DR-NTU (Digital Repository of NTU)
Not a member yet
116018 research outputs found
Sort by
The future of Iran's axis of resistance in Syria and Lebanon
No full textThe war between Israel, Hamas and Hezbollah transformed the Middle East’s balance of power, in particular, weakening Iran’s influence in Syria and Lebanon. What is the future of Iran’s Axis of Resistance, and can Iran rebuild its influence in the Levant?Published versio
An energy-based numerical continuation approach for quasi-static mechanical manipulation
No full textRobotic manipulation inherently involves contact with objects for task accomplishment. Traditional motion planning techniques, while having shown their success in collision-free scenarios, may not handle manipulation tasks effectively because they typically avoid contact. Although geometric constraints have been introduced into classical motion planners for tasks that involve interactions, they still lack the capability to fully incorporate contact. In addition, these planning methods generally do not operate on objects that cannot be directly controlled. In this work, building on a recently proposed framework for energy-based quasi-static manipulation, we propose an approach to manipulation planning by adapting a numerical continuation algorithm to compute the equilibrium manifold (EM), which is implicitly derived from physical laws. By defining a manipulation potential energy function that captures interaction and natural potentials, the numerical continuation approach is integrated with adaptive ordinary differential equations that converge to the EM. This allows discretizing the implicit manifold as a graph with a finite set of equilibria as nodes interconnected by weighted edges defined via a haptic metric. The proposed framework is evaluated with an inverted pendulum task, where the explored branch of the manifold demonstrates effectiveness.National Research Foundation (NRF)Published versionThis research was supported by grants from the National Research Foundation, Singapore, under the NRF Medium Sized Centre scheme (CARTIN)
Enhancing stereo vision estimation with custom Berhu-Gradient loss
No full textStereo vision and disparity estimation play a critical role in allowing machines to perceive depth, with its applications spanning from various real-time systems such as autonomous driving, robotics to medical imaging. The practicality of these algorithms in real-time systems are hence tied closely to its efficiency and accuracy. As such, exploration have been done to employ deep learning for stereo vision, aimed at improving the performance in terms of accuracy and efficiency. Deep learning models generally outperform traditional stereo vision techniques in handling challenging regions like occlusions and texture-less regions. Despite these advancements, optimising disparity estimation models for higher accuracy and efficiency is still an ongoing challenge. Thus, we would like to explore a novel loss function and evaluate its effects on the accuracy of disparity estimation. This research aims to analyse the accuracy of the resulting model from our novel loss function. To achieve this, we train a baseline model with Huber loss function and compare it against our model which is trained on Berhu-Gradient loss function. The methodology involves modifying the loss function from a smooth L1 function (Huber Loss) to a hybrid between Berhu Loss and Gradient Loss. We will compare the results of the baseline model against our model and discuss the implications of the new loss function. The evaluation metric used includes end-point error (EPE) and disparity accuracy thresholds to assess improvements in depth estimation performance. To further improve on the accuracy, an ablation study is done to fine tune hyperparameters within the new hybrid function.Bachelor's degre
Deep learning and computer chess
No full textComputer chess began with traditional engines that used handcrafted evaluation functions and brute-force search methods. Over the years, advancements in deep learning have introduced new ways to evaluate positions by learning from past chess game data. This project explores how deep learning can improve position evaluation in chess engines by implementing and comparing two models: Giraffe and DeepChess. Giraffe uses a feature-based representation with a deep neural network to assign position scores, while DeepChess uses a Siamese network to compare positions and determine which is better.
Both models were trained on large datasets of chess positions and compared based on how accurately they evaluated a set of test positions, as well as how closely their move recommendations aligned with Stockfish. The results show that DeepChess had a higher evaluation accuracy (94.47%) than Giraffe (78.65%) and generally selected better moves. On the other hand, Giraffe required less computational resources to train. Further improvements were tested, including using DeepChess’ feature representation in Giraffe and adding transformer-based architectures. A web-based chess application was also developed to demonstrate how these trained models can assist users in their chess gameplay.
Overall, this project successfully integrated neural network-based evaluation into chess engines, demonstrating how deep learning can enhance position evaluation without relying on predefined heuristics or even requiring chess knowledge.Bachelor's degre
Bunker supply chain analysis for the adoption of ammonia as an alternative bunker fuel
No full textAmmonia is one of the alternative marine fuels to reduce greenhouse gas emissions for international shipping. Ammonia bunker supply chain includes ammonia production, storage, transportation, loading and bunkering operations and consumption process. Current research topics include ammonia bunkering supply and demand configurations, ammonia safety study for marine applications, ammonia bunkering port selection, and its bunker supply chain resilience. This study aims to analyse the adoption of ammonia as bunker fuel from the bunker supply chain perspective, comprising five studies:
Study 1 conducts a systematic literature review on ship bunkering from perspectives of bunkering management, alternative fuel bunkering and bunkering operational risk. A concept of sustainable, safe and smart (3S) bunkering is proposed for multi-fuel bunkering, digital bunkering, and risk-based approaches for implementing alternative marine fuels.
Study 2 describes the development of a discrete event simulation model for bunker supply chain, emphasising how ammonia bunkering affects the operational and economic performance of the system. The results show that the increase in bunkering service time compared with pure MFO bunkering is up to 113.7% for ammonia-MFO dual fuel bunkering, and up to 61.6% for pure ammonia bunkering. Ammonia bunkering flow rate is the most sensitive parameter for bunkering service time, with an effect of up to 51.3% decrease when it is increased by 50%. Moreover, the increase in annual operational cost for a bunker supplier is up to 19.1% under the condition of ammonia-MFO dual fuel bunkering. The number of ammonia bunker supply vessels is the most sensitive parameter for annual operational cost, and if it is increased by 50%, the increased annual operational cost variability is as high as 15.2%.
Studies 3 and 4 investigate the operational risk of accidental ammonia release in bunkering ports, considering various bunkering supply scenarios, release events, meteorological conditions, seasons, and geographical locations. Findings indicate that under the risk assessment criteria of 1100 ppm maximum cloud footprint, wind speed emerges as the most influential factor for both small and medium release scales, exhibiting sensitivities of up to -4 and -13.25, respectively. In the case of large release scales, hose diameter emerges as the most important factor, exhibiting a sensitivity of +33.77 to 1100 ppm maximum cloud footprint and a sensitivity of +26.17 to the 3% lethality footprint. Moreover, study 4 shows the data-driven framework, including a user layer, simulation-based knowledge layer and decision layer. An integrated approach incorporating energy balance modelling, Gaussian dispersion modelling and an adaptive neuro-fuzzy inference system (ANFIS) is developed. The synergy effects of different parameters are also presented.
Study 5 investigates the ripple effect of ammonia supply disruptions on an ammonia bunkering port, considering ammonia bunker demand dynamics. Low-frequency and high-impact disruption types are discussed in this study. A Coloured Petri Nets model shows the path of spreading disruptions, their interrelationships and their effects on the system performance. The cost, lead time and GHG emission reduction are selected as indicators to evaluate the system performance.
This PhD work contributes to the field of ammonia as a marine fuel, advancing both scientific knowledge and practical applications. It fills critical gaps in the literature on ammonia bunkering in the maritime sector and is the first to examine the combined impact of various factors related to the ammonia bunker supply chain, operational risk, and disruption risk. The proposed models, including a discrete event system simulation model, Gaussian dispersion model, simulation-based ANFIS model, and Petri Nets model, can be applied to analyse other types of fuels. Additionally, the insights and discussions on the risks associated with ammonia bunkering provide valuable references for stakeholders in making decisions about bunkering supply configurations and risk mitigations.Doctor of Philosoph
Environment data processing for a data centre (2)
No full textData centres form the backbone and brains behind contemporary digital services. But in order for data centers to function and be reliable, a significant amount of energy is used for the purpose of cooling servers. This project aims to develop predictive models to accurately forecast the cooling power consumption required to cool data centres, using the TDC2.0 dataset from an 11-month experiment conducted in a Singapore air-cooled data centre testbed. By leveraging machine learning techniques, such as tree-based models, neural networks and foundation models, robust models were created that can model cooling power consumption based on environmental and operational parameters. From the experiments, it was found that just using the temporal dependencies and autocorrelation properties of the cooling power consumption was sufficient to give accurate predictions, outperforming models which incorporated environmental factors and setpoint information as the features. It was also found that zero-shot forecasting performance of the time-series foundation model TimesFM v1.0 and v2.0 compared to the other models used was poor, possibly due to the limited context length.
Additionally, using feature importance metrics and recursive feature elimination from LightGBM and XGBoost, the most important environmental and setpoint features that influenced cooling power consumption were explored in-depth and analyzed to better optimize for cooling power efficiency.Bachelor's degre
Is Singapore burning out? Unpacking the FIRE movement and its implications
No full textAmidst soaring costs of living and intensifying pressures of financial self-reliance in Singapore,
the Financial Independence, Retire Early (FIRE) movement has emerged as a pragmatic
response to these challenges. However, upon deeper inspection, the FIRE movement sits
uneasily within the landscapes of financial subjectivities. Drawing on qualitative interviews
with 15 young working adults actively pursuing FIRE, the study utilises a financial
citizenship framework to interrogate how FIRE strategies—characterised by frugality,
disciplined investing, and a rejection of conventional wage labour—simultaneously reinforce
and undermine state-driven financial subjectivities. The findings reveal a paradox: while
adherents rigorously embody state-promoted ideals of self-reliance and rationality, affirming
the effectiveness of its neoliberal governmentality project, their actions may subvert and
destabilise state developmental goals.Bachelor's degre
Natural multimodal fusion-based human-robot interaction: application with voice and deictic posture via large language model
No full textTranslating human intent into robot commands is crucial for the future of service robots in an aging society. Existing Human-Robot Interaction (HRI) systems relying on gestures or verbal commands are impractical for the elderly due to difficulties with complex syntax or sign language. To address the challenge, this paper introduces a multimodal interaction framework that combines voice and deictic posture information to create a more natural HRI system. Visual cues are first processed by the object detection model to gain a global understanding of the environment, and then bounding boxes are estimated based on depth information. By using a large language model (LLM) with voice-to-text commands and temporally aligned selected bounding boxes, robot action sequences can be generated, while key control syntax constraints are applied to avoid potential LLM hallucination issues. The system is evaluated on real-world tasks with varying levels of complexity using a Universal Robots UR3e manipulator. Our method demonstrates significantly better HRI performance in terms of accuracy and robustness. To benefit the research community and the general public, we will make our code and design open-source.National Research Foundation (NRF)Submitted/Accepted versionThis work is supported by the EFRE and MWK ProFö-R&D program, under Grants FEIH_ProT_2517820 and MWK32-7535-30/10/2. This work is also supported by “CAIpirinha—Conversational AI and Personalized Interaction for Risk-Aware Navigation With Human Awareness,” under Grant Förderkennzeichen: BW7_1030/02, and Funding Program “Invest BW—Innovation III.” This work is additionally supported by the National Research Foundation, Singapore, under its Medium-Sized Center for Advanced Robotics Technology Innovation
An analysis of environmental consciousness in Singaporean households
No full textEnvironmental issues have been growing increasingly relevant in the past few decades. In Singapore, the government has recognised the pressing need to devote greater attention to environmental-related efforts, thus implementing various policies accordingly. Surveys conducted report that Singaporeans are highly environmentally conscious (Awang, 2021; Schneider Electric, 2022), yet it is unclear whether these reports accurately reflect current sentiments at the household level. As such, to determine if Singaporeans really possess high levels of environmental consciousness and whether governmental efforts have genuinely contributed to said levels, a series of ten interviews has been conducted as part of a qualitative study on Singaporean households. This paper adopts a novel methodological approach by examining environmental consciousness through three dimensions (cognitive, affective, and dispositional), finding that Singaporean households fall short in the cognitive and affective dimensions, with a middling performance in the dispositional dimension. Subsequently, this paper has included two policy recommendations aimed at simultaneously bolstering all three dimensions together, in hope that they may prove useful in truly elevating the environmental consciousness of Singaporean households in general.Bachelor's degre
Generate vulnerable transaction sequences for smart contract using large language models
No full textAs blockchain technologies expand, smart contract security remains critical, particularly in complex state-dependent exploits in decentralised finance (DeFi). Traditional vulnerability detection methods, like fuzzing and symbolic execution, often miss intricate, multi-step exploit scenarios, forcing reliance on costly manual audits. Addressing this gap, this research introduces a comprehensive framework utilizing advanced Large Language Models (LLMs), static analysis integration, a multi-agent system, and Retrieval-Augmented Generation (RAG). Evaluations on benchmark datasets (CTFBench) and real-world exploit scenarios (DeFiHackLabs) demonstrate the framework’s capability in accurately detecting vulnerabilities and generating executable Proof-of-Concept exploits. Key innovations include specialized multi-agent workflows, static analysis guidance, semantic grounding via RAG, and self-healing exploit refinement mechanisms. Future work includes curating multi-file evaluation datasets, fine-tuning reasoning models, parameter optimisation, and computational efficiency improvements, aiming to further enhance smart contract security and development practices.Bachelor's degre