Intelligent transportation makes our travel more convenient and efficient. With the development of mobile Internet and position technologies, it is reasonable to collect spatio-temporal data and then leverage these data to achieve the goal of intelligent transportation. However, effectively understanding and modeling the complex spatiotemporal correlations in traffic data remains a challenge. In this talk, we present a comprehensive literature review on spatiotemporal modeling in intelligent transportation, focusing on the latest techniques and approaches. Finally, we will discusses the current challenges and limitations in spatiotemporal modeling and proposes potential future directions.

Seemingly free to move around in urban space, people from "mainstream" neighborhoods and disadvantaged neighborhoods still demonstrate considerably different visitation patterns, due to both personal choices and accessibility to important urban facilities. Such phenomenon is known as experienced segregation, the measurement of which is a ubiquitous and dynamic indicator of equity and inclusiveness from the microscopic structure of urban life. Meanwhile, the city as a whole possesses a macroscopic structure, i.e., the spatial distribution of various public and private spaces, which shapes the possible microscopic segregation but is also influenced in turn. A number of works have focused on either the microsocopic segregation or the macroscopic urban morphology, there is still a gap in connecting and even uniting these two levels of investigation. In this talk, I will review several important papers published in the past few years and discuss potential research directions.

Learning the underlying slow and fast dynamics of a system is instrumental for many practical applications related to the system. However, existing approaches are limited in discovering the appropriate time scale to separate the slow and fast variables and effectively learning their dynamics based on correct-dimensional representation vectors. In this work, we introduce a framework that effectively learns slow and fast system dynamics in an integrated manner. We propose a novel intrinsic dimensionality (ID) driven learning method based on a time-lagged autoencoder framework to identify appropriate time scales to separate slow and fast variables and their IDs simultaneously. Further, we propose an integrated framework to concurrently learn the system’s slow and fast dynamics, which is able to integrate prior knowledge of time scale and IDs and model the complex coupled slow and fast variables. Extensive experimental results on representative dynamical systems show that our proposed framework is able to efficiently learn slow and fast system dynamics.

The mobile cellular network connects the world, i.e., users access base stations and communicate with the world via mobile traffic, e.g., the world consumed 115 EB traffic in Q4 of 2022. Thus, accurately predicting mobile traffic and accessed user amount is of great importance to network resource allocation, energy saving, etc.However, due to the complicated environmental contexts and complex interaction between mobile traffic and connected users, mobile network prediction is still challenging. Besides, the existing works could not be applied to large-scale networks because of the limited hardware resources and unacceptable time cost. In this work, we propose the Spatial-Temporal Transformer Framework (STTF) to solve the problem of the joint prediction of mobile traffic and the number of connected users. Our proposed model contains three key parts. First, to capture the complex interaction between mobile traffic and connected users, we propose the temporal cross-attention encoder. Then, to identify and extract the most relevant information from various semantic relationships, we propose the hierarchical spatial encoder. This information is then used to create a more comprehensive representation of the network. Finally, the subgraph sampling method could significantly reduce the amount of computing power required and have comparable performance to the methods that input the whole networks enabling the model for real-world applications. Extensive experiments demonstrate that our proposed model significantly outperforms the state-of-the-art models by over 17\% in both mobile traffic prediction and connected user prediction.

Offline RL is a paradigm that learns exclusively from static datasets of previously collected interactions, making it feasible to extract policies from large and diverse training datasets. Effective offline RL algorithms have a much wider range of applications than online RL, being particularly appealing for real-world applications such as education, healthcare, and robotics. This talk will present research works related to offline RL.

With the increasing popularity of mobile applications, understanding user behavior has become crucial for app developers. The diffusion model offers a novel approach to app behavior generation by examining how users adopt and spread behaviors within a social network. This report discusses the utilization of the diffusion model for app behavior generation. The diffusion model, rooted in social psychology, provides valuable insights into the spread of ideas and behaviors among individuals. By adapting this model to app development, we can simulate and predict user behaviors, leading to enhanced app design, personalized recommendations, and targeted marketing strategies. The utilization of the diffusion model in app behavior generation offers several benefits. Firstly, it enables app designers to create more engaging and user-centric interfaces by understanding how behaviors spread and evolve. Secondly, personalized recommendations based on the diffusion model can enhance user satisfaction and increase app usage. Lastly, targeted marketing strategies can be developed by identifying influential users who can drive behavior adoption within the app's user base. By conducting research on several models based on diffusion for time series forecasting, we aim to achieve the desired outcome by making modifications to them. By conducting experiments using the existing dataset, we have organized the existing experimental results and conclusions, and planned for future work.

Urban villages, defined as informal residential areas in or around urban centers, are characterized by inadequate infrastructure and poor living conditions, closely related to the Sustainable Development Goals (SDGs) on poverty, adequate housing, and sustainable cities. However, traditional field survey methods for detecting urban villages are time-consuming, labor-intensive, and possibly delayed. In recent research, widely available satellite imagery has become an important tool for detecting urban villages. However, existing methods either focus on simple image classification without accurate boundary or lack precision in boundary identification. To overcome these challenges and harness the power of the Segment Anything Model (SAM), we propose a large-model-based framework for urban village boundary identification. In order to enhance the quality of bounding box prompts, the framework utilizes the distinction between urban villages and their surrounding environments to incorporate contrastive learning into the object detection model. Furthermore, we hope to analyze the spatiotemporal changes in multiple cities to gain deeper insights into the development trends of urban villages.

Deep reinforcement learning has shown its advantages in solving multiple combinatorial optimization problems in recent years. However, challenges still exist, especially in solving those with complicated constraints. The application of RL to constrained problems raises concerns about violating constraints during the learning process. In this paper, we propose a novel approach, called Safe Reinforcement Learning for Constrained Combinatorial Optimization (SRL-CCO), which addresses the challenge of optimizing constrained combinatorial problems using RL while guaranteeing constraint satisfaction.First, we train an RL by considering both constraints and objectives via a trust region iteration approach. Second, we expand the parameter space after training to continue solution search to find better solution incumbent. Results demonstrate that both approaches bring siginificant performance improvement.

Digital Twin brings a new realization approach to the modeling of mobile networks. Mobile networks, as complex systems comprising multiple components, such as mobile users, base stations, and wireless environments, have intricate interactions and relationships with each other. By creating a virtual replica of each physical mobile network entity in a virtual space, we build a scalable digital twin system for mobile communication networks. Due to the privacy issues in data sharing and the high expenses in data acquisition, our system integrates generative algorithms to generate mobile trajectories and communication traffic demand. Moreover, the system can interact with multiple optimizers to evaluate and display real-time simulation results.

For operation-oriented autonomous driving systems, simulators are usually indispensable. On the one hand, the iteration of the system needs to be tested on the simulator to compare performance, and to test safety performance before deployment to real vehicles. On the other hand, various modules based on learning can use simulators to generate more data to help with training. In recent years, more and more companies and research institutions have begun to focus on the problems and improvement solutions in simulators. Companies like Waymo and Tesla have talked about their achievements in simulation in recent years. With the advancement of autonomous driving algorithms, the industry has put forward higher and higher requirements for the authenticity, reliability, and scalability of simulators. In autonomous driving systems, modeling and predicting human intent is especially difficult when using simulators for testing trajectory prediction and trajectory planning modules. Compared with the open-loop testing still widely used in the industry, closed-loop simulation can more realistically and effectively test the performance of real vehicles on the road. This report will analyze and compare the relevant technologies of closed-loop simulation of driving behavior and introduce some of the main research ideas.

Most model-free reinforcement learning methods leverage state representations (embeddings) for generalization, but either ignore structure in the space of actions or assume the structure is provided a priori. Existing research decomposes a policy into a component that acts in a low-dimensional space of action representations and a component that transforms these representations into actual actions via auto-encoder. These representations improve generalization over large, finite action sets by allowing the agent to infer the outcomes of actions similar to actions already taken. However, existing action standards cannot distinguish good actions from bad ones. Therefore, we propose a self-supervised learning based action space representation method to solve this problem.

Trajectory generation involves using algorithms to create a sequence of ordered data points that follow existing rules and data. This can generate spatio-temporal trajectory data for AI applications like traffic optimization and epidemic simulation. Having control over user profile attributes can enhance its value, but there's currently limited research on this. In this talk, we will introduce a controllable trajectory generation framework based on a diffusion probability model and classifier guidance to capture human movement patterns and trajectory data characteristics.

In major cities such as Beijing or Shanghai, taxis and subways are the most commonly utilized modes of transportation. In the taxi-hailing scenario, the online ride-hailing platform needs to predict the number of passengers in advance from one region to another to obtain the flow pattern of passengers. We have formulated this problem as a source-destination flow prediction problem. In the subway scenario, the prediction of source-destination flow can help passengers make reasonable travel plans, and enhance passenger travel experience. In this presentation, I will provide an overview of the current solutions to the source-destination flow prediction problem.

In the last decades, cities are widely accepted to be disequilibrium and organized from bottom to up, regarding the importance of evolving systems with emergent phenomena built from local decisions. Yet, we are still far from understanding the driving mechanisms of the emergence of positive social phenomena such as the formation of innovation hubs and the increase in cultural activities. In particular, approaches to analyzing urban phenomena are limited in scope by neglecting both temporal non-stationarity and spatial heterogeneity. Here we develop an approach to describe the spatial, temporal, and periodic variations in urban rhythm. With data from New York, we characterize how the periodicity of activities varies spatially across the city over time. We confirm one-year activity cycles and show that this periodicity occurs unevenly across the city. These ‘waves of activity’ keep traveling across the city: while cities have a stable number of regions with a circannual period, the regions exhibit non-stationary series.

The development of autonomous driving technology is gradually changing the real players in road traffic. Many autonomous driving strategies may have an impact on the energy efficiency and access efficiency of traffic, especially V2X-related strategies. In this study, we try to use simulation to analyze the impact of three autonomous driving strategies on traffic efficiency in the future when autonomous driving and connected vehicle technologies are fully rolled out. These three strategies include (1) platooning, (2) intersection rhythm control, and (3) dynamic traffic assignment. These three strategies correspond to the three spatial levels of cooperative autonomous driving, adjacent space, intersection, and city. In this group meeting, the overall research idea and preliminary experimental results will be reported.

In the era of big data, fully unleashing the value of data elements and leveraging the key supporting role of data in the development of the digital economy is an inevitable law of the development of the digital society. Currently, there is a strong demand for data applications, and data is gradually penetrating various fields. Various data applications are developing rapidly, but the effective supply of data is insufficient. Support for urban governance, economic development, and livelihood services needs to be further strengthened. Therefore, it is necessary to form a complete market-oriented trading platform within the range of tradable data elements. In this paper, we take the use authorization of data elements as an example to first propose the concept of distributed resource authorization, and explain how to use blockchain to build a complete framework for distributed resource authorization, so as to achieve efficient transmission and reliable circulation of data elements. At the same time, it can be expanded to many social service fields such as real estate. The research content includes the following four points: the concept and flow requirements of data components, the overall design and specific implementation of networks using CDN and blockchain.

A large amount of mobility data is often needed to support downstream applications. However, the quality and quantity of real trajectories are limited due to user privacy concerns or constraints in data acquisition.Therefore, trajectory generation becomes a great solution. Many existing models for generating individual trajectories do not consider population distribution, which is actually crucial in guiding trajectory generation. Therefore, in this work, we propose a trajectory generation framework based on a diffusion model that enables the generation of individual mobility trajectories while considering the dynamic population distribution.

Apps like TikTok and Kuaishou have leveraged recommender systems to achieve unprecendented success in terms of user growth and business profit, but their explosive popularity has some downsides. With ubiquitous data and smartphones at everyones disposal, the issue of smartphone addiction has become a hot topic, leading both researchers and industry professionals to seek out solutions. In our ongoing work, we investigate the factors that can lead to addiction of Kuaishou based on analysis of user behavior across a one year period. Our results should bring light to the question of whether addiction is primarily driven by user factors of platform factors, and give us a direction for the design of future, more responsible recommender systems that prevent addiction. Preliminary results show that demographic factors such as age and gender could be risk factors for addiction.

In logistics service, the delivery timely rate is a key experience indicator which is highly essential to the competitive advantage of express companies. Prediction on it enables intervention on couriers with low predicted results in advance, thus ensuring employee productivity and customer satisfaction. Currently, few related works focus on couriers’ level delivery timely rate prediction, and there are complex spatial correlations between couriers and road districts in the express scenario, which make traditional real-time prediction approaches hard to utilize. What's more, when facing unexpected events such as pandemics or natural disasters, the logistics indicators may experience significant fluctuations and are difficult to predict due to regional lockdowns and transportation disruptions. In this talk, we will study couriers delivery timely rate prediction problem under anomaly condition based on spatiotemporal analysis.

Traffic forecasting is a very important research direction, which can help us find the potential congestion risks and trends in time, so as to take corresponding measures to reduce or avoid the occurrence of traffic congestion. Recently, the most popular models are Spatial-Temporal Graph Neural Networks (STGNNs). STGNNs mainly use the graph convolutional network to capture the correlation of traffic speeds at different locations on the road network and use sequence models such as RNN, LSTM and TCN to capture the autocorrelation of traffic speed. Due to the limitation of computing capacity, STGNNs only model short-term historical data, such as data over the past one hour, resulting in a large amount of accumulated long-term historical information (e.g., the past two weeks) is not well utilized. Therefore, this paper proposes a method named Contrastive Learning Enhanced Spatial-temporal Graph Neural Network (CLE-STGNN) to utilize long-term historical information. Specifically, this paper first maps the very long-term historical sequences to a low-dimensional hidden vector via contrastive learning. While, the model extracts the periodic features of long-term traffic speed in the frequency domain through Fourier transform. Then, the low-dimensional hidden vector can be input into STGNNs. Since the diffusion intensity of traffic flow changes with time, this paper adopts the method of learning dynamic graph to model it. Extensive experimental results show that our model exceeds the baseline model in traffic prediction performance.

To address the problem of road congestion and provide personalized travel experiences for specific passengers, customized bus services have been introduced in several cities. This approach involves designing optimal routes for customized buses, which poses a challenging combinatorial optimization problem. Existing research has relied on heuristic algorithms and precise formula methods to tackle the problem. However, these methods are task-specific, and they fail to fully leverage available data. Moreover, they are time-consuming and cannot meet the demanding response speed requirements of customized bus route design. To overcome these drawbacks, this paper proposes a novel customized bus route design method based on deep reinforcement learning. The proposed approach involves initially using the K-means clustering algorithm based on spherical distance to group passenger addresses and then computing the customized bus stops based on the generated clusters. The Tencent map interface is utilized to identify the actual road network distance matrix and driving time matrix between each stop and destination. Additionally, the optimization goal incorporates the total travel time of customized buses to avoid the use of congested sections on the route. Finally, an improved attention mechanism algorithm is adopted to effectively embed and represent the real road network distance matrix and travel time matrix between each site and destination. A decoder comprising an attention mechanism is utilized to gradually obtain optimal customized bus routes based on the embedded representation. The effectiveness of the proposed method is validated through experiments on real datasets.

The novel coronavirus pandemic hits the world in late 2019, and people around the world have been more or less affected by the pandemic in countless ways. It also happens in the academic world. This pandemic is one of the few incidents in history that can cause impact to almost all areas of research. Ever since the outbreak of Covid-19, research on this pandemic has been one of the most popular topics in numerous disciplines. Undoubtedly, we can feel the influence it brings to the academic world. Covid has gathered researchers of different disciplines, working together to produce novel ideas. No doubt covid has influenced the academic world, but to our best knowledge, questions of A. How impactful it is? B. In what aspect does it affect? C. What features do those impacts have, compared to other topics? remain widely unexplored. In this special era, it would be very critical to answer: how did the pandemic affect interdisciplinary research, so that researchers of different disciplines can better understand Covid-19-related research? This ongoing work aims to answer the above questions by examining the features of Covid-19, as an academic topic, in 2 aspects: A. Interdisciplinary Collaborations B. Impact Factors of Interdisciplinary Publications. Then concludes its impact done to the development of interdisciplinary research by comparing it with its counterparts of identical nature in both the short and long term.

Generalization in reinforcement learning over unseen environment variations is crucial for real-world scenarios. From our observation, an agent trained on a wide range of environment variations (generalist) have better generalization, but they may converge to inferior solutions. On the other hand, an agent trained only on a limited set of environment variations (specialist) can converge to better solutions but have poor generalization. To have the best of both worlds, we propose a novel framework called generalist and hyper-specialist (GHS). GHS consists of two parts: a generalist trained on all environment variations and a hyper-specialist conditioned on a few environment variations. Specifically, we first encoder the context of environment variation. Then we use a hyper-network to generate the policy network (hyper-specialist) based on the environment context. Finally, we add the generalist network and hyper-specialist network together, thus leveraging the advantages of both. We conduct experiments on ride-sharing real-world scenario and Procgen benchmark. Our results demonstrate that our approach effectively improves generalization.

Electric Vehicles (EV) have become increasingly popular as a sustainable mode of transportation, making crucial contributions to emission reduction and climate goals. California has set a mandated EV transition goal that all new cars sold in the state be electric by 2035, meaning 12 million more EVs on California’s roads. Despite the fast commercialization of EV, the newness of electric vehicles, their high upfront cost, and need for charging access meant equity has been overlooked. Previous works on EV charging infrastructure inequity mainly focus on the existence of charging facilities on neighborhood or ZIP-code level, neglecting their spatial concentration, the affordability to use them, and the economic and environmental outcomes of EV charging infrastructure inequity. In this talk, we propose to probe the inequity in EV charging infrastructure from above perspectives and list some preliminary analytical results in California.

Resilience, the ability to maintain fundamental functionality when failures and errors occur, is a vital property for many complex networked systems. Existing works mainly seek to provide theoretical estimation, which requires prior knowledge of network dynamics and relies on the assumption of low degree correlations. However, real-world networked systems often have diverse topology patterns, and their dynamics are usually unknown, hindering the application of the theoretical tools and making the inference of their resilience an open challenge. Here, we propose a powerful deep learning framework ResInf for resilience inference based on the topology and node activity series of a broad range of networked systems. Our numerical results demonstrate that ResInf can accurately predict the resilience of real-world systems via the empirical interaction network and activity series, shedding light on a research paradigm that introduces deep learning methods to infer complex system properties.

Precisely estimating high-resolution carbon emissions is of significant importance to planning emission mitigation efforts. Existing carbon emission accounting methods generally fall into two categories: bottom-up ones and top-down ones. Specifically, the former type calculates and aggregates carbon emissions from point sources. In contrast, the latter type allocates city-level total carbon emission, which is usually calculated by regional fuel consumption statistics, to high-resolution grids by proxy factors like population, road length, and nighttime light. However, the bottom-up methods highly rely on massive data collection, while the top-down ones are limited in precision and resolution. In today's talk, I will introduce the existing commonly used carbon emission accounting methods and the current pain points in the field of carbon emission estimation. I will also introduce my work on predicting carbon emission with multi-source urban data, which manages to relieve the data collection expense of existing accounting methods.

The structure of neural network is a key factor to the network performance, while it’s still mystery how network structure influences the performance. The frustration as a source of disorder provides flexibility for network organization in various complex networks such as brain network. Through the lens of complex network, we try to design corresponding measurement and identify the pattern of frustration in the neural network. Extensive experiments show there’s significant relation between the frustration pattern and gradient stability, which could help explain the performance distinction between networks with different structures. Our results may provide a novel view of neural network behavior from the perspective of complex network.

Is it possible to predict the emergence of high-impact papers? Prior research has developed models to identify patterns of hot streaks among individual researchers during which they produce a series of high-impact papers in a concentrated period of time. Additionally, there are models that estimate the awakening probability of "sleeping beauties," which are initially overlooked papers that gain attention after a significant time since publication. However, these studies generally overlook the dynamic interaction between collective research trends and individual research interests, which can play a significant role in determining a paper's popularity. In this work, we aim to deconstruct the emergence of high-impact papers with a complex network approach. First, we highlight the limited attention within the research community, emphasizing the importance of understanding the top trending keywords. Next, we compare the evolution speed of collective attention and individual attention, demonstrating that the former changes significantly faster than the latter. In the next step, we will dive deep into the keyword embedding space to show how certain keyword combinations are more likely to go popular. Our work may provide insights into identifying critical research directions in a timely and dynamic manner.

Crowd flow represents the number of people entering or leaving an urban region within a given time interval. Estimating the crowd flow of urban regions plays an important role in public safety and transportation system. Existing works mostly predict crowd flow in the future based on historical flow, which is not always available. Some other works generate OD flow among regions while ignoring its temporal variance. In this work, we aim to generate crowd flow for regions in different time periods without information about real flow. We first construct a knowledge graph to model regions and other elements in the city like POIs as well as complex relationships between them. Then we leverage diffusion model as the backbone to generate crowd flow for regions in the KG. Specifically, in the denoising module, we propose a knowledge enhanced ST-block to capture both the semantic dependencies among regions and temporal flow patterns. Extensive experiments on four real world datasets demonstrates our model’s superiority over baselines. Further experiments show the effectiveness of our model design and its benefit for downstream tasks.

In this presentation, we discuss the diffusion probabilistic modeling (DPM) framework for spatiotemporal data, which is a fairly new topic that is barely explored compared with its counterparts in computer vision and natural language processing. We will talk about the unique challenges, but also opportunities, brought by intrinsic characteristics of spatiotemporal data. After introducing a wide range of elementary tasks in DPM (class-conditioned sampling, imputation, solving inverse problems, etc.), we will discuss the possibility of combining prediction, interpolation, generation, and anomaly detection of spatiotemporal data in a unified manner.

The heatedly discussed issues of open access (OA) has acquired increasing attention since U.S. government’s announcement of OA requirements on all research papers it funds. However, it remains largely unclear how OA relates to the discrepancies across identities and whether it alleviates or exacerbates the prevailing inequalities in science. Drawing upon 39 million papers published between 1990 and 2020 across all scientific fields, we show that as OA and female-majority publishing teams grow, female authors more actively engage in OA, but are paradoxically even more disadvantaged in visibility in terms of OA publications. Our findings suggests the necessity of transformation in science to make the future scientific workforce more equitable and inclusive.

In this talk, I will introduce a paper about assessing the livelihood impact of electricity access by machine learning (ML) and then briefly present my ongoing work. Insufficient data on significant economic outcomes in several parts of the world have impeded the progress, targeting, and assessment of public policy. To tackle these data and inference challenges, a nature paper proposes leveraging satellite imagery and machine learning advances. The study focuses on the expansion of the electrical grid in Uganda, where the researchers demonstrate how combining satellite imagery and computer vision can lead to the development of localized livelihood measurements, appropriate for inferring the causal impact of electricity access on livelihoods. Following this paper, we intend to measure the impact of road construction on economic development for poverty alleviation by applying both CV and satellite images.

Introducing social cognition modeling into urban simulations can enhance the realism and complexity of urban dynamics by capturing the social interactions and feedback effects that shape urban systems. In this talk, we will present a design framework for integrating social cognition modeling into our urban simulation systems. The framework includes a system implementation of a cognitive simulator, techniques for integrating cognitive and urban simulators, and methods for modeling interactions between physical and social domains. By introducing social cognition modeling into urban simulations, we can gain a deeper understanding of the social dynamics that shape our urban environment, which can inform more effective urban planning and policy-making.

Cities built environment greatly shapes citizens’ health outcomes. Previous studies point out that the walkability of a city correlates with the dementia incidence, and the severity of COVID-19 transmission is also influenced by the urban resilience. With the recent advances of deep learning and big data processing, geo-tagged imagery provides us a new perspective to assess the built environment of cities and how it affects public health. In this talk, we will introduce the definition of transmission risks, through which we unveil the built environment features by geo-tagged imagery and propose a model accurately assess the COVID-19 incidence in urban spaces.

Fragmentation, a system's process of breaking itself into completely isolated or weakly connected small parts, is a ubiquitous phenomenon in many complex systems. Despite widespread consequences for urban, information, and ecological systems, the mechanisms governing the emergence of fragmentation are still unknown. The limitation is rooted in a theoretical gap: the current framework only focuses on homogeneous networks, which is unsuitable for systems consisting of heterogenous components that interact in a bipartite relationship. Here we bridge this theoretical gap by developing a general framework to uncover the mechanisms governing the universal fragmentation patterns in co-evolving bipartite systems.

The concept of resilience can be realized in natural and engi- neering systems, representing the ability of a system to adapt and recover from various disturbances. Although resilience is a critical property needed for understanding and managing the risks and collapses of transportation systems, an accepted and useful definition of resilience for urban traffic as well as its statistical property under perturbations are still missing. Here, we define city traffic resilience based on the spatiotemporal clusters of congestion in real traffic and find that the resilience follows a scale-free distribution in 2D city road networks and 1D highways with different exponents but similar exponents on different days and in different cities. The traffic resilience is also revealed to have a scaling relation between the cluster size of the spatiotemporal jam and its recovery duration independent of microscopic details. Our findings of universal traffic resilience can provide an indication toward better understanding and designing of these complex engineering systems under internal and external disturbances.

The movement of individuals is influenced by various factors, which can impact the collective behavior of crowds, making it difficult to predict. Knowledge-driven approaches, such as crowd dynamics, provide fundamental rules and models with good interpretability and generalizability across different scenarios, but they cannot capture the complexity and randomness of data. Deep learning methods have shown good performance in data fitting, but their generalization ability may suffer from extrapolation or observation bias, leading to physically inconsistent or unreliable predictions. Our research aims to combine crowd dynamics and deep learning for micro-level pedestrian simulation. Firstly, we generate crowd trajectories based on a diffusion probability model, where the model takes the current movement data as input and denoises the acceleration distribution of the next timestep from Gaussian noise. Secondly, we embed social force knowledge as conditional information into the probabilistic model and decompose the acceleration into repulsive and attractive forces, allowing the model to focus on learning the dynamic interaction between pedestrians and the environment. Finally, we propose a new diffusion model training mechanism for long-term simulation. Further more, I will briefly introduce our future exploration of incorporating geometric equivariance in the current task.

How to use reinforcement learning technology to optimize wireless network coverage is a topic that we are very interested in. In this seminar, two works on wireless network coverage optimization using multi-agent reinforcement learning will be shared. Remote Electrical Tilt (RET) optimization is an efficient method for adjusting the vertical tilt angle of Base Stations (BSs) antennas in order to optimize Key Performance Indicators (KPIs) of the network. The first work [1] demonstrates how to use coordination graphs and reinforcement learning for RET optimization in a complex application involving hundreds of cooperating agents. The work shows how mobile networks can be modeled using coordination graphs and how network optimization problems can be solved efficiently using multiagent reinforcement learning.The second work [2] focuses Unmanned Aerial Vehicles-BaseStations (UAV-BSs). UAV-BSs can provide communication services for ground devices (GDs) in various scenarios, such as emergency communication and traffic offloading in hotspots. The work characterizes the fairness at user-level based on proportional fairness scheduling and formulate a weighted-throughput maximization problem via designing UAV-BSs’ trajectory. The approach adopts the framework of centralized training with distributed execution. Simulation results show that the method can improve fairness of communication service by sacrificing a small amount of throughput. [1] Bouton M, Farooq H, Forgeat J, et al. Coordinated reinforcement learning for optimizing mobile networks[J]. arXiv preprint arXiv:2109.15175, 2021. [2] Z. Qin, Z. Liu, G. Han, C. Lin, L. Guo, and L. Xie, "Distributed UAV-BSs Trajectory Optimization for User-Level Fair Communication Service With Multi-Agent Deep Reinforcement Learning," IEEE Transactions on Vehicular Technology, vol. 70, no. 12, pp. 12290-12299, Dec. 2021

The widespread spread of COVID-19 and H1N1 has had a huge impact on public health. To accurately predict the spread of infectious diseases using machine learning models, we proposed a data-driven macro-micro collaborative model to predict disease dynamics. We first simulate the dynamics of disease transmission based on a human mobility dataset. Secondly, the macro graph and micro graph were generated. Then, the prediction of macro and micro graph collaboration is realized, and the epidemic dynamics in the future prediction window can be realized on the granularity of the epidemic dynamic prediction region according to the limited history window. Our innovation is not limited to separate macro or micro data, but to use a macro-micro collaborative approach to capture the changes of the micro graph under different dynamics of the disease by considering the dynamics of the user granularity along with the regional granularity prediction.

Base station energy saving aims to reduce total energy consumption by optimizing and adjusting the operation of base station equipments. This paper considers how to adjust the control statuses of base stations and cells to meet predicted traffic demands and simultaneously minimize energy consumption. First, this paper analysis the relationship between the energy consumption of different devices in the base station and their traffic load, and then set up a compatible energy consumption model. Then, this paper decomposes the optimization problem and proposes a reinforcement learning method to adjust the cell control statuses, and then adapt the base station control statuses and traffic load conditions to near-optimal based on the cell control statuses.Then, this paper decomposes the optimization problem and proposes a reinforcement learning method to adjust the cell control statuses, and then based on them adapt the base station control statuses and traffic load conditions to near-optimal. Finally, the algorithm is evaluated under realistic base station distribution and traffic demand distribution. The experimental results show that the algorithm can save about 55% of energy consumption compared to the original control scheme and has a significant improvement compared to other baselines.

Meituan is the Groupon e-commerce site in China. It helps consumers find the most trustworthy businesses and low discount products. We introduce two recommendation tasks in Meituan: 1) knowledge graph based recommendation and 2) recommendation on biased data. The knowledge graph is a natural and general representation of data in modern large commercial recommender systems which involve heterogeneous types of data. Knowledge graph based recommenders faces two problems: how to represent the semantic relations of knowledge graph and how to fuse the knowledge graph information to make recommendations. Recommendation algorithms typically build models based on user-item interactions (e.g., clicks, likes, or ratings) to provide a personalized ranked list of items. These interactions are often distributed unevenly over different groups of items due to varying user preferences. The main challenge is how to adapt the recommendation model to handle a new biased data.

Inequality is a long-standing issue existing across various scenarios, such as gender inequality of scientific impact, infrastructure inequality in cities. However, inequality of people's online consumption has not recieved much attention in related researches. In this work, we aim to reveal widespread inequality of vulnerable users' content accessibility and consumpotion in a short-video platform. Furthermore, we will explore the essential mechanism of how this inequality forms with large-scale and long-term interaction records, from both perspectives of user behaviors and algorithmic effects. Preliminary results have shown that boys and women are consuming more homogeneous content, even for new users, demonstrating the inequality from recommendation algorithms.

The onset of the Covid-19 pandemic in February 2020 prompted the U.S. federal government and states “stay-at-home” orders that led to a surge in telecommuting adoption The purpose of this project is to explore the influence of telecommuting on human mobility, and to study how the change of human mobility influence urban center, community structure and the rationality of infrastructure distribution. We first verify the correlation between telecommuting data and human mobility data, and find that there is a negative correlation between their amount of change. This is consistent with our hypothesis that an increase in telecommuting will cause a decrease in human mobility. Then, we move on to explore the impact of human mobility on urban structure. Most of the existing researches on urban structure are based on static indicators, such as population density and height of urban buildings. The innovation of our project is to use dynamic metrics, human movement, to study urban structure and its changes. We build a human mobile network before and after the epidemic based on the safegraph data set, and then (1) divide urban centers by calculating the centrality index of the network, (2) divide communities by community detection algorithm, and (3) study the impact on infrastructure distribution. I will present our preliminary conclusions in the presentation.

With the deployment of 5G networks, the huge energy consumption and carbon emissions caused by 5G networks have become a difficult problem. To solve urgent development for sustainable traditional energy-saving methods have acceptable performance on 4G networks, but they fail to solve the problem of energy saving in heterogeneous networks consisting of both 5G and 4G base stations. In this paper, we propose a data-driven method to mitigatecarbon emissions of heterogeneous 4G and 5G networks. Specifically, to realize large-scale network energy saving, we propose to divide the grid based on the co-coverage relationship of cells, use historical data to predict traffic and select active cells for the grid, and model energy efficiency profiles for the grid. Then, we offload 5G network traffic to its corresponding equivalent grid with the highest energy efficiency Profiles to save energy and reduce carbon emissions through 4G and 5G joint shutdown control. We conduct an experiment to verify the energy-saving effect of our method in Nanchang City. The experiment results show that our method can save 3442.7 MWh of energy consumption and reduce 3276.5 tons of CO2 emissions in Nanchang in a week, and its energy saving rate is 18.15% higher than the greedy method and 38.85% higher than threshold-based method.

With the development of mobile communication technologies, mobile networks have become more complicated, while there is an increasing requirement for the quality of communication services. The digital twin for communication networks can abstractly represent a real communication environment and be used as a foundation for testing services provided under actual or simulated conditions. Mobile networks consist of multiple base stations (BSs), and various parameters could be tuned to provide adequate services. Automatically optimizing such parameters remains challenging since they are sensitive to environmental fluctuations and might impact each other. Reinforcement learning (RL) techniques are potent solutions for learning configuration strategies from interactive data, but they are typically challenging to adapt to multiple agents. In this paper, we illustrate how to employ digital twin technique and reinforcement learning in mobile communication networks consisting of many collaborating agents. We explain how mobile networks may be modeled in the form of digital twins and how to handle network optimization issues effectively using multi-agent reinforcement learning (MARL). Especially, in our proposed MARL algorithm, multiple features are extracted elaborately from historical knowledge through UNet and allows to explicitly learn coordinating behaviors among all antennas through value functions. We demonstrate experimentally that the proposed MARL algorithm outperforms alternative approaches. The utilization of the optimized MARL can well generalize to a high number of agents without reducing cooperation.

Diversity in recommendations has become a growing area of interest, but attempts to improve diversity typically come at a cost of accuracy. In this work, we propose a model for improving diversity without hurting accuracy by adapting to users' individual diversity sensitivity: namely, whether certain users respond positively or negatively to more diverse recommendations. We first validate the presence of variation in users' diversity sensitivity in a representative dataset. With the help of psychology literature, we define two levels of diversity preferences based on intrinsic user traits: 1) an overall preference for variety in recommendations, and 2) a preference for diversity per item based on the strength of category interests. Then we design a method for inferring the two types of sensitivities for each user so we can adapt recommendations accordingly. To do so, we propose DivSense, a method to adaptively sample diverse neighbors in a graph convolutional network according to users' inferred diversity preferences. Our proposed approach consists of two major components: 1) entropy adjusted neighbor sampling, which addresses overall diversity sensitivity, and 2) cluster based sub-interest neighbor sampling, which addresses category based diversity sensitivity. Through extensive experiments, we show that our solution exceeds the state of the art in both diversity and accuracy. We further evaluate results for users at different extremes of sensitivity to determine that the model does, indeed, adapt to users' sensitivities to in a more optimal way than the state of the art method.

Polarization is a ubiquitous phenomenon in social systems. Empirical studies document substantial evidence for opinion polarization across social media, showing a typical bipolarized pattern devising individuals into two groups with opposite opinions. While coevolving network models have been proposed to understand polarization, existing works cannot generate a stable bipolarized structure. Moreover, a quantitative and comprehensive theoretical framework capturing generic mechanisms governing polarization remains unaddressed. In this paper, we discover a universal scaling law for opinion distributions, characterized by a set of scaling exponents. These exponents classify social systems into bipolarized and depolarized phases. We find two generic mechanisms governing the polarization dynamics and propose a coevolving framework that counts for opinion dynamics and network evolution simultaneously. Under a few generic assumptions on social interactions, we find a stable bipolarized community structure emerges naturally from the coevolving dynamics. Our theory analytically predicts two-phase transitions across three different polarization phases in line with the empirical observations for the Facebook and blogosphere datasets. Our theory not only accounts for the empirically observed scaling laws but also allows us to predict scaling exponents quantitatively.

To better understand how the wireless environment characterizes the electromagnetic signal, it is inevitable to study channel modeling. The ultimate purpose of channel modeling is to achieve high accuracy and efficiency and to support the system design and performance evaluation. The conventional methods for channel modeling are classified into two categories, the stochastic model and the deterministic model. However, they have the shortcomings of low fidelity and high computational time consumption, respectively. Recently, approaches based on machine learning revealed the feasibility of modeling channels in a data-driven way. In this work, we jointly consider the physical knowledge and the measured data. Specifically, we solve the integral form of Maxwell's equations to roughly derive the electric field distribution and then leverage the neural network to learn to correct the error. Moreover, we apply a neural network to replace the integral operation in the physical knowledge to accelerate the computation. By comparing with representative baselines, the results reveal the superiority of our proposed method.

Spatio-temporal point process (STPP) is a stochastic collection of events accompanied with time and space. Due to computational complexities, existing solutions for STPPs compromise with conditional independence between time and space, which consider the temporal and spatial distributions separately. The failure to model the joint distribution leads to limited capacities in characterizing the spatio-temporal entangled interactions given past events. In this work, we propose a novel parameterization framework for STPPs, which leverages diffusion models to learn complex spatio-temporal joint distributions. We decompose the learning of the target joint distribution into multiple steps, where each step can be faithfully described by a Gaussian distribution. To enhance the learning of each step, an elaborated spatio-temporal co-attention module is proposed to capture the interdependence between the event time and space adaptively. For the first time, we break the restrictions on spatio-temporal dependencies in existing solutions, and enable a flexible and accurate modeling paradigm for STPPs. Extensive experiments from a wide range of fields, such as epidemiology, seismology, crime, and urban mobility, demonstrate that our framework outperforms the state-of-the-art baselines remarkably, with an average improvement of over 50%. Further in-depth analyses validate its ability to capture spatio-temporal interactions, which can learn adaptively for different scenarios.

The growing complexity of next-generation networks exacerbates the modeling and algorithmic flaws of conventional network optimization methodology. In this paper, We propose a mobile network digital twin (MNDT) architecture for 6G networks. To address the modeling and algorithmic shortcomings, the MNDT uses a simulation-optimization structure. The feedback from the network simulation engine, which serves as validation for the optimizer's decision outcomes, is specifically used by an AI-based optimizer to iteratively optimize. We develop a network digital twin prototype system in practice using data-driven technology to accurately model the behavior of mobile network elements (e.g., mobile users and base stations), wireless environments, and network performance. An artificial intelligence (AI)-powered network optimizer has been developed based on the deployed MNDT prototype system. The results of the experiments demonstrate that the proposed MNDT infrastructure can provide practical network optimization solutions while being able to adapt to the more complex environment that mobile network systems are operating in.

The two-dimensional bin packing problem (2DBP) is a critical optimization problem in the industry of furniture production, glass cutting, and so on, where the objective is to cut smaller-sized items from a minimum number of large standard-sized raw materials. In practice, hundreds of customer orders (sets of items) are manufactured every day in a factory. To relieve the pressure in management, a common practice is grouping the orders into batches for production so that the items from one order are in the same batch rather than scattered all over the production line. In this work, we formulate it as the grouped 2D bin packing problem, a bi-level problem where the upper level partitions orders into groups and the lower level solves 2DBP for items in each group. The main challenges are (1) the coupled optimization of upper and lower levels, and (2) the high computational efficiency requirement for practical application. To tackle these challenges, we propose an iteration-based hierarchical reinforcement learning framework, which can learn to solve the optimization problem in a data-driven way and provide fast online performance after offline training. Extensive experiments demonstrate that our method not only achieves the best performance compared to all the baselines but is also robust to the change in dataset distribution and problem constraints. Finally, our method is deployed in the factory of ARROW Home, China, reducing the cost of raw materials by 4.1%.

Vehicle trajectory boosts various city service and management applications, and also scientific researches. The city-scale and full-amount vehicle trajectory data is especially valuable and vital in some critical applications, such as city simulation with city-scale OD generation, vehicle pollution emission calculation, and road network science with knowing the flow of each road. My talk consists of two parts. The first part introduces a city-scale vehicle trajectory recovery system based on traffic camera video data towards the release of a city-scale vehicle trajectory dataset. In this system, we address the computation resource restriction problem in this city-scale task by dividing city into a few regions. We perform recovery in each region and then carefully piece up the trajectories across regions with a trajectory merging algorithm. The second part introduces my personal brain storming and some investigations on the promising and important research directions which can especially benefit from the city-scale and full-amount vehicle trajectory data. I’m looking forward to your kind suggestions on any other interesting research directions on trajectory applications.

How to use simulation technology to promote urban science and applied research is a topic of great interest to us. In this seminar, two works on simulation technology-assisted urban science research will be shared. The first work [1] simulates the traffic congestion process in several cities with the help of microscopic traffic simulation, and the analysis finds that in the congestion process the maximum number of congested clusters (the key parameter in percolation theory) and the maximum macroscopic fundamental diagram (MFD) flow occur at the same moment. The work provides a new perspective to understand the propagation of traffic congestion. The second work [2] focuses on the evaluation of the resilience of urban transport networks. It first simulates road traffic in 40 U.S. cities under everyday conditions and calibrates the delay time calculations using real-world statistics. Second, it simulates changes in the average delay time for urban access with different percentages of road damage, giving an evaluation of the resilience of transport networks in 40 U.S. cities.

People’s beliefs about COVID-19 are constructed by not only their own knowledge and experience with COVID but also other people’s attitudes and shared information. However, not all people’s words have the same impact on us. Those that are closer to us often have more influence on us than others. In this project, through conducting surveys and experiments, we investigate how people’s beliefs toward COVID are influenced by other people, with whom they have different strengths of social ties. In this presentation, I will share our research design and preliminary findings related to the experiments. Specifically for the design, we let participants, who self-confirmed COVID infection, share their COVID experience with five people separately, with whom they are related differently, i.e., a family member, a friend, a classmate or colleague, an acquaintance, and a stranger, and measure these five people’s change of COVID beliefs. I will also showcase our preliminary findings during the presentation.

For walking pedestrians, when they are blocked by obstacles or other pedestrians, they adjust their speeds and directions to avoid colliding with them, which is called collision avoidance behavior. This behavior is the most complex part of pedestrians' walking processes and its modeling and simulation are the keys to realistic crowd simulation, which serves as the foundation for various applications. However, most existing methods either lack the representation power to accurately model the complex collision behavior or do not model it explicitly, which leads to a poor level of realism of the simulation. To realize realistic crowd simulation, we propose to analyze, understand and model the collision avoidance behavior in a data-driven way. First, to automatically detect collision avoidance behavior for further analysis, we propose a domain transformation algorithm that detects it by transforming the trajectories in the spatial domain into a new domain where the behavior is much more apparent and is thus easier to detect. The new domain also provides a new perspective for understanding collision avoidance behavior. Second, since there are no mature metrics to evaluate the level of realism, we propose a new evaluation metric based on the least-effort theory, which evaluates the realism of collision avoidance behavior by its physical and mental consumption. This evaluation metric also provides the foundation of modeling. Third, for realistic crowd simulation, we design a reinforcement learning model. It trains agents with our proposed reward function that models pedestrians' intrinsic needs of "reducing effort consumption" and thus can guide agents to behave realistically when avoiding collisions. Extensive experiments show our model is 55.9% and 52.5% more realistic in collision avoidance behavior than the best baselines on two real-world datasets.

Sequential recommendation predicts the next item based on users' historical behaviors. Longer historical behaviors often improve recommendation accuracy but bring efficient problems. Existing works for long sequential behaviors only consider the relation between historical items and target items, failing to capture the relation between historical items. Such a manner indeed fails to make full use of the historical information because there is interest drifting within the historical sequence that is helpful to predict the target item. As sequences get longer, the following two main challenges have not been addressed: (1) efficient modeling under increasing sequence length and (2) interest drifting within historical items. In Wechat's Channels, we deploy Sparse Attentive Diffusion for Long-sequence Recommendation (DiffuRec) with Sparse Attention Layer and Diffusion Attention Layer. Specifically, we first propose three sparse attention modules - global attention, window attention, and random attention - to capture the global interest drifting for short-term items, local interest drifting, and casual interest drifting for historical items. Then to expand the interest drifting of the sparse attentions, we further propose Diffusion Attention Layer to expand the receptive field of each historical item by multi-step Personalized PageRank iteration. The theoretical analysis illustrates that our proposed diffusion method can approximate full attention efficiently. Experiments on two real-world datasets show the superiority of our proposed method against state-of-the-art baselines. Further ablation study and visualization also sustain its effectiveness.

With recent advances in reinforcement learning, we have witnessed countless successes of intelligent agents from game-playing to industrial applications. Especially, multi-agent reinforcement learning (MARL) is suitable for many real-world scenarios and has vast potential applications in the real world. However, typical MARL methods can only handle tens of agents, leaving scenarios with up to hundreds or even thousands of agents almost unexplored. There exist two key challenges in scaling up the number of agents: (1) agent-agent interactions are critical in multiagent systems while the number of interactions grows quadratically with the number of agents, causing great computational complexity and difficulty in strategies-learning; (2) the strengths of interactions vary among agents and over time, making it difficult to precisely model such interactions. In this paper, we propose a Graph Attention Mean Field (GAT-MF) method. By converting agent-agent interactions into interactions between each agent and a weighted mean field, which precisely models the dynamics of the interactions, we greatly reduce the computational complexity. We mathematically prove the correctness of this approach. We design a graph attention mechanism to automatically capture the different and time-varying strengths of interactions, ensuring that our method can precisely model interactions among the agents.We conduct extensive experiments in both manual and real-world scenarios with more than 3000 agents. The results demonstrate that our method achieves superior performance with an improvement of 42.7%, and saves 86.4% training time and 19.2% GPU memory comparing the best baseline method.

Millions of slum dwellers suffer from poor accessibility to urban services due to inadequate road infrastructure within slums, and road planning for slums is critical to the sustainable development of cities. Existing re-blocking or heuristic methods are either time-consuming which cannot be generalized to different slums, or yield sub-optimal road plans in terms of accessibility and construction costs. In this paper, we present a deep reinforcement learning based approach to automatically layout roads for slums. We propose a generic graph model to capture the topological structure of a slum, and devise a novel graph neural network to select locations for the planned roads. Through masked policy optimization, our model can generate road plans that connect places in a slum at minimal construction costs. Extensive experiments on real-world slums in different countries verify the effectiveness of our model, which can significantly improve accessibility by 14.3% against existing baseline methods. Further investigations on transferring across different tasks demonstrate that our model can master road planning skills in simple scenarios and adapt them to much more complicated ones, indicating the potential of applying our model in real-world slum upgrading.

The Origin-Destination~(OD) matrix provides an estimation of number of individuals traveling between regions, i.e., mobility flow in the city, which is widely-used in urban planning, transportation, etc. Given various city characteristics of urban regions, generating the city-wide OD matrix without using historical flow information has become increasingly appealing to both researchers and practitioners. However, existing works are limited in independent generation of each element, i.e., flow, in OD matrix, overlooking the element relations within the matrix that can be well formulated as a network. In this paper, we instead propose to generate the city-wide OD matrix from the network perspective, and design a graph denoising diffusion method to learn the conditional joint probability distribution of all elements in the OD matrix given city characteristics at region level. To overcome the learning difficulty of the city-wide OD matrix covering over thousands of regions, we decompose the original one-shot generative modeling of the diffusion model into two cascaded stages, corresponding to the generation of network topology and mobility flow, respectively. To further reproduce important network properties contained in city-wide OD matrices, we design an elaborated graph denoising network structure including a node property augmentation module and a graph transformer backbone. Empirical experiments on data collected in two large US cities have verified that our method can generate OD matrices for new cities with network statistics remarkably similar with the ground truth, further achieving superior outperformance over competitive baselines in terms of the generation realism.