DR-NTU (Digital Repository of NTU)
Not a member yet
116018 research outputs found
Sort by
Viscosities of hcp iron alloys under Earth's inner core conditions
No full textViscosity is critical for controlling the dynamics and evolution of the Earth's inner core (IC). The viscosities of hexagonal close-packed (hcp) and body-centred cubic (bcc) Fe were studied experimentally and theoretically under Earth's core conditions. However, Earth's inner core is mainly composed of Fe-Ni alloys with some light element impurities (Si, S, C, H, O), and the influence of impurities (Ni, Si, S, C, H, and O) on viscosity is still unknown. In this study, the diffusion coefficients of Fe, Ni, Si, S, C, H, and O were calculated under IC conditions using ab initio molecular dynamics (AIMD) and deep learning molecular dynamics (DPMD) methods. Among them, C, H, and O are highly diffusive like liquids in the lattice, while Fe, Ni, Si, and S diffuse through Fe site vacancies. In binary alloys, the influence of these impurities (Ni: 12.5%, S: 3.6%, Si: 3.1%, C: 1.3%, O: 1.7%, H: 0.4% by weight) on viscosity is insignificant. Based on the dislocation creep mechanism, the predicted viscosities of the hcp Fe alloys are 1 × 1014–2 × 1016 Pa·s, which is consistent with the values predicted by free inner core nutation and seismic wave attenuation observations.Published versionWe acknowledge the support of the National Natural Science Foundation of China (42350002, 42074104), the CAS Youth Interdisciplinary Team (JCTD-2022-16), and the Youth Innovation Promotion Association of CAS (2020394). This study was also supported by the Guizhou Provincial 2020 Science and Technology Subsidies (No. GZ2020SIG)
Still voices
No full textThe nature of this project is a 20-30 minute short film. "Still Voices" explores themes of
communication, grief, and intergenerational silence within a fractured family. Set largely within the confined space of an HDB flat, the film uses subtle visual cues and stillness to reflect the emotional distance between a mother and her non-verbal daughter. As the producer of this film, my responsibility was to shape and guide the project from inception to final delivery, ensuring that the team had the resources, timelines, and structure needed to bring the director's vision to life.Bachelor's degre
Solvent-responsive deformation of magnetic pollen platform for cargo transportation
No full textPollen paper, made from pollen grains, is an innovation that contributes to a sustainable future and mindset. Repurposing an abundant and renewable material, pollen paper has the potential to be used for applications in various fields. This study explores the application of pollen paper as a platform for underwater cargo transportation while also analyzing its properties. After bleaching, its unique solvent-responsive behaviour enables it to shrink and expand in response to different solvents. During the shrinking process, the pollen paper actuated and folded into a bucket to collect and transport small cargos. Additionally, magnetic nanoparticles were added to enable the controlled movement of pollen paper during transportation. The magnetic pollen paper was tested for its actuation mechanism and properties under different parameters and solvents to ensure that there was no modification to the paper's fundamental properties. The magnetic pollen paper was also tested for its biodegradability and recyclability to show and understand its sustainability. By studying and exploring the properties of bleached and unbleached, non-magnetic and magnetic pollen paper, this research aims to uncover their potential for actuation-based applications of a bio-inspired underwater cargo transport system.Bachelor's degre
Decision-making with cooperative multi-agent reinforcement learning
No full textReinforcement learning, a machine learning technique, has already shown advancement in solving complex sequential decision-making problems. Many tasks involve multiple agents and require sequential decision-making policies to achieve common goals, such as warehouse automation, autonomous driving, and game-playing. To derive the policies for all agents, these problems can be modelled as multi-agent systems and addressed by multi-agent reinforcement learning (MARL). However, optimizing policies in multi-agent scenarios presents significant challenges due to the intricate behaviours of multiple agents and the non-stationary nature of the environments’ complex dynamics. Firstly, optimizing policies in multi-agent scenarios presents significant challenges due to the complexity of multi-agent behaviours, especially in partially observable environments. Also, the dynamic nature of agents’ behaviours and their interactions with other agents lead to changes in the environment’s states and agents’ observations over time, which is more complicated
in open environments. Furthermore, the need to balance individual and collective objectives in certain real-world multi-agent environments also complicates the decision-making process. This doctoral thesis aims to tackle the following three fundamental multi-agent research problems and propose a solution for each. Our research covers from theoretical analysis to practical application.
We start by studying the problem of learning an efficient policy under partially
observable environments. We notice that a group of agents cooperates to com-
pete against another group of agents (named opponents), with the limitation that
information about the opponents is inaccessible due to partial observability. To address this issue, we propose a novel multi-agent distributional actor-critic algorithm to achieve speculative opponent modeling with purely local information (i.e., the controlled agent’s observations, actions, and rewards). The actor maintains a speculated belief of the opponents, which we call the speculative opponent models, to predict opponents’ actions using local observations and make decisions accordingly. The distributional critic models the return distribution of the policy. It reflects the quality of the actor and thus can guide the training of the imaginary opponent model that the actor relies on. Extensive experiments confirm that our method successfully models opponents’ behaviours without their data and delivers superior performance against baseline methods with a faster convergence speed.
Furthermore, in some environments, teammates’ numbers and policies change with the market, requiring workers to adapt to perform different sets of tasks across time. To solve this issue, we propose an RL-based method that allows the controlled agent to collaborate with dynamic teammates in open environments. The controlled agent maintains a dual teamwork situation inference model to capture the current teamwork state and facilitate reasonable decision-making under partial observability. Considering the dynamic types of teammates, we first leverage the Chinese Restaurant Process-based model to categorize versatile teammate policies into distinct clusters, which improves the efficiency of identifying current teamwork situations. Next, to model the heterogeneous relationships among agents and accommodate the varying number of teammates, we apply heterogeneous graph attention neural networks to learn the representation of the teamwork situation. Extensive experiments confirm that our method outperforms state-of-the-art baselines with faster convergence in various ad hoc teamwork tasks.
Lastly, in certain real-world applications, such as routing problems and warehouse management, decision-makers must balance overall benefits with individual fairness among agents. Achieving both learning efficiency and fairness concurrently presents a complex, multi-objective, joint-policy optimization challenge. Moreover, existing approaches are predominantly confined to simulation environments. To address the above issues, we present a pioneering MARL approach to balance individual and collective objectives for agents collaboration. Extensive experiments on synthetic and real-world datasets demonstrate that our method not only surpasses the state-of-the-art DRL methods, but also exhibits advantages over established heuristics, notably in achieving significantly faster optimization speeds. This method underlines the essential integration of fairness into real-world applications, representing a major advancement in creating equitable and efficient logistics solutions.
To conclude, this doctoral thesis investigates three fundamental multi-agent decision-making research problems that are ubiquitous and unsolved. The proposed three MARL methodology solutions achieve efficient policy training and performance for agents in multi-agent environments with uncertainties raised by the partially observable, the open environment, and the individual-collective objectives of MARL. This thesis delves into novel designs of various critical components of MARL, including MDP formulations, policy networks, training algorithms, and inference methods. These contributions significantly elevate the effectiveness and efficiency of cooperative MARL, establishing new performance benchmarks.Doctor of Philosoph
Hardware development of non-contact online impedance analyzer
No full textThis dissertation presents the development of a non-contact, real-time impedance measurement system that addresses the limitations of traditional sinusoidal signal-based methods by introducing multi-harmonic signals such as square and triangular waves at a test frequency of 100 kHz. The proposed methodology combines advanced signal processing techniques, including moving-window discrete Fourier transform (MWDFT), and precise calibration procedures to achieve high accuracy and efficiency in dynamic and high-frequency environments. Experimental results demonstrated that while sinusoidal signals provided reliable accuracy with an average deviation of 1.60%, square and triangular waves significantly improved measurement efficiency and accuracy, achieving deviations of 1.48% and 1.38%, respectively. Triangular waves emerged as the most effective signal type due to their smooth harmonic spectrum and reduced noise sensitivity. The proposed system's ability to provide rapid and accurate impedance tracking highlights its suitability for applications in power electronics, biomedical diagnostics, and material science. The findings underscore the potential of multi-harmonic signals to transform impedance measurement practices, offering a robust, scalable solution for real-time applications and paving the way for future advancements in the field.Master's degre
Enhancing financial market volatility prediction: a machine learning approach integrating sentiment analysis and macroeconomic indicators
No full textThis report aims to explore the integration of sentiment-derived features with machine-
learning (ML) algorithms – LightGBM (LGBM) and XGBoost to improve volatility
forecasting. Financial market volatility forecasting continues to be a significant challenge in
risk management and investment decision-making.
In this study, sentiment data from Reddit posts, representing retail investor sentiment, and
The Wall Street Journal, reflecting institutional perspectives alongside historical market
index data from 2022 to 2024 were added into our ML models to assess their impact on
volatility forecasting. Utilising these ML models, the study evaluates whether sentiment
enhanced models offer additional predictive power beyond baseline models that rely solely
on historical data and traditional econometric approaches, such as EGARCH.
The results show that LGBM and XGBoost models consistently outperform EGARCH,
demonstrating lower mean squared errors (MSE). Incorporating sentiment data from both
Reddit and the Wall Street Journal with lagged historical data significantly improves
predictive accuracy, achieving the lowest MSE for both Dow Jones and S&P 500 indexes.
Additionally, while macroeconomic indicators contribute to model performance, they are less
effective than historical market trends in short-term volatility forecasting. Diebold-Mariano
test results along with a simulation study reinforces the statistical significance of machine
learning models' enhanced accuracy over EGARCH.
These findings strengthen the case for integrating machine learning techniques into volatility
forecasting as viable alternatives to traditional econometric models in financial applicationsBachelor's degre
Multi-agent solution for airdrop hunting on social network and blockchain
No full textIn this project, we propose the development of an advanced multi-agent system designed to optimize the discovery and acquisition of airdrops on online social networks and blockchains. Central to this system is a principal agent responsible for orchestrating the collaborative efforts of subordinate AI agents, allocating tasks, and synthesizing outcomes. This architecture allows for a division of labor among the agents, wherein specific agents are dedicated to gathering relevant airdrop information on online social networks and executing on-chain operations. Our approach leverages the strengths of distributed intelligence to efficiently identify and exploit airdrop opportunities in the dynamic blockchain community.Bachelor's degre
Wearable neuromorphic ion circuits
No full textWearable neuromorphic ion circuit is a frontier research field which includes advanced ion materials, flexible electronics, artificial synaptic device, biological emulation, human information and neuron networks. Traditional circuits have tremendous restrictions when they are applied to the bare surface of human skin. In this case, ion circuits, especially neuromorphic form, have take the advantage of its biocompatibility and neuron-like characteristics to execute many information collective works on the surface of skins.
Neuromorphic ion components constitute the very basic of ion circuits and play a vital role in the stability and performance for them. In these components, artificial synaptic device is a key pillar which has the ability to reproduce the phenomena in a real neuron synapse and is the main topic this dissertation will cover.
This dissertation talks about comprehensive aspects of a Mg-ion artificial synaptic device, the device is comprised of a three-layer structure (CNTF, aqueous MgCl2 solution and NiOOH/CNTF) and shows excellent synaptic behaviors.
To start, a necessary introduction of wearable neuromorphic ion circuits and the components will be discussed. Relative literatures that contribute to the previous fundamental labor are reviewed to get lasted ideas and insights. Then, material characterizations and electrochemical properties of NiOOH/CNTF are given. Moreover, the hippocampal neuron LTP is realized by the device. Furthermore, mechanism of working synaptic simulation is introduced to achieve controllable synaptic plasticity for voltage input case. In the final section, applications consist of mental state recording (current input, voltage output) and sensor signal (voltage input, current output) are presented.Master's degre
Retracing the footsteps of Tudi Gong and Tua Pek Kong: diffusion and evolution across South China and localisation in the Nanyang
No full textTua Pek Kong is a popular deity worshipped across Southeast Asia. Every temple venerating Tua Pek Kong has its own interpretation of his origins, beliefs and worship practices. The belief of Tua Pek Kong had evolved across the migratory networks of South Chinese migrants to the Nanyang, but the evolution of the deity first begun in South China, as the different dialect subcommunities migrated southwards and eastwards and acculturated to the dominant culture in the host regions. As such, it is difficult to trace the origins of the deity by observing an individual temple in either Southeast Asia or South China and instead requires a multi-sited ethnographic approach that considers the historical continuity and evolution of the belief along the migratory routes of the Hakka, Teochew and Minnan subcommunities. The present study sheds light on the “genealogical link” between Tudi Gong, the prototype in South China, and Tua Pek Kong, and evaluates that the Tua Pek Kong belief is neither wholly a cultural inheritance transmitted in its entirety by the Chinese migrants, nor a native creation devoid of the cultural link. Instead, the translocal and transnational diffusion has transformed him into a “New Tua Pek Kong”, a locally rooted divinity.Bachelor's degre
Simulation of stably controlled biped walking by educational humanoid robot
No full textThe study of humanoid robots has become an important direction in the research of
robotics, and has been extensively applied across various areas such as the service
industry and healthcare. Among these applications, the “humanoid” features of
humanoid robots give them greater affinity and interactivity in the educational
applications than other robots. Given that educational robots are mainly designed for
children and students, research on the motion stability and safety is very important.
In this case, this dissertation reviews the development history of humanoid robot
technology and investigates into several theoretical methods of stability control for
biped walking robots. Based on the Linear Inverted Pendulum Model (LIPM), this
dissertation conducts a walking simulation of a biped model in MATLAB using the
theory of orbital energy. The simulation analyzes the impact of various walking
parameters on the walking control of biped walking robots and explores the
approaches to achieve a stable walking process.Master's degre