About frankielei

This author has not yet filled in any details.
So far frankielei has created 0 blog entries.
13 2024-11

IOTSC Distinguished Visiting Scholar Series: AI-empowered Engineering Solutions

2024-12-13T00:01:15+08:00

Dear Colleagues and Students,

The State Key Laboratory of Internet of Things for Smart City would like to invite you to join our “IOTSC Distinguished Visiting Scholar Series” on 19/11/2024 (Tuesday). We are pleased to invite Prof. Michael Beer from Leibniz Universität Hannover as the speaker.

AI-empowered Engineering Solutions
Speaker: Michael Beer
Date: 19/11/2024 (Tuesday)
Time: 10:00 – 11:00
Language: English
Venue: N21-5010

Abstract:
Engineering analyses are associated with two key challenges; they must be realistic and numerically efficient. A realistic analysis requires a proper description of the physics of the underlying problem in the numerical model. In the case of complex problems or physics this can easily lead to a quite high computational cost in order to arrive at reasonably realistic results. If the available information about the physics and the problem is vague and limited, a numerical model cannot be formulated with sufficient confidence. In engineering design additional requirements need to be considered to ensure products to serve their purpose. This includes robust design to compensate deviations from normal conditions and even unforeseen events. Also, decision margins are often desired to provide flexibility in variant development and more freedom in use. Challenges are then to translate the requirements into numerical descriptions, to identify the most suitable design solutions that meet the various requirements, to find variants thereof and to compare them with one another. In these contexts, engineers have sought help from computational intelligence in various forms and for various purposes. The seminar provides insight in civil and mechanical engineering approaches to develop solutions to the described challenges with the aid of computational intelligence. Selected developments are discussed with focus on the added value for engineering analyses and are demonstrated on industrial examples. These developments include processing of vague information as fuzzy sets with evolutionary concepts and their use in design, efficient stochastic analysis with meta models and process simulation based on neural networks, robust design and identification of critical mechanical behavior with the aid of cluster analysis methods. In addition, the scope of the approaches is expanded to systems engineering, demonstrating advancements on computational efficiency analyzing complex systems reliability problems with methods of computational intelligence. The examples include dynamical analyses of civil engineering structures and of an aerospace structure, nonlinear dynamical problems in crashworthiness analysis, as well as civil infrastructure systems.

Speaker’s Bio:
Michael Beer is Professor and Head of the Institute for Risk and Reliability, Leibniz Universität Hannover, Germany. He is also part time Professor at the University of Liverpool and guest Professor at Tongji University and Tsinghua University, China. He obtained a doctoral degree from Technical University Dresden, Germany, and worked for Rice University, National University of Singapore, and the University of Liverpool, UK. Dr. Beer’s research is focused on uncertainty quantification in engineering with emphasis on imprecise probabilities. Dr. Beer is Editor in Chief of the ASCE-ASME Journal of Risk and Uncertainty in Engineering Systems, Part A Civil Engineering and Part B Mechanical Engineering. He is also Editor in Chief (joint) of the Encyclopedia of Earthquake Engineering, Associate Editor of Information Sciences, and Editorial Board Member of Engineering Structures and several other international journals. He has won several awards including the Alfredo Ang Award on Risk Analysis and Management of Civil Infrastructure of ASCE. Dr. Beer is the Chairman of the European Safety and Reliability Association (ESRA) and a Co-Chair of Risk and Resilience Measurements Committee (RRMC), Infrastructure Resilience Division (IRD), ASCE. He is serving on the Executive Board of the International Safety and Reliability Association (IASSAR), on the Executive Board of the European Association of Structural Dynamics (EASD), and on the Board of Directors of the International Association for Probabilistic Safety Assessment and Management (IAPSAM). He is a Fellow of the Alexander von Humboldt-Foundation and a Member of ASCE (EMI), ASME, CERRA, IACM and GACM. 

 

For enquiries: Tel: 8822 9159
Email: frankielei@um.edu.mo

Best Regards,
State Key Laboratory of Internet of Things for Smart City

IOTSC Distinguished Visiting Scholar Series: AI-empowered Engineering Solutions2024-12-13T00:01:15+08:00
8 2024-11

IOTSC Distinguished Visiting Scholar Series: Machine Learning in the Air: From Semantic Communications to Distributed Learning and Inference

2024-12-08T00:00:08+08:00

Dear Colleagues and Students,

The State Key Laboratory of Internet of Things for Smart City would like to invite you to join our “IOTSC Distinguished Visiting Scholar Series” on 11/11/2024 (Monday). We are pleased to invite Prof. Deniz Gündüz from Imperial College London as the speaker.

Machine Learning in the Air: From Semantic Communications to Distributed Learning and Inference
Speaker: Prof. Deniz Gündüz
Date: 11/11/2024 (Monday)
Time: 15:00 – 16:00
Language: English
Venue: N21-5007 (Exhibition Hall)

Abstract:
Machine learning is poised to transform wireless communication technologies in the next decade, and seamless integration of AI technologies into 6G networks will be a transformative leap beyond previous generations. Towards this future, synergies between communications and machine learning are two-fold: data-driven methods coupled with established model driven solutions provide novel approaches and state-of-the art results to many long-standing challenges in communications, from coding to resource optimization. On the other hand, distributed learning and inference problems introduce new challenges imposing more stringent reliability and latency constraints on future communication networks. In this talk, I will touch upon both aspects, and present my group’s recent contributions spanning data compression, communications, and distributed learning and inference problems, bringing together tools from information and coding theories with recent advances in machine learning.

Speaker’s Bio:
Deniz Gündüz received the Ph.D. degree in electrical engineering from NYU Tandon School of Engineering in 2007. He is a Professor of Information Processing at Imperial College London. He held various positions at the University of Modena and Reggio Emilia, University of Padova, CTTC, Princeton University, and Stanford University. Deniz Gündüz is a Fellow of the IEEE. He serves as an Area Editor for the IEEE Transactions on Information Theory and IEEE Transactions on Communications. He is the recipient of the IEEE Communications Society CTTC Early Achievement Award in 2017, Starting (2016) and Consolidator (2022) and Proof-of-Concept (2023) Grants of the European Research Council (ERC), and has co-authored several award-winning papers, including the IEEE Communications Society – Young Author Best Paper Award (2022), and IEEE ICC Best Paper Award (2023).

 

For enquiries: Tel: 8822 9159
Email: frankielei@um.edu.mo

Best Regards,
State Key Laboratory of Internet of Things for Smart City

IOTSC Distinguished Visiting Scholar Series: Machine Learning in the Air: From Semantic Communications to Distributed Learning and Inference2024-12-08T00:00:08+08:00
8 2024-11

IOTSC TALK SERIES: High Frequencies, Higher Potential: Redefining Movable Antennas with Polarization

2024-12-08T00:00:09+08:00

Dear Colleagues and Students,

The State Key Laboratory of Internet of Things for Smart City would like to invite you to join our “IOTSC Talk Series” on 12/11/2024 (Tuesday). We are pleased to invite Prof. Yulin SHAO from State Key Laboratory of Internet of Things for Smart City as the speaker.

High Frequencies, Higher Potential: Redefining Movable Antennas with Polarization
Speaker: Yulin SHAO
Date: 12/11/2024 (Tuesday)
Time: 15:00 – 16:00
Language: English
Venue: N21-5007 (Exhibition Hall)

Abstract:
This talk introduces the polarization-aware movable antenna (PAMA) framework, a novel approach that incorporates polarization effects into the design and optimization of movable antennas (MAs). While MAs have shown promise in enhancing wireless communication performance, existing research largely overlooks the impact of polarization, especially at higher frequencies where line-of-sight (LOS) links dominate. By rigorously modeling three-dimensional (3D) antenna translation and rotation with explicit consideration of polarization, PAMA demonstrates that significant performance gains can be achieved even in higher frequency bands by facilitating optimal polarization matching between transmitter and receiver. This advancement extends the effectiveness of MAs beyond low-frequency, multipath-rich environments and highlights the crucial role of incorporating polarization considerations into the system design. Our findings pave the way for enhanced efficiency, better link reliability, and increased data throughput in future wireless networks with MAs.

Speaker’s Bio:
Yulin Shao is an Assistant Professor with the University of Macau, and a Visiting Researcher with Imperial College London. He received the Ph.D. degree from the Chinese University of Hong Kong in 2020. He was a Research Assistant with the Institute of Network Coding, a Visiting Scholar with Massachusetts Institute of Technology, a Research Associate with Imperial College London, and a Lecturer with University of Exeter. He was a Guest Lecturer at 5G Academy Italy and IEEE Information Theory Society Bangalore Chapter. Dr. Shao is an Editor for IEEE Communications Magazine, IEEE Transactions on Communications, and IEEE Communications Letters. He received the Best Paper Awards at IEEE ICC 2023, and IEEE WCNC 2024.

For enquiries: Tel: 8822 9159
Email: frankielei@um.edu.mo

Best Regards,
State Key Laboratory of Internet of Things for Smart City

IOTSC TALK SERIES: High Frequencies, Higher Potential: Redefining Movable Antennas with Polarization2024-12-08T00:00:09+08:00
5 2024-11

IOTSC Postgraduate Forum: Urban Safety and Disaster Prevention

2024-12-05T00:01:05+08:00

Dear Colleagues and Students,

The State Key Laboratory of Internet of Things for Smart City would like to invite you to join our “IOTSC Postgraduate Forum” on 19/11/2024 (Tuesday). The event aims to bring together postgraduate students from various disciplines to share their research, exchange ideas, and engage in meaningful discussions. We are pleased to invite six outstanding PhD students to give presentations related to Urban Safety and Disaster Prevention.

IOTSC Postgraduate Forum: Urban Safety and Disaster Prevention
Date: 19/11/2024 (Tuesday)
Time: 15:00 – 17:00
Language: English
Venue: N21-5007 (Exhibition Hall)
Moderator: Prof. Huabin SHI

Presenters Abstract
Shuangao YAO

Bayesian Selective Surrogate Gaussian Process for Structural Health Indicator Modeling

The Bayesian selective surrogate Gaussian process (BSSGP), as a novel progressive probabilistic approach, is proposed for nonparametric modeling. In Gaussian process (GP) regression, the kernel for specifying the similarity between the input data is commonly assigned through trial and error with the captured input variables. However, an improper choice of the kernel type and/or redundant input variables can significantly degrade the modeling performance. To address this problem, the proposed approach provides a generative kernel augmentation scheme to develop the optimal kernel with the appropriate input variables. The scheme starts with a candidate kernel set. By adopting more features, these candidates evolve as augmented kernels. A Bayesian indicator is formulated to assess the performance of the potential kernels. Hence, the set of kernels that strike the optimal balance between fitting capacity and robustness is chosen for further enhancement. The generation procedure is conducted iteratively until further augmentation ceases to provide considerable improvement in the kernel performance. To illustrate the efficacy of the proposed approach, a case study with three-year continuous monitoring of a 22-story reinforced concrete building is presented.
Cong TANG

Probabilistic Settlement Prediction and System Serviceability Assessment of Immersed Tunnel

Based on the engineering background of the Hong Kong-Zhuhai-Macao Bridge (HZMB) immersed tunnel, a simple and efficient method for predicting tunnel settlement is proposed. This method employs Bayesian inference combined with field measurement data to update the model and quantify uncertainties, enabling accurate prediction of tunnel settlement. Furthermore, drawing on reliability theory and considering multiple failure modes related to differential settlement of tubes and joints, a framework is established for evaluating the system serviceability performance of the immersed tunnel.
Linfeng MEI

Online Bayesian Learning for Uncertainty-Aware Anomaly Detection with Streaming Transmissibility Measurements

Transmissibility function (TF) is widely applied in structural anomaly detection due to its sensitivity to damage and robustness against external excitations, while its application in online detection is limited, primarily due to challenges in handling data streams. To address this issue, a Bayesian nonparametric clustering approach is introduced for online structural anomaly detection with TF, which integrates a truncation-free variational inference-based Dirichlet process Gaussian mixture model (VI-DPGMM) within a streaming variational inference (SVI) framework. As a Bayesian nonparametric approach, the truncation-free VI-DPGMM dynamically adapts model complexity during online learning with increasing data by setting the variational distributions of parameters for components without data to their prior distributions, which enables computing the probabilities to assign data points to these components and determining the creation of new components. As a result, the truncation-free VI-DPGMM allows dynamically adding components to the mixture model, providing the flexibility to automatically adapt the number of components for arbitrary amounts of data. This characteristic enables its intuitive integration into the SVI featured as the variational posterior conditioned on the previous data as the prior when new data are observed, facilitating continuous refinement of the mixture model without repeatedly making inference of previous data. Therefore, the proposed method is highly efficient and well-suited for online structural anomaly detection. Moreover, as a density estimator, this approach supports outlier-based anomaly detection, providing robustness to environmental effects and comprehensive uncertainty quantification.
Jiaming LI

Physical-Knowledge-Guided and Explainable CNNs for Gear Fault Severity Level Diagnosis

Despite much success that deep-learning (DL) models have achieved, their interpretability, however, remains opaque and thus often impedes their real applications in the scenarios of fault diagnosis for risk-sensitive equipment. Fortunately, Class Activation Mapping (CAM) has significantly enhanced the interpretability of convolutional neural networks (CNNs) because it can perform saliency analysis and then illuminate which areas of an input contribute more to a classification result c. Nevertheless, CAM fails to enhance diagnostic accuracy and actively leverage interpretability due to its passively explanatory property for the trained models. To address this issue, in this paper, a Physically Meaningful Regularization (PMR) term is proposed by using Grad-CAM, to guide the models in focusing on the same frequency bands of the input spectra and ignoring other parts of noisy and irrelevant signals. Based on the PMR term, a two-step back propagation training algorithm is accordingly proposed to train the diagnostic models embedded with physical knowledge. Consequently, the obtained physical-knowledge-guided and explainable CNNs can offer not only higher interpretability but also stronger anti-noise ability. Finally, the proposed method is validated in two datasets containing multiple fault severity levels. The diagnostic results and the saliency analysis demonstrate the efficacy of the proposed method.
Wenjing ZHANG

Two Stage Nonparametric Framework for Missing Data Imputation, Uncertainty Quantification and Incorporation in System Identification

In many engineering applications, missing data during system identification can hinder the performance of the identified model. The proposed framework is a two-stage approach where the first stage is to impute the missing values in the dataset with plausible values inferred from the observed data and quantify the corresponding imputation uncertainty. The second stage proposes to incorporate imputation uncertainty in a single modeling process by mitigating the influence of samples with imputations during training and testing. The framework is applied to the structural health monitoring data of East Asian Hall. Results reveal a minimum reduction of 50% in root mean squared error values compared to those achieved by directly removing incomplete samples.

 

For enquiries: Tel: 8822 9159
Email: frankielei@um.edu.mo

Best Regards,
State Key Laboratory of Internet of Things for Smart City

IOTSC Postgraduate Forum: Urban Safety and Disaster Prevention2024-12-05T00:01:05+08:00
5 2024-11

IOTSC TALK SERIES: High Frequencies, Higher Potential: Redefining Movable Antennas with Polarization

2024-12-05T00:01:05+08:00

Dear Colleagues and Students,

The State Key Laboratory of Internet of Things for Smart City would like to invite you to join our “IOTSC Talk Series” on 12/11/2024 (Tuesday). We are pleased to invite Prof. Yulin SHAO from State Key Laboratory of Internet of Things for Smart City as the speaker.

High Frequencies, Higher Potential: Redefining Movable Antennas with Polarization
Speaker: Yulin SHAO
Date: 12/11/2024 (Tuesday)
Time: 15:00 – 16:00
Language: English
Venue: N21-5007 (Exhibition Hall)

Abstract:
This talk introduces the polarization-aware movable antenna (PAMA) framework, a novel approach that incorporates polarization effects into the design and optimization of movable antennas (MAs). While MAs have shown promise in enhancing wireless communication performance, existing research largely overlooks the impact of polarization, especially at higher frequencies where line-of-sight (LOS) links dominate. By rigorously modeling three-dimensional (3D) antenna translation and rotation with explicit consideration of polarization, PAMA demonstrates that significant performance gains can be achieved even in higher frequency bands by facilitating optimal polarization matching between transmitter and receiver. This advancement extends the effectiveness of MAs beyond low-frequency, multipath-rich environments and highlights the crucial role of incorporating polarization considerations into the system design. Our findings pave the way for enhanced efficiency, better link reliability, and increased data throughput in future wireless networks with MAs.

Speaker’s Bio:
Yulin Shao is an Assistant Professor with the University of Macau, and a Visiting Researcher with Imperial College London. He received the Ph.D. degree from the Chinese University of Hong Kong in 2020. He was a Research Assistant with the Institute of Network Coding, a Visiting Scholar with Massachusetts Institute of Technology, a Research Associate with Imperial College London, and a Lecturer with University of Exeter. He was a Guest Lecturer at 5G Academy Italy and IEEE Information Theory Society Bangalore Chapter. Dr. Shao is an Editor for IEEE Communications Magazine, IEEE Transactions on Communications, and IEEE Communications Letters. He received the Best Paper Awards at IEEE ICC 2023, and IEEE WCNC 2024.

For enquiries: Tel: 8822 9159
Email: frankielei@um.edu.mo

Best Regards,
State Key Laboratory of Internet of Things for Smart City

IOTSC TALK SERIES: High Frequencies, Higher Potential: Redefining Movable Antennas with Polarization2024-12-05T00:01:05+08:00
23 2024-10

IOTSC TALK SERIES: Unlocking the demand side: virtual power plants, energy communities and microgrids

2024-11-23T00:00:58+08:00

Dear Colleagues and Students,

The State Key Laboratory of Internet of Things for Smart City would like to invite you to join our “IOTSC Talk Series” on 29/10/2024 (Tuesday). We are pleased to invite Prof. Pierluigi Mancarella from University of Melbourne as the speaker.

Unlocking the demand side: virtual power plants, energy communities and microgrids
Speaker: Prof. Pierluigi Mancarella
Date: 29/10/2024 (Tuesday)
Time: 10:00 – 11:30
Language: English
Venue: N21-5007 (Exhibition Hall)

Abstract:
Distributed Energy Resources (DER) are becoming widespread in most countries worldwide. In particular, at the household level rooftop solar photovoltaics and battery energy storage systems, and soon electric vehicles too, are poised to play an essential role towards power system decarbonisation and more consumer-centric markets. Local energy systems with shared assets such as community storage are also emerging as important options of interest. Key ongoing debates now refer to how different DER technologies and schemes should be fully integrated into power and energy system and market operation.
Fundamental techno-economic aspects of key aggregation approaches that could enable large-scale deployment of DER, namely, virtual power plants, energy communities, and microgrids will be presented. More specifically, we will discuss the key features of these different aggregation models (“what”), under which conditions each of them should/could be considered (“where”), and potential technical and commercial opportunities and challenges for different stakeholders involved with DER development, e.g., aggregators, energy managers, distribution system operators, and so forth (“who”). Several case studies from a number of recent projects in Australia, UK, Europe, and around the world will be used to exemplify the concepts presented.

Speaker’s Bio:
Pierluigi Mancarella is the Chair Professor of Electrical Power Systems at the University of Melbourne, Australia, and Professor of Smart Energy Systems at the University of Manchester, UK. His key research interests include techno-economic modelling and analysis of multi-energy systems, grid integration of renewables and distributed energy resources, energy infrastructure planning under uncertainty, and security, reliability, and resilience of low-carbon networks.
Pierluigi is the Energy Systems Program Leader at the Melbourne Energy Institute, a Fellow of the IEEE (Institute of Electrical and Electronics Engineers), an IEEE Power and Energy Society Distinguished Lecturer, the Convenor of the CIGRE (International Council of Large Electric Systems) C6/C2.34 Working Group on “Flexibility Provision from Distributed Energy Resources”, holds the 2017 veski innovation fellowship for his work on urban-scale virtual power plants, and is a recipient of the international Newton Prize 2018 for his work on power system resilience in Chile. He is author of several books and over 300 research papers and reports, and is a Senior Editor of the IEEE Transactions on Power Systems, and the Oxford Open Energy journal.

For enquiries: Tel: 8822 9159
Email: frankielei@um.edu.mo

Best Regards,
State Key Laboratory of Internet of Things for Smart City

IOTSC TALK SERIES: Unlocking the demand side: virtual power plants, energy communities and microgrids2024-11-23T00:00:58+08:00
15 2024-10

IOTSC Postgraduate Forum: Intelligent Transportation

2024-11-14T00:00:49+08:00

Dear Colleagues and Students,

The State Key Laboratory of Internet of Things for Smart City would like to invite you to join our “IOTSC Postgraduate Forum” on 15/10/2024 (Tuesday). The event aims to bring together postgraduate students from various disciplines to share their research, exchange ideas, and engage in meaningful discussions. We are pleased to invite six outstanding PhD students to give presentations related to Intelligent Transportation.

IOTSC Postgraduate Forum: Intelligent Transportation
Date: 15/10/2024 (Tuesday)
Time: 14:20 – 17:00
Language: English
Venue: N21-5007 (Exhibition Hall)
Host: Prof. Zhenning LI

Presenters Abstract
Yiming WANG

Rethinking Exploration in Reinforcement Learning with Effective Metric-Based Exploration Bonus

Enhancing exploration in reinforcement learning (RL) through the incorporation of intrinsic rewards, specifically by leveraging  state discrepancy measures within various metric spaces as exploration bonuses, has emerged as a prevalent strategy to encourage agents to visit novel states. The critical factor lies in how to quantify the difference between adjacent states as novelty for promoting effective exploration. Nonetheless, existing methods that evaluate state discrepancy in the latent space under  or  norm often depend on count-based episodic terms as scaling factors for exploration bonuses, significantly limiting their scalability. Additionally, methods that utilize the bisimulation metric for evaluating state discrepancies face a theory-practice gap due to improper approximations in metric learning, particularly struggling with hard exploration tasks. To overcome these challenges, we introduce the Effective Metric-based Exploration-bonus (EME). EME critically examines and addresses the inherent limitations and approximation inaccuracies of current metric-based state discrepancy methods for exploration, proposing a robust metric for state discrepancy evaluation backed by comprehensive theoretical analysis. Furthermore, we propose the diversity-enhanced scaling factor integrated into the exploration bonus to be dynamically adjusted by the variance of prediction from an ensemble of reward models, thereby enhancing exploration effectiveness in particularly challenging scenarios. Extensive experiments are conducted on hard exploration tasks within Atari games, Minigrid, Robosuite, and Habitat, which illustrate our method’s scalability to various scenarios, including pixel-based observations, continuous control tasks, and simulations of realistic environments.
Haicheng LIAO

Towards Human-like Trajectory Prediction for Autonomous Vehicles: A Behavior-aware Approach

Accurately predicting the trajectories of surrounding vehicles is essential for safe and efficient autonomous driving. This paper introduces a novel behavior-aware trajectory prediction model (BAT) that incorporates insights from traffic psychology, human behavior, and decision-making. BAT seamlessly integrates four modules: behavior-aware, interaction-aware, priority-aware, and position-aware. These modules perceive and understand underlying interactions and account for uncertainty and variability in prediction, enabling higher-level learning and flexibility without rigid categorization of driving behavior. Importantly, BAT eliminates the need for manual labeling in the training process and addresses the challenges of non-continuous behavior labeling and the selection of appropriate time windows. Furthermore, we evaluate BAT’s performance across the Next Generation Simulation (NGSIM), Highway Drone (HighD), Roundabout Drone (RounD), and Macao Connected Autonomous Driving (MoCAD) datasets, showcasing its superiority over prevailing state-of-the-art (SOTA) benchmarks in terms of prediction accuracy and efficiency. Remarkably, even when trained on reduced portions of the training data (25%) and with a much smaller number of parameters, BAT outperforms most of the baselines, demonstrating its robustness and efficiency in challenging traffic scenarios, including highways, roundabouts, campuses, and busy urban locales. This underlines its potential to reduce the amount of data required to train autonomous vehicles, especially in corner cases. In conclusion, the behavior-aware model represents a significant advancement in the development of autonomous vehicles capable of predicting trajectories with the same level of proficiency as human drivers.
Wencheng HAN

Learning 3D Geometry from Visual Consistency for Monocular Vision Tasks in Self-Driving Cars

Monocular vision tasks, like depth estimation and 3D object detection in self-driving cars, face challenges due to the limited 3D information in 2D images. This highlights our recent achievements that leverage visual consistency to enhance monocular vision models. We first introduce the Direction-aware Cumulative Convolution Network (DaCCN) for self-supervised depth estimation. DaCCN addresses direction sensitivity and environmental dependency, improving feature representation and achieving state-of-the-art performance. Next, we present the Rich-resource Prior Depth estimator (RPrDepth), which uses rich-resource data as prior information. RPrDepth estimates depth from a single low-resolution image by referencing pre-extracted features. Finally, we discuss a weakly-supervised approach for monocular 3D object detection that relies only on 2D labels. By using spatial and temporal view consistency, this method achieves results comparable to fully supervised models and enhances performance with minimal labeled data.
Chunlin TIAN

An Asymmetric LoRA Architecture for Efficient Fine-Tuning

Adapting Large Language Models (LLMs) to new tasks through fine-tuning has been made more efficient by the introduction of Parameter-Efficient Fine-Tuning (PEFT) techniques, such as LoRA. However, these methods often underperform compared to full fine-tuning, particularly in scenarios involving complex datasets. This issue becomes even more pronounced in complex domains, highlighting the need for improved PEFT approaches that can achieve better performance. Through a series of experiments, we have uncovered two critical insights that shed light on the training and parameter inefficiency of LoRA. Building on these insights, we have developed HydraLoRA, a LoRA framework with an asymmetric structure that eliminates the need for domain expertise. Our experiments demonstrate that HydraLoRA outperforms other PEFT approaches, even those that rely on domain knowledge during the training and inference phases.
Tianxiao GAO

Night-Voyager: Consistent and Efficient Nocturnal Vision-Aided State Estimation in Cross-Modal Maps

Accurate and robust state estimation at nighttime is essential for autonomous navigation of mobile robots to achieve nocturnal or round-the-clock tasks. An intuitive and practical question arises: Can low-cost standard cameras be exploited for nocturnal state estimation? Regrettably, most current visual methods tend to fail due to adverse illumination conditions, even with active lighting or image enhancement. A crucial insight, however, is that the streetlights in most urban scenes can provide static and salient visual information at night. This inspires us to design an object-level nocturnal vision-aided state estimation framework, named Night-Voyager, which leverages cross-modal maps and keypoints to enable the versatile all-day localization. Night-Voyager starts with a fast initialization module which solves the global localization problem. With the effective two-stage cross-modal data association approach, the system state can be accurately updated by the map-based observations. Meanwhile, to address the challenge of large uncertainties in visual observations during nighttime, a novel matrix Lie group formulation and a feature-decoupled multi-state invariant filter are proposed for consistent and efficient state estimation. Comprehensive experiments on both simulation and diverse real-world scenarios (about 12.3 km total distance) showcase the effectiveness, robustness, and efficiency of Night-Voyager, filling a critical gap in nocturnal vision-aided state estimation.

 

For enquiries: Tel: 8822 9159
Email: frankielei@um.edu.mo

Best Regards,
State Key Laboratory of Internet of Things for Smart City

IOTSC Postgraduate Forum: Intelligent Transportation2024-11-14T00:00:49+08:00
7 2024-10

IOTSC Postgraduate Forum: Intelligent Transportation

2024-11-07T00:00:44+08:00

Dear Colleagues and Students,

The State Key Laboratory of Internet of Things for Smart City would like to invite you to join our “IOTSC Postgraduate Forum” on 15/10/2024 (Tuesday). The event aims to bring together postgraduate students from various disciplines to share their research, exchange ideas, and engage in meaningful discussions. We are pleased to invite six outstanding PhD students to give presentations related to Intelligent Transportation.

IOTSC Postgraduate Forum: Intelligent Transportation
Date: 15/10/2024 (Tuesday)
Time: 14:20 – 17:00
Language: English
Venue: N21-5007 (Exhibition Hall)
Host: Prof. Zhenning LI

Presenters Abstract
Yiming WANG

Rethinking Exploration in Reinforcement Learning with Effective Metric-Based Exploration Bonus

Enhancing exploration in reinforcement learning (RL) through the incorporation of intrinsic rewards, specifically by leveraging  state discrepancy measures within various metric spaces as exploration bonuses, has emerged as a prevalent strategy to encourage agents to visit novel states. The critical factor lies in how to quantify the difference between adjacent states as novelty for promoting effective exploration. Nonetheless, existing methods that evaluate state discrepancy in the latent space under  or  norm often depend on count-based episodic terms as scaling factors for exploration bonuses, significantly limiting their scalability. Additionally, methods that utilize the bisimulation metric for evaluating state discrepancies face a theory-practice gap due to improper approximations in metric learning, particularly struggling with hard exploration tasks. To overcome these challenges, we introduce the Effective Metric-based Exploration-bonus (EME). EME critically examines and addresses the inherent limitations and approximation inaccuracies of current metric-based state discrepancy methods for exploration, proposing a robust metric for state discrepancy evaluation backed by comprehensive theoretical analysis. Furthermore, we propose the diversity-enhanced scaling factor integrated into the exploration bonus to be dynamically adjusted by the variance of prediction from an ensemble of reward models, thereby enhancing exploration effectiveness in particularly challenging scenarios. Extensive experiments are conducted on hard exploration tasks within Atari games, Minigrid, Robosuite, and Habitat, which illustrate our method’s scalability to various scenarios, including pixel-based observations, continuous control tasks, and simulations of realistic environments.
Haicheng LIAO

Towards Human-like Trajectory Prediction for Autonomous Vehicles: A Behavior-aware Approach

Accurately predicting the trajectories of surrounding vehicles is essential for safe and efficient autonomous driving. This paper introduces a novel behavior-aware trajectory prediction model (BAT) that incorporates insights from traffic psychology, human behavior, and decision-making. BAT seamlessly integrates four modules: behavior-aware, interaction-aware, priority-aware, and position-aware. These modules perceive and understand underlying interactions and account for uncertainty and variability in prediction, enabling higher-level learning and flexibility without rigid categorization of driving behavior. Importantly, BAT eliminates the need for manual labeling in the training process and addresses the challenges of non-continuous behavior labeling and the selection of appropriate time windows. Furthermore, we evaluate BAT’s performance across the Next Generation Simulation (NGSIM), Highway Drone (HighD), Roundabout Drone (RounD), and Macao Connected Autonomous Driving (MoCAD) datasets, showcasing its superiority over prevailing state-of-the-art (SOTA) benchmarks in terms of prediction accuracy and efficiency. Remarkably, even when trained on reduced portions of the training data (25%) and with a much smaller number of parameters, BAT outperforms most of the baselines, demonstrating its robustness and efficiency in challenging traffic scenarios, including highways, roundabouts, campuses, and busy urban locales. This underlines its potential to reduce the amount of data required to train autonomous vehicles, especially in corner cases. In conclusion, the behavior-aware model represents a significant advancement in the development of autonomous vehicles capable of predicting trajectories with the same level of proficiency as human drivers.
Wencheng HAN

Learning 3D Geometry from Visual Consistency for Monocular Vision Tasks in Self-Driving Cars

Monocular vision tasks, like depth estimation and 3D object detection in self-driving cars, face challenges due to the limited 3D information in 2D images. This highlights our recent achievements that leverage visual consistency to enhance monocular vision models. We first introduce the Direction-aware Cumulative Convolution Network (DaCCN) for self-supervised depth estimation. DaCCN addresses direction sensitivity and environmental dependency, improving feature representation and achieving state-of-the-art performance. Next, we present the Rich-resource Prior Depth estimator (RPrDepth), which uses rich-resource data as prior information. RPrDepth estimates depth from a single low-resolution image by referencing pre-extracted features. Finally, we discuss a weakly-supervised approach for monocular 3D object detection that relies only on 2D labels. By using spatial and temporal view consistency, this method achieves results comparable to fully supervised models and enhances performance with minimal labeled data.
Chunlin TIAN

An Asymmetric LoRA Architecture for Efficient Fine-Tuning

Adapting Large Language Models (LLMs) to new tasks through fine-tuning has been made more efficient by the introduction of Parameter-Efficient Fine-Tuning (PEFT) techniques, such as LoRA. However, these methods often underperform compared to full fine-tuning, particularly in scenarios involving complex datasets. This issue becomes even more pronounced in complex domains, highlighting the need for improved PEFT approaches that can achieve better performance. Through a series of experiments, we have uncovered two critical insights that shed light on the training and parameter inefficiency of LoRA. Building on these insights, we have developed HydraLoRA, a LoRA framework with an asymmetric structure that eliminates the need for domain expertise. Our experiments demonstrate that HydraLoRA outperforms other PEFT approaches, even those that rely on domain knowledge during the training and inference phases.
Tianxiao GAO

Night-Voyager: Consistent and Efficient Nocturnal Vision-Aided State Estimation in Cross-Modal Maps

Accurate and robust state estimation at nighttime is essential for autonomous navigation of mobile robots to achieve nocturnal or round-the-clock tasks. An intuitive and practical question arises: Can low-cost standard cameras be exploited for nocturnal state estimation? Regrettably, most current visual methods tend to fail due to adverse illumination conditions, even with active lighting or image enhancement. A crucial insight, however, is that the streetlights in most urban scenes can provide static and salient visual information at night. This inspires us to design an object-level nocturnal vision-aided state estimation framework, named Night-Voyager, which leverages cross-modal maps and keypoints to enable the versatile all-day localization. Night-Voyager starts with a fast initialization module which solves the global localization problem. With the effective two-stage cross-modal data association approach, the system state can be accurately updated by the map-based observations. Meanwhile, to address the challenge of large uncertainties in visual observations during nighttime, a novel matrix Lie group formulation and a feature-decoupled multi-state invariant filter are proposed for consistent and efficient state estimation. Comprehensive experiments on both simulation and diverse real-world scenarios (about 12.3 km total distance) showcase the effectiveness, robustness, and efficiency of Night-Voyager, filling a critical gap in nocturnal vision-aided state estimation.

 

For enquiries: Tel: 8822 9159
Email: frankielei@um.edu.mo

Best Regards,
State Key Laboratory of Internet of Things for Smart City

IOTSC Postgraduate Forum: Intelligent Transportation2024-11-07T00:00:44+08:00
23 2024-09

IOTSC TALK SERIES: Fair Allocation of Chores with Subsidy

2024-10-23T00:01:28+08:00

Dear Colleagues and Students,

The State Key Laboratory of Internet of Things for Smart City would like to invite you to join our “IOTSC Talk Series” on 25/09/2024 (Wednesday). We are pleased to invite Prof. Xiaowei WU from State Key Laboratory of Internet of Things for Smart City as the speaker.

Fair Allocation of Chores with Subsidy
Speaker: Xiaowei WU
Date: 25/09/2024 (Wednesday)
Time: 15:00 – 15:55
Language: English
Venue: N21-5007 (Exhibition Hall)

Abstract:
The fair allocation problem has gained significant attention recently in the fields of theoretical computer science, artificial intelligence, and economics. In this presentation, I will discuss our latest research on ensuring fairness for the allocation of chores using subsidies. We consider the allocation of m indivisible chores among n agents with subsidies. Specifically, we focus on scenarios where agents have additive cost functions and assume that the maximum cost of an item to an agent can be offset by one dollar, we show that a total subsidy of n/4 dollars is sufficient to achieve a proportional allocation. Furthermore, we prove that n/4 is the minimum necessary subsidy, as there exists an instance with n agents where any proportional allocation requires at least n/4 dollars in subsidies. Additionally, we explore the weighted case and show that a total subsidy of ( n/3 ) dollars is sufficient to ensure weighted proportionality.

Speaker’s Bio:
Xiaowei Wu is an Assistant Professor in the Department of Computer and Information Science with the State Key Laboratory of Internet of Things for Smart City at the University of Macau. He received his Ph.D. degree from the University of Hong Kong and his B.Eng. degree from University of Science and Technology of China. His research interests span various topics in online approximation algorithms, algorithmic game theory, and computational social choice. He has published more than 50 papers in top theory and artificial intelligence conferences and journals including JACM, SICOMP, AIJ, STOC, FOCS, SODA, EC, WINE, AAAI and IJCAI. He is an Associate Editor for the Journal of Combinatorial Optimization. He has served as the PC chair and local organizing chairs of several international conferences and competitions, including IJTCS-FAW 2023, ICPC 2020 – 2023, MCSCT 2022 – 2024 and GPC 2024.

For enquiries: Tel: 8822 9159
Email: frankielei@um.edu.mo

Best Regards,
State Key Laboratory of Internet of Things for Smart City

IOTSC TALK SERIES: Fair Allocation of Chores with Subsidy2024-10-23T00:01:28+08:00
19 2024-09

IOTSC TALK SERIES: Fair Allocation of Chores with Subsidy

2024-10-19T00:00:25+08:00

Dear Colleagues and Students,

The State Key Laboratory of Internet of Things for Smart City would like to invite you to join our “IOTSC Talk Series” on 25/09/2024 (Wednesday). We are pleased to invite Prof. Xiaowei WU from State Key Laboratory of Internet of Things for Smart City as the speaker.

Fair Allocation of Chores with Subsidy
Speaker: Xiaowei WU
Date: 25/09/2024 (Wednesday)
Time: 15:00 – 15:55
Language: English
Venue: N21-5007 (Exhibition Hall)

Abstract:
The fair allocation problem has gained significant attention recently in the fields of theoretical computer science, artificial intelligence, and economics. In this presentation, I will discuss our latest research on ensuring fairness for the allocation of chores using subsidies. We consider the allocation of m indivisible chores among n agents with subsidies. Specifically, we focus on scenarios where agents have additive cost functions and assume that the maximum cost of an item to an agent can be offset by one dollar, we show that a total subsidy of n/4 dollars is sufficient to achieve a proportional allocation. Furthermore, we prove that n/4 is the minimum necessary subsidy, as there exists an instance with n agents where any proportional allocation requires at least n/4 dollars in subsidies. Additionally, we explore the weighted case and show that a total subsidy of ( n/3 ) dollars is sufficient to ensure weighted proportionality.

Speaker’s Bio:
Xiaowei Wu is an Assistant Professor in the Department of Computer and Information Science with the State Key Laboratory of Internet of Things for Smart City at the University of Macau. He received his Ph.D. degree from the University of Hong Kong and his B.Eng. degree from University of Science and Technology of China. His research interests span various topics in online approximation algorithms, algorithmic game theory, and computational social choice. He has published more than 50 papers in top theory and artificial intelligence conferences and journals including JACM, SICOMP, AIJ, STOC, FOCS, SODA, EC, WINE, AAAI and IJCAI. He is an Associate Editor for the Journal of Combinatorial Optimization. He has served as the PC chair and local organizing chairs of several international conferences and competitions, including IJTCS-FAW 2023, ICPC 2020 – 2023, MCSCT 2022 – 2024 and GPC 2024.

For enquiries: Tel: 8822 9159
Email: frankielei@um.edu.mo

Best Regards,
State Key Laboratory of Internet of Things for Smart City

IOTSC TALK SERIES: Fair Allocation of Chores with Subsidy2024-10-19T00:00:25+08:00
© University of Macau
Go to Top