New doctoral dissertation

May 2, 2025

VILNIUS TECH Library invites you to follow the published new dissertations. The dissertation „Data-driven method for increasing efficiency in road freight transportation“ prepared by VILNIUS TECH, Edvardas Liachovičius. The dissertation was prepared in 2020–2025. Scientific Consultant – Assoc. Prof. Dr Viktor Skrickij.

The dissertation was defended at the public meeting of the Dissertation DefenceCouncil of the Scientific Field of Transport Engineering in the Aula Doctoralis Meeting Hall of Vilnius Gediminas Technical University at 10 a.m. on 2 May 2025.

Large and medium-sized asset-based road transport companies have long-term growth strategies. The logistics business is cyclical, and companies, especially asset-based, must align their growth ambitions with market cycles. Growth during a cycle of decreasing demand for transportation services and declining prices can negatively impact businesses, leading to underutilisation of resources and significant financial losses. Conversely, when demand recovers, and transportation rates increase, there is often a delayed readiness – a shortage of human and other resources (e.g., to achieve the desired number of trucks). This situation can result in unused market potential, competitive losses, and the inability to achieve strategic goals on time. Business digitalisation brings an opportunity to develop novel technological solutions that may increase efficiency in the transportation sector. Efficiency in transportation companies is analysed in this dissertation, focusing on developing a freight rate and demand a prognosis method. By taking different approaches, the research offers a comprehensive subject analysis. It helps bridge the literature gap on demand and freight rate prognosis in road freight transportation. The study employs artificial intelligence-based and econometric models, including multivariate models, to comprehensively analyse the subject matter and address the problem. These models consider various factors influencing demand and pricing, such as economic indicators, accumulated practical experience and market trends. The analysis results provide valuable insights into demand prognosis and pricing in the road freight transport industry during significant market fluctuations. It can serve as a valuable reference for companies in the sector. Based on the investigation, a data-driven method for increased efficiency in road freight transportation has been developed, and the possibility of its integration into the company’s IT system has been analysed. The dissertation includes an introduction, three chapters, general conclusions and references. The main results of the dissertation were published in four scientific publications: two in journals referenced by the Web of Science database with Impact Factor and two in Conference Proceedings. Research results were presented during the international conference Transbaltica plenary session (2019) and conference Transport Problems (2024).
 
Doctoral dissertation readers can search via VILNIUS TECH Virtual Library.

Related news

Assoc. Prof. Dr T. Sledevic: „We aim to 'compress' the beekeeper's experience into an algorithm“
Assoc. Prof. Dr T. Sledevic: „We aim to 'compress' the beekeeper's experience into an algorithm“
Bees are increasingly settling not only in gardens or rural homesteads, but also in cities – on building rooftops, in small courtyards and parks. Here, their lives unfold right alongside people, yet observing what is happening at the hive entrance is not always easy. VILNIUS TECH researchers are looking for a way to do this using technology – they are developing a system that can detect changes that would normally only be recognised by an experienced beekeeper. In the face of technological progress, there are hardly any areas that have not been touched by innovation in one way or another. And although beekeeping remains firmly rooted in traditions cultivated over centuries, researchers are looking for ways to modernise the field. One of them is Assoc. Prof. Dr Tomyslav Sledevic from the Department of Electronic Systems at the Faculty of Electronics of VILNIUS TECH, who studies bee behaviour at the hive entrance and analyses it using smart technologies. [caption id="attachment_136651" align="alignnone" width="2560"] Tomyslav Sledevič[/caption] „It all started with a simple hypothesis: if an experienced beekeeper can understand the condition of the entire hive by observing bee behaviour at the hive entrance, then computer vision and deep learning methods should be able to automate this behavioural analysis,“ says the researcher. The goal – to develop a bee behaviour recognition system The origins of this research date back to 2018, when Assoc. Prof. Dr T. Sledevic began training convolutional neural networks to recognise pollen carrying bees as well as to look for the optimal balance between speed and accuracy. The research has now become much broader: the aim is to develop a bee behaviour recognition system that could operate effectively on compact embedded computers located right next to the hive. „The main goal is to develop a bee behaviour recognition system that can analyse bee actions in the hive accurately and quickly. It also aims to optimise neural networks so that they can operate effectively on compact embedded computers located right next to the hive. At present, we are investigating how accurately and how quickly trained networks recognise a specific action in an image, especially when bees are densely clustered and obscure one another, and the action itself lasts for a very short time,“ explains the researcher. The system should recognise and track insects – bees, bees carrying pollen loads, drones and wasps – and determine their coordinates, direction, speed or behavioural class. Different types of bee behaviour are being studied: the intensity of bees entering and leaving the hive, ventilation, trophallaxis, cleaning the hive entrance and defensive reactions. Particularly useful for urban beekeeping The urban environment has its own particular characteristics for bee life. Due to the urban heat island effect, it is warmer here, so bees often start flying earlier in spring and remain active for longer in autumn. Urban vegetation is also more diverse and blooms throughout almost the entire season – in parks, gardens, trees growing along streets and even on balconies. Nevertheless, urban hives are often located on rooftops or in small courtyards, making remote monitoring particularly useful. An urban beekeeper often looks after several hives located in different places, so an automated system would allow the monitoring even when they are not nearby. Thus, the most important thing is not the behavioural recognition itself, but what it signals. „The hive entrance is like a window into the overall internal condition of the colony; by observing it, a beekeeper can predict a great deal without opening the hive. An automated system transfers part of this observation to a computer that can operate 24 hours a day and immediately alert the beekeeper when it detects an anomaly,“ explains Assoc. Prof. Dr T. Sledevic. Computer vision does not learn this intuitively, but by analysing statistical data: researchers prepare thousands of video frames in which a person marks where the bee and wasp are, where trophallaxis is taking place or where a defensive reaction is occurring. By „showing“ these examples to a deep learning network, it learns to recognise the same visual features that an experienced beekeeper recognises subconsciously. „Essentially, we are trying to 'compress' the accumulated experience of a beekeeper into an algorithm that becomes accessible even to a novice beekeeper who does not yet have such intuition,“ says the researcher. The vision for the future – a smart hive In the future, individual behaviour recognition modules could be combined into a single mobile system that would operate continuously on a compact computer located next to the hive. Upon detecting something unusual, it would immediately send a notification to the beekeeper's phone. However, a fully smart hive would not rely solely on visual monitoring. The view at the hive entrance does not reveal everything happening inside the hive, so in the future it could be complemented by scales, temperature and humidity sensors, as well as sound analysis. „This is still a vision for the future – for now, we are working on reliable recognition of individual behaviours and their optimisation for embedded systems, but these are precisely the fundamental building blocks from which a fully smart hive could be assembled in the future,“ says Assoc. Prof. Dr T. Sledevic. Prepared by: Giedre Gedeikyte, Head of External Communications Projects This article was published in the autumn 2026 issue of "Sapere Aude"
More