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

A World Model Reinforcement Learning Approach for Vehicle Control: A vehicle controller based on DreamerV3 and a research platform for training and evaluating machine learning based controllers
(2026) Munck af Rosenschöld, Andreas; Wir, Oscar
The purpose of this thesis was to investigate whether a world-model-based reinforcement learning approach could be used for path following on a small off-road unmanned ground vehicle. The method involved building a 1/10 scale RC car platform equipped with a binocular camera and developing a ROS 2 graph to manage interprocess communication, including pose estimation and vehicle control. A worldmodel reinforcement learning controller was developed by adapting the DreamerV3 implementation for the path-following task. This controller was evaluated across various trajectories, including sine waves and clothoid turns on uneven grass, as well as backtracking on sand. A Pure Pursuit controller was used as a baseline. However, it was not tuned to individual paths and did not represent state-of-the-art controllers. The results indicate that the controller tracked all paths effectively and mostly smoothly, outperforming Pure Pursuit on all tasks in terms of cross-track error, successfully bridging the sim-to-real gap using a limited amount of training data. While it exhibited minor over-adjustments on certain paths and underutilisation of the steering range, the controller demonstrated an emergent behaviour of reversing to adjust its alignment at sharp corners. Consequently, this validated the successful development of the underlying research platform.
Theoretical and experimental analysis of inductor characteristics for hearing implants
(2026) Asadour, Christian; Safi, Abdulrahman
Implantable hearing devices rely on inductive coils to transfer power and signals wirelessly across the skin. The electrical and magnetic properties of these coils directly affect device effi-ciency, and size. This thesis investigates how core material, core diameter, wire diameter, and number of turns affect the inductance, DC resistance, quality factor, magnetic flux, and physical size of wire-wound inductors intended for hearing implant applications. The work was carried out in collaboration with Oticon Medical. A MATLAB model was implemented and evaluated for five core materials across seven core di-ameters and two wire diameters at a target inductance of 100 μH. Four prototype coils were also hand-wound on ferrite rod cores at Oticon Medical's laboratory and measured using a Source-tronic ST2827A precision LCR meter. The theoretical results were compared with the experi-mental measurements to evaluate how well the model reflected practical coil behavior. The findings provide a comparative framework for understanding how material and geometric choices affect coil performance and highlight key considerations for future coil design in im-plantable hearing devices.
Collaborative Robotic Arm and Humanoid Interaction for Kitting Tasks in Simulated Factory Environment: A Simulation-Based Evaluation of Motion Planning Methods for Automated Kitting in Dynamic Industrial Environments
(2026) Olsson, Marcus; Wilsborn, Elias
Future industrial automation requires robotic systems that can operate in workspaces where objects, equipment, and humans may be present. This thesis presents the development and evaluation of a simulation-based system for motion planning in a kitting task. The system was built around a gantry-mounted UR10e robot arm with a Robotiq gripper in a simulated factory environment containing a flow rack, crates, static and dynamic obstacles. The work integrated ROS 2, Isaac Sim, MoveIt 2, and several motion-planning frameworks in a containerised software architecture. Sampling-based planning with OMPL, GPU-accelerated planning with cuMotion and cuRobo, and a hybrid planner based on cuRobo MotionGen and MPC were implemented and evaluated. The planners were tested in simple motion cases, complete pick-and-place workflows, and a dynamic obstacle benchmark where the robot had to react to a newly introduced obstacle during execution. The results showed that cuRobo provided the strongest overall balance between planning speed, success rate, and motion efficiency in the static benchmark cases. cuMotion also achieved high success rates, but generally required longer planning times. The OMPL planners were computationally cheap and could be fast in successful cases, but showed lower robustness in several scenarios. The hybrid planner was able to react to dynamic changes in the environment, but its success depended on how close the obstacle appeared to the robot. When enough clearance was available, the hybrid planner recovered reliably, while recovery became less likely when the obstacle was inserted very close to the robot. The thesis shows that simulation is a useful tool for evaluating motion-planning methods for constrained industrial tasks before real-world deployment. It also shows that GPU-accelerated and hybrid planning methods are promising for robotic operation in dynamic environments, but that further work is needed before the system can fully represent realistic human-robot collaboration.
Lighthouse++ - A Search and Generative Engine Optimization Tool for VSCode
(2026) Jöeäär, Adam; Alm-Eriksson, Daniel; Hamdan, Mohammad; Johansson, Simon; Noufal, Yousef
Search engine optimization (SEO) remains an important challenge for e-commerce businesses in the ever-expanding and competitive World Wide Web (WWW) landscape. The emerging AI-driven information retrieval trends have added new dimensions to the WWW utility and further complicated the SEO problem. Traditional SEO is largely performed using external tooling, disrupting the flow as developers move back and forth between environments. In addition, current tools do not adequately consider the evolving AI-driven search landscape. This thesis presents the design and implementation of Lighthouse++, a Visual Studio Code (VS Code) extension for automated analysis and improvement of webpage quality with a primary focus on on-page SEO and Generative Engine Optimization (GEO). The system is built around Google Lighthouse and extends it with AI-assisted suggestions, a set of custom metrics, and an exploratory GEO component intended to provide heuristic feedback rather than predictive ranking-based estimates. In this regard, the thesis proposes a GEO-oriented evaluation framework based on semantic similarity and source visibility, intended to capture how webpage content may align with retrieval and answer generation in large language model systems. The developed extension audits webpages, identifies technical and structural issues, and generates suggested code changes intended to improve Lighthouse-related outcomes while also exposing GEO related feedback as an exploratory complement to traditional SEO analysis. Finally, Lighthouse++ comes with improved usability by enabling developers to code and test their websites for S&GEO in their IDEs. The project demonstrates the effectiveness of the proposed tool by evaluating and showing improved metrics across 80 pages from 18 different websites. The evaluation shows an increase in all metrics and that AI-assisted SEO and GEO are possible.
Logikprogrammering för styrning av autonoma system
(2026) Eriksson, Viktor; Fogelstrand, Erik; Frännhag, Evelina; Siik, Hampus
Detta arbete behandlar utvecklandet av ett logikprogrammeringsspråk för att styra autonoma system. En konceptuell modell presenteras där ett autonomt system beskrivs som en uppsättning inkanaler, utkanaler och regler som definierar hur data flödar mellan dessa. Utifrån modellen utvecklas språket Channelog, inspirerat av Datalog med stöd för kanaler och tidsordnade datamängder. Arbetet resulterar i en prototyp av exekveringsmiljö för Channelog samt en konceptuell modell. Prototypen består huvudsakligen av en parser genererad av BNFC och en översättare som översätter från den abstrakta syntaxen till SQL. Med denna prototyp kan man realisera en styrmodell.