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.
Senast publicerade
- Closed-loop Real-time Safety Critic as Guardrail in Robotics(2026) Chen, Xingyu; Nan, ShipengRobots are increasingly used in manufacturing, logistics, service, and manipulation settings, where physical safety is a basic premise rather than an optional supplement. While robot policies may be obtained through reinforcement learning, imitation learning, optimal control, etc., the development normally involves training, simulation, testing, or rollout data. Apart from learning task behavior, the interactive data can also be used to extract knowledge about danger. To address this issue, this thesis adapts the idea of safe reinforcement learning to a deployment setting in which the task policy is already available and should remain frozen. Instead of retraining the nominal policy with a new constrained objective, the work asks whether a learned safety module can act as a last-step guardrail: before each action is executed, it evaluates the nominal action, accepts it when risk is low, applies a small correction when risk is moderate, or switches to a recovery action when collision risk is high. Specifically, the proposed method is a closed-loop robot safety guardrail built around a real-time dual-head safety critic. Mixed-behavior trajectories are first used to construct hard-collision labels and continuous soft-risk labels; the critic is then pre-trained to estimate hard-constraint violation risk and cumulative soft-risk cost. During online execution, directional gating, candidate triggering, random shooting, gradient-projection soft correction, and emergency Recovery Actor takeover are combined to produce hierarchical and minimally invasive action correction. The framework is evaluated on multiple task suites, including a PointMaze navigation problem, a Franka Panda reach-obstacle task, and a Meta-World pick-place-wall manipulation, where the proposed guardrail consistently reduces collision risk while preserving task performance. To start with, collision rates prominently drop in all scenarios. For Panda reach-obstacle, the average minimum clearance also improves. On Meta-World, while the collision instances dramatically decrease, the overall task success also rises. Additional ablation results show that soft correction and Recovery address different risk stages and work best when combined.
- TransFURmer: Hair and Fur rendering using SwinIR - Transformer-based Neural Rendering of Hair and Fur using SwinIR(2026) Jin, Ryan; Jönsson, KevinWhile ray tracing algorithms offer photorealistic rendering of hair and fur, their high computational cost makes them unfeasible for real-time applications. Conversely, fast rendering techniques such as rasterization fail to capture accurate lighting effects and high-frequency details. This thesis investigates the feasibility of a Transformer-based pipeline tailored for rendering hair and fur. A SwinIR-based architecture is adapted within a Conditional Generative Adversarial Network (CGAN) framework to translate input buffers into high-quality renders. Trained on the SyntheticFur dataset, the model utilizes a combination of losses, notably Frequency Domain (FFT) loss, Learned Perceptual Image Patch Similarity (LPIPS) loss, and Wasserstein GAN with gradient penalty (WGAN-GP) loss, to accurately reconstruct the fine details and improve perceptual realism. The final model successfully generalizes to unseen geometry and the implemented FFT and LPIPS loss proved highly effective at preserving the texture of individual hair strands. However, the current model yields an average throughput of 0.13 fps, failing to satisfy real-time latency constraints. Additionally, hardware limits restricted the training crop size, and the training dataset was limited to simple primitives lacking complex self-shadowing scenarios. This likely contributed to shading inconsistencies and visual artifacts observed in the final renders. Despite these limitations, the final model is a faster alternative to path tracing for offline rendering tasks, and demonstrates the potential viability of Transformer-based neural rendering for complex hair and fur.
- Exploring Load Localization for Automated Guided Vehicles: Pallet pose estimation using a depth sensing camera(2026) Nordström, JonatanAutomated guided vehicles (AGVs) typically require pallets to be placed in precise locations to be able to pick them without advanced detection systems. This limits AGV systems in environments where pallets are also handled by human fork lift operators. Modern solutions to this problem often rely on machine learning, which requires powerful and costly computers inside the AGVs. In this thesis a computationally efficient method for estimating the pose of a pallet using a depth sensing camera is proposed and evaluated. The method uses a novel approach to find the region of interest directly in the depth map by segmenting it into depth slices and detecting the fork pockets of the pallet with fast 2D image processing techniques. A rough pose estimate is calculated from the detected fork pockets, after which only a small region around the pallet is deprojected into a point cloud where the pose is refined using the iterative closest point (ICP) algorithm. The method was evaluated on 2678 depth maps of EU-pallets captured on the floor and in a rack at angles up to ±15◦. A pallet was detected in 90.14% of the frames and 73.61% of the frames resulted in an accepted pose estimation. The standard deviation of the estimated vertical position of a stationary pallet was 4.59 mm in the best case, indicating a high repeatability. The complete pipeline was 30.32% faster than performing ICP on the full point cloud, showing that the proposed approach reduces the computational load while maintaining an accurate localization.
- From Home to Community Care: Transforming a Community Building as a Spatial Mediator in Chinese Post-Industrial Contexts(2026) Zhang, JingwenThis research explores how spatial design can support ageing in place in Chinese post-industrial urban communities. It focuses on Dongshan Community in Dalian, a university-affiliated residential area developed under the Danwei work-unit system, where long-term residence has fostered strong social ties and a relatively stable neighbourhood structure. With an ageing population and the gradual transformation of the former welfare system, care resources have become increasingly fragmented, and responsibilities for everyday care are shifting from the private home towards the community level. The research examines how a community building can function as a spatial mediator connecting healthcare support, everyday social interaction, and community activities. The study adopts a co-production approach to understand residents’ lived experiences and emerging needs. The research process includes a key information interview with a long-term resident, a community survey, and focus workshops that provide insights into daily life patterns, care expectations, and existing social networks within the neighbourhood. Based on these findings, the design explores spatial strategies to reorganise the community building and integrate diverse care-oriented and social functions within a shared spatial framework. The project proposes an adaptable community environment that supports health care ser vices, everyday interaction, and collective activities. Principles from Traditional Chinese Medicine inform the design by linking health, daily life, and spatial experience. The research demonstrates how community architecture can strengthen local care infrastructures and sustain ageing in place in Chinese post-industrial neighbourhoods.