DR-NTU (Digital Repository of NTU)
Not a member yet
116018 research outputs found
Sort by
Wayfarer
No full textMoving to a new country is full of excitement but with its unique challenges as well. Majority of the international students have experienced difficulties when adapting to the unfamiliar culture and new education system. Language and cultural difficulties in particular might cause them to feel isolated even with the supporting systems and programs offered by academic institutions which might severely impact their social integration and academic results.
This project targets Chinese students in Singapore, focusing on those who prefer a more gradual, self-directed transition process. While the research highlights the need for flexible supporting systems, ‘Wayfarer’ aims to transform difficulties into adventures with five seasoned animal characters. A bilingual mobile application that encourages Chinese students to explore their new surroundings at their own pace by combining interesting stories, cultural exploration, and practical advice to facilitate the transition.Bachelor's degre
Performance study of rotated QAM modulation scheme over fading channels
No full textDigital Quadrature Amplitude Modulation (QAM) combines both Amplitude Shift Keying (ASK) and Phase Shift Keying (PSK), by using different combinations of amplitudes and phases to encode and create unique data symbols. The modulated signal is transmitted over a communication channel which may be demodulated at the receiving end of a communication system.
Fading Channels are communication channels where the transmitted signal is affected by Environmental, Mobility, and Atmospheric factors. These may be in the form of large objects obstructing the direct path between a transmitter and receiver, movement of the transmitter or the receiver, and weather conditions such as rain or snow. The transmitted signal may be degraded significantly due to the impact of fading channels.
To study the impact of fading channels on communication systems, mathematical models have been developed to mimic its effects in various conditions. Bit-Error-Rate (BER) is a key indicator in evaluating the performance of a communication system. Signal-to-Noise-Ratio (SNR) or Energy per Bit to Noise Spectral Density Ratio (Eb/No) is usually used along with BER to evaluate the performance of a wireless communication system.
In this project, MATLAB programming was used to generate the BER vs Eb/No graphs for QAM signals under different types of fading channel models. The simulated results will be verified with the calculated theoretical results.Bachelor's degre
Establishing collaborative robot arms through digital twin-enabled decision support systems
No full textWith waste volumes continuing to rise globally, the need for smarter and more efficient waste management solutions is becoming increasingly urgent. Traditional sorting methods are often manual and not sustainable in the long run. To address this, this project presents a framework that integrates a digital twin of a robotic arm with deep reinforcement learning for autonomous waste sorting. A digital twin of the robot arm was developed in Unity, and real-time connectivity was established between the physical robot and its virtual counterpart, using a combination of User Datagram Protocol (UDP) and Open Platform Communications Unified Architecture (OPC UA). A DRL policy was trained using the Proximal Policy Optimization (PPO) algorithm to perform pick-and-place tasks in a dynamic simulated environment, where multiple waste objects move continuously to simulate a conveyor belt system.
The results demonstrate that the DT can accurately reflect the physical robot's behavior, and the trained DRL model achieved a reasonable success, even under more complex testing conditions. The framework lays a strong foundation for future development toward a fully autonomous, intelligent, and scalable robotic sorting system for real-world deployment.Bachelor's degre
Integrating raspberry PI 5 with emoji response recording and predicted pleasantness and eventfulness metrics
No full textThis project details the work of deploying an International Organisation for Standardisation (ISO) prediction model into an edge device, Raspberry PI 5 using the Docker platform for a real-time soundscape analysis. This edge device tool records a real-time audio signal utilising a Log-Mel Spectrogram conversion to predict ISO Pleasantness (PL) and Eventfulness (EV) values from a machine learning model. Additionally, a human response graphic user interface (GUI) is deployed to collect real-time responses from participants through emoticon-based evaluations that work hand in hand with the prediction model to record model PL and EV as well as participant's responses and store results onto Google Sheet API for real-time monitoring. The data stored is made available via a Shiny Python dashboard for further analysis.
This research aims to enhance research and urban planners in soundscape assessment, offering a more efficient and cost-effective research device.Bachelor's degre
Finite element analysis modeling of fatigue problems
No full textUniaxial fatigue tests are conducted in specimens with defects to determine the crack
propagation geometry and fatigue life under uniaxial cyclic stress conditions. This project
presents the Finite Element modelling of two common uniaxial fatigue samples of the
BS4360 50D steel using the Abaqus Extended Finite Element Method (XFEM).
First, a preliminary study was conducted on the rectangular bar sample to probe issues with
XFEM in Abaqus. Solutions were proposed to reduce the likelihood of the PHILSM value
error (Element loop error) and improve the meshing scheme to enhance accuracy. The effects
of the Paris Law constants magnitude on result accuracy and simulation stability were also
investigated. The rectangular bar model simulation results were compared with the
experimental results to validate the proposed solutions.
Next, an Hourglass model was studied. The effects of the proposed solution to reduce the
probability of an element loop error were studied in detail, and a preliminary study into
meshing schemes for small cracks was conducted. The results from both models were
compared with a C program developed based on the Newman and Raju equations to predict
fatigue life for a sample under uniaxial loading. Following the findings from the 2 Finite
element models, modelling procedures to predict crack propagation and fatigue life prediction
were developed.Bachelor's degre
Design of wideband power amplifier for cognitive radio communication systems
No full textThis project focuses on designing a Class AB RF Power Amplifier aimed at maximizing
output power for a low-power driver stage amplifier intended for use in GSM 900, 2.1 GHz
WCDMA, and 2.4 GHz WLAN applications. The design process is conducted in Cadence
using TSMC transistor device models. The primary goal is to achieve 12 dBm output power
within the linear region, ensuring minimal distortion and optimal signal integrity. At the same
time, the amplifier must maintain high efficiency across the different frequency bands,
minimizing power loss and improving overall performance across various wireless standards.
To accommodate the varying frequency requirements of each standard, a multiband approach
has been adopted. This involves designing individual narrowband matching networks tailored
to each wireless standard. Each network is optimized for the specific frequencies and
impedance characteristics of the corresponding band, ensuring maximum power transfer and
minimal reflection for each standard. Results show the performance of the Power amplifier
across different wireless standards using the same amplifier circuit, including the PAE, P1dB
and power gains.
Additionally, the project will explore the gm/id methodology for sizing the device, which
represents the transconductance-to-current ratio, determining how efficiently the transistor
can convert input voltage into output current. This methodology will be employed to
precisely adjust device parameters, ensuring the desired output power is achieved. It provides
a systematic approach for selecting the optimal transistor sizes, which is especially important
as MOSFET technologies continue to shrink, requiring careful consideration of
size-dependent performance factors.
Overall, the approach focuses on achieving high output power and efficiency while
maintaining the versatility needed to support multiple wireless communication standards
within the same RF power amplifier design.Bachelor's degre
Towards robust and effective visual localization
No full textVisual localization is the task of determining an agent's location using visual sensors such as cameras and LiDARs. It plays a pivotal role in applications including autonomous navigation, robotics, and augmented reality, where precise positioning in complex environments is essential. However, existing approaches are not robust against environmental perturbations. Moreover, the fusion of multi-modal information from different sensors is not optimized for effective visual localization in both same-view and cross-view settings. Furthermore, there is a gap in efficiently transferring rich multi-modal information into single-modal pipelines, limiting the potential for lightweight and computationally efficient operation.
To address these challenges, we leverage neural ordinary differential equations to achieve robust visual localization under challenging conditions. The non-intersection property of neural ordinary differential equations also ensures the consistency between the scene feature and the actual geo-distance. Additionally, we propose to embed features into spaces of distinct curvatures to enhance their geometric representation ability. Furthermore, we incorporate learnable manifolds with flexible chart functions and tangent space metrics to effectively fuse features from different modalities. This supports more comprehensive sensor data interaction and thus improves multi-modal visual localization performance. In addition, we propose to combine self-agent and cross-agent relationships to enhance the effectiveness of localization knowledge transfer. The multi-relationship knowledge transfer increases the potential for lightweight single-modal localization with improved efficiency.
We evaluate these approaches across various visual localization scenarios. We also conduct extensive experiments on benchmark datasets to demonstrate the effectiveness of these approaches. These approaches hold significant potential for advancing autonomous systems and related technologies.Doctor of Philosoph
SFDM: robust decomposition of geometry and reflectance for realistic face rendering from sparse-view images
No full textIn this study, we introduce a novel two-stage technique
for decomposing and reconstructing facial features from
sparse-view images, a task made challenging by the unique
geometry and complex skin reflectance of each individual.
To synthesize 3D facial models more realistically, we endeavor to decouple key facial attributes from the RGB color,
including geometry, diffuse reflectance, and specular reflectance. Specifically, we design a Sparse-view Face Decomposition Model (SFDM): 1) In the first stage, we create a general facial template from a wide array of individual faces, encapsulating essential geometric and reflectance characteristics. 2) Guided by this template, we
refine a specific facial model for each individual in the second stage, considering the interaction between geometry
and reflectance, as well as the effects of subsurface scattering on the skin. With these advances, our method can reconstruct high-quality facial representations from as few as
three images. The comprehensive evaluation and comparison reveal that our approach outperforms existing methods
by effectively disentangling geometric and reflectance components, significantly enhancing the quality of synthesized
novel views, and paving the way for applications in facial
relighting and reflectance editing. Visit our project page for
more details https://kingjg.github.io/SFDM.github.io/.Ministry of Education (MOE)Agency for Science, Technology and Research (A*STAR)Submitted/Accepted versionThis work was supported in part by the Ministry of Education, Singapore, under its Academic Research Fund Grants (MOE-T2EP20220-0005 & RT19/22) and the RIE2020 Industry Alignment Fund–Industry Collaboration Projects (IAF-ICP) Funding Initiative, as well as cash and in-kind contribution from the industry partner(s). J.Hu was supported by the National Natural Science Foundation of China under Grants 62402083
Design and analysis of wellhead tubing hanger metal seal
No full textThis final year project presents the design and analysis of a novel wellhead tubing
hanger metal seal aimed at mitigating leakage caused by shuttling-induced
movements. In the context of an evolving global energy landscape, where traditional
oil and gas operations remain critical despite the shift towards cleaner energy,
robust and reliable wellhead equipment is paramount. Sponsored by SPM Oil &
Gas, the project focuses on addressing a key challenge identified on the well
site—namely, the overcompression of sealing elements due to subtle vertical
displacements of tubing hangers.
The design process commenced with an extensive product research phase,
reviewing existing sealing technologies. Building upon these foundations, several
conceptual designs were generated and refined, incorporating features like straight
bore sealing bumps. Verification of the proposed designs was achieved through
rigorous mathematical calculations and Finite Element Analysis (FEA) simulations,
which assessed structural integrity and sealing performance under cyclic pressure
loading. Key performance criteria—including von Mises stress, equivalent plastic
strain and contact pressure—were evaluated, revealing both the advantages and
limitations of each design concept. Furthermore, shuttling analysis was conducted to
estimate probable vertical displacements, providing critical insights for the
development of a test fixture for subsequent qualification tests.
The project demonstrates a viable pathway for developing the tubing hanger seal,
while also outlining clear directions for future work in tool design, further seal
refinement, and robust rigidization strategies to fully mitigate shuttling effects in
high-pressure wellhead applications.Bachelor's degre
India and Pakistan dangerously close to crisis after deadly Kashmir attack
No full textAmid domestic pressure to take military action against Pakistan, there are lessons from India’s response to two previous attacks during Indian Prime Minister Narendra Modi’s term.Published versio