
Tasks and expected results:
1. Data collection and development of deep neural network architectures for SOH monitoring (identification, characterization, and classification) and dealing with data heterogeneity as well as time dependence.
2. Application and testing of developed models and risk analysis approach for the case of various data (e.g. sample of annotated BMS data), taking into consideration data transformation optimization and calibration results.
3. Software development, for deep learning of SOH monitoring, when taking into consideration data heterogeneity and risk analysis. The software will be released under the GNU General Public License for the free usage of interested parties.
For further information on the topic, and relevant R&D work please contact the supervisor.
|
