For background information on the topic as a whole, scroll to the end of this page.
Available Project Proposals
If you are interested in the general topic of Machine Learning, or if have your own project idea related to the topic, please contact us directly. Alternatively, you can also work on one of the following concrete project proposals:
Optimising path planners using temporal specifications (Moritz Hahn) Optimising path planners using temporal specifications
Contact person: Ernst Moritz Hahn, e.m.hahn@utwente.nl
Title: Optimising path planners using temporal specifications
Description: The master thesis is in the direction of optimising global path planners for autonomous driving vehicles. Global path planners synthesise vehicle trajectories to go from point A to point B while avoiding obstacles. Although this task seems simple, the complication occurs when the trajectories generated must follow rules of engagement and avoid obstacles (static and dynamic).
The goal is to utilise the previous work [1] on optimising robot trajectories published at IROS 2022 and adapt it for an autonomous driving application. Your tasks would be,
- to replicate the results from [1],
- extend Signal Temporal Logic specifications to work for one of the autonomous driving applications,
- setup machine learning based global path planner and,
- optimise path planner trajectories based on logical constraints.
[1] Dhonthi, Akshay, et al. "Optimizing demonstrated robot manipulation skills for temporal logic constraints." 2022 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS). IEEE, 2022.
Point cloud classification, segmentation, and visulization (Faizan Ahmed) Supervisors: Faizan Ahmed
Point clouds are sets of spatial data points captured by 3D scanning techniques such as lidar. These point clouds contain many million points of data, resulting in 3D representations of the railway environment. Point cloud data can be used to create machine learning models that can classify the object in rail infrastructure automatically.
The main research question is: How to reliably segment, classify, and visualize point clouds for railway catenary systems with scant computational power, memory, and ground truth labels?
However, the student can focus on a combination of topics given below:
- leverage existing trained models for point cloud labeling
- pre-process (down-sampling. Augmentation, filtering, voxelization etc.) point clouds for accurate model training
- train deep learning models for point cloud segmentation and object detection
- analyze the variation in object sizes with relation to the accuracy of the trained models.
- analyze the effect of point cloud density on accuracy
- determine the accuracy of the trained model objectively (or What are the excellent metric for such machine learning tasks?)
- connecting point clouds with existing 3D object libraries
Leveraging large language models for computer algebra (Benedikt Peterseim) Supervisors: Milan Lopuhaä-Zwakenberg, Benedikt Peterseim (contact: benedikt.peterseim@utwente.nl)
Large language models (LLMs), particularly recent advances in so-called reasoning models, demonstrate remarkable performance in classical computer algebra tasks such as symbolic integration. However, they remain unreliable—sometimes showing what is referred to as “hallucinations”—without any formal guarantee of correctness. This limits their practical usefulness, as manually verifying LLM-generated solutions can be cumbersome. The issue is exacerbated by the fact that incorrect outputs may appear logical and convincing at first glance due to the nature of LLM training.
This Master’s thesis project aims to integrate LLMs with formal methods to ensure correctness in specific, well-defined computer algebra tasks.
This may include, but will not be limited to:
- Designing a domain-specific formal grammar that the LLM must adhere to when outputting its final result.
- Applying grammar-constrained decoding [1] to force the LMM to provide its final answer in this specified formal language.
- Utilising external verifiers to check the parsed results or, if necessary and sufficiently simple, developing simple verification tools for this purpose.
- Leveraging LLMs to translate informal user input into the specified formal language. This presents a creative challenge: the formal language should be close enough to standard calculus notation to allow users to intuitively assess the correctness of the translation.
- Evaluating this approach by comparing its performance to existing computer algebra systems.
[1] Scholak, Torsten, Nathan Schucher, and Dzmitry Bahdanau. "PICARD: Parsing incrementally for constrained auto-regressive decoding from language models." Proceedings of the 2021 Conference on Empirical Methods in Natural Language Processing, November 2021.
Contact
Background
Machine learning helps to learn from enormous amounts of data gathered as a direct or indirect product of a process, whether it be an industrial product line, efficient farming for food production, prediction of bacterial growth, or any other industrial, healthcare-related, or technological endeavor. Similar to humans, machine learning extracts patterns from data if enough examples are provided. Reinforcement learning is an action-oriented type of learning in which an agent tries to maximize a given reward signal without being told explicitly which actions to take. This type of learning is often combined with other types of machine learning. For instance, deep learning is widely applied in scenarios with large state spaces.
The topic of machine learning leads to projects of varying nature such as theoretical (algorithm development, computational efficiency, convergence of machine learning algorithm, and formal representation), algorithm training for classification or regression (see some example projects below), predictive maintenance, and reinforcement learning.
Related Courses: Data Science
