Chalmers Open Digital Repository

Välkommen till Chalmers öppna digitala arkiv!

Här hittar du:

Studentarbeten utgivna på lärosätet, såväl kandidatarbeten som examensarbeten på grund- och masternivå
Digitala specialsamlingar, som t ex Chalmers modellkammare
Utvalda projektrapporter

Forskningspublikationer, rapporter och avhandlingar hittar du i research.chalmers.se

Enheter i Chalmers ODR

Välj en enhet för att se alla samlingar.

Visar 1 - 2 av 2

Senast publicerade

Hälsa och Säkerhet - Reducera tunga lyft av tvättsäckar för lokalvårdare på Östra Sjukhuset
(2025) Al Masri , Yousef; Friberg, Jakob; Jirdén, Tora; Jönsson , Moa; Löwenberg , David; Sahlqvist, Johan
This bachelor thesis explores how user-centered design with focus on ergonomics can contribute to a safer handling of lifting heavy laundry bags for environmental service workers in their daily work. The aim of the project is to reduce the risk of injury when lifting laundry bags into a laundry cart for environmental service workers at Östra Sjukhuset during clearance of laundry from the healthcare personnel’s locker rooms. The project has been executed through an iterative design process with interviews, ideation, sketches, concept development, prototyping and user tests. Initially, information gathering and analysis of the current conditions were carried out to identify the target group's needs. From the result of these, the process continued further until ideation and concept development where a number of concepts were produced. The concepts were evaluated and refined, then prototypes were built for the three further developed concepts which were presented to the client. The final result was two of the concepts in combination with each other where one concept distributes the weight more equally between laundry bags and the other assists in lifting said laundry bags. The conclusion is that the developed concept fulfills the purpose of the project and that further work in the area is advocated. It is suggested that future evaluations be carried out to complete the concept further with calculations and technical sketches.
Synthetic Data Generation for Industrial Perception Using 3DGS
(2026) Gurusinghe, Dineth; Khalid, Yahya
Autonomous mobile robots operating in industrial environments require reliable camera-based perception to detect pedestrians, forklifts, and other vehicles in dynamic shared workspaces. However, collecting and manually annotating sufficiently large and diverse real-world datasets in factory environments is costly and timeconsuming. This thesis investigates the use of 3D Gaussian Splatting (3DGS) as the basis of a synthetic-data-generation pipeline for training industrial object-detection models and examines factors affecting simulation-to-reality transfer. A 3DGS reconstruction of a real factory environment was integrated into Unreal Engine as a digital-twin background and combined with dynamic simulated actors representing pedestrians, forklifts, and other vehicles. Synthetic images and corresponding object-detection annotations were generated from multiple virtual camera viewpoints and compiled into datasets for training YOLOv8-based detection models. A model trained on a 21-camera synthetic dataset achieved an mAP50 of 0.971 on synthetic test data but only 0.227 on real factory images, demonstrating a substantial simulation-to-reality domain gap. Experiments showed that both camera-viewpoint diversity and environmental representation affected real-world generalisation. Models trained using the 3DGS factory representation transferred better to real images than those trained using a conventional Unreal Engine environment. However, because the appearance encoded by 3DGS is tied to the lighting conditions under which the scene was captured, lighting differences could not be varied independently, limiting the extent to which this improvement could be attributed solely to the background representation. Diffusionbased style transfer substantially improved zero-shot real-domain performance, increasing mAP50 from 0.125 to 0.440, while subsequent pseudo-labelling increased it further to 0.504 without requiring manually annotated real training images. Synthetic pre-training followed by fine-tuning on real data increased mean mAP50 from 0.876 to 0.891 compared with real-only fine-tuning, although this difference fell within the observed variability across data splits. In contrast, joint training with style-adapted synthetic and real data produced more substantial gains when only limited annotated real data was available. Overall, the results indicate that 3DGSbased synthetic data is most valuable as a complement to real-world data rather than as a replacement for it, particularly when viewpoint diversity, visual domain alignment, and training strategy are considered together.
Falsification of Safety-Critical Decision and Control Software
(2026) Zhang, Pingjie; Zhang, Yang
Safety testing of autonomous driving systems requires methods that can expose rare but safety-critical failures beyond nominal driving behavior. Since exhaustive realworld testing is infeasible, simulation-based falsification provides a practical way to actively search for scenarios in which a system under test violates desired safety requirements. This thesis investigates deep reinforcement learning for falsification of safety-critical decision and control software in closed-loop simulation. The falsification task is formulated as an adversarial control problem in a simplified longitudinal two-vehicle scenario, where a learning agent controls the target vehicle and attempts to induce collisions with an ego vehicle controlled by the system under test. A TD3- based falsifier is first evaluated against Monte Carlo Tree Search and uniform random sampling to study collision discovery performance across randomly sampled initial conditions. The thesis then extends the framework towards diversity-oriented falsification using a DTW-based(dynamic time wrapping) Series RL method. In this method, previously discovered collision trajectories are stored in a repository and used to reward later agents for finding trajectories with different gap and relativespeed evolution. A collision replay buffer is also introduced to increase the reuse of rare collision-related experiences during training. The results show that the TD3- based falsifier discovers collisions more effectively than the non-learning baselines in the tested setting. They also show that the DTW-based diversity reward leads to a broader and less repetitive set of collision trajectories than the baseline Series RL method. These findings suggest that reinforcement learning can support both efficient failure discovery and diversity-oriented exploration of failure modes, although the conclusions are limited to the simplified simulation setting considered in this thesis.
Vibration Condition Monitoring of an Electric Drive Unit
(2026) Nalawade, Aditya
The increasing competition and rapid development in electromobility are driving accelerated product development, higher drivetrain performance, and consequently greater mechanical stress on components. Ensuring reliability and safety therefore requires effective condition monitoring during testing. Vibration-based condition monitoring is widely used for this purpose; however, the effectiveness of such monitoring strongly depends on the measurement setup, particularly sensor placement. In practice, sensor positioning is often performed without a structured approach, which could lead to inconsistent and potentially unreliable results. This study focuses on developing a standardized procedure for vibration measurement setup during drivetrain testing. Multiple real-time measurements were conducted and analysed using various evaluation metrics to assess the influence of sensor placement and operating conditions. The results demonstrate that sensor placement must be carefully optimised for different machine configurations. In addition, the classification of load cases requires refinement to support a consistent and standardized testing methodology. The proposed approach contributes to improved reliability and comparability of condition monitoring results.
Design and Optimization of a Wideband Feed Antenna for the Onsala 20-m Telescope
(2026) Valsan, Abhijith
Wideband feed antennas are an essential part of reflector antenna systems, as they affect the radiation characteristics and, in turn, the reflector performance, which is the focus of the present thesis. A wideband quad-ridge horn feed was designed and optimized for the Onsala Cassegrain reflector system. The main objective of this work was to achieve a narrow radiation beam that satisfies the required edge taper while providing acceptable impedance matching and antenna efficiency. The antenna geometry is determined and studied by means of electromagnetic simulation. The ridge profile, horn geometry, and backshort, as well as antenna dimensions and radiation characteristics, were studied in detail. The 3D spline-based approach was used to design the ridge profile, providing flexibility in controlling the antenna geometry. Several geometric configurations and scaling factors were investigated to study their influence on radiation beamwidth and efficiency. The scaled design of 1.8 has shown approximately a 12-degree beam width at -12 dB point at 4 GHz and has retained a narrow beam with the increase of frequency and better antenna efficiency. The efficiency of the antenna has been assessed through simulated reflection coefficient, radiation patterns, spillover efficiency, illumination efficiency and combined aperture efficiency as measuring parameters. Although the target beamwidth could be achieved or closely approached, the overall efficiency remained below the target value of 60%. The results indicate that modifying the geometry and scaling of a wideband quad ridge horn provides significant control over its radiation pattern. Nevertheless, further optimization and validation are required to improve the overall efficiency while maintaining the required beam characteristics.