Despite having billions of data sensing devices that virtually acquire data about anything, anywhere, not all smart applications are intelligent. Context-awareness being at the core of intelligence, it is a contradiction founded on the impotence of these applications to effectively leverage this plethora of big data and infer context in real-time. Hence, in this paper, we propose a conceptual architecture and instantiate a working prototype of the Distributed Contextual Intelligence System (DCIS) - a ubiquitous, mobile edge-based context management mechanism. Unlike any traditional Context Management System, DCIS is an ultra lightweight, event-driven, dynamically scaling agent-based system that overcomes the barriers to ubiquitous contextfacilitating context anywhere, anytime, and everywhere so that any application can be intelligent. Our prototype - Zoolocity, is based on edge devices that execute light-weight context inference on data that it can acquire and dynamically collaborate with nearby edges in real-time. Thus, unlike any previous works that do not provide any evidence or motivation towards distributed context management, we demonstrate how applications can use DCIS agents to become fully context-aware about any scene it is involved in, unlocking any intelligent feature. This is enabled by two key novel contributions: a method to spontaneously share interoperable context and infer geospatially sparse complex activities in near real-time using generalised entity interaction modelling.
Vehicular traffic is one of the major sources of air pollution in urban settings, making it essential to clearly understand how much and where vehicle emissions impact residents. Recent approaches manage to yield pollution maps at the microscopic level by processing GPS trajectories of vehicles. That is achieved by applying mathematical models to estimate instantaneous emissions from GPS data, extending estimates to areas without data through missing data imputation, and further considering air dispersion factors. In this work, we leverage such inferred knowledge to implement an emission-aware pedestrian routing strategy and to study its impact on the reduction of exposure to vehicular pollutants and walking time. The study is realized through simulations of large masses of pedestrians over a medium-sized city in Italy, analyzing the interplay between the two factors - exposure versus walking time - in terms of time efficiency of paths and changes over existing habits both at a global and at an individual level. Experiments suggest that exposure-aware routing can yield a significant margin of improvement in health over most paths with minor effects on mobility, making it feasible and effective.
The rapid proliferation of Unmanned Aerial Vehicles (UAVs) and UAV swarm technologies has raised critical concerns about security and safety in low-altitude airspace. In response, we propose a vision-based system for detecting and tracking UAV swarms, which combines a novel UAV detection mechanism with a swarm tracking strategy. Our UAV detector incorporates parallel receptive field blocks alongside an attention mechanism to enhance detection performance. This design effectively captures multiscale features of UAVs while prioritizing salient features, ensuring robust detection under diverse conditions. For swarm tracking, we leverage the inherent formation constraints typically maintained by UAV swarms. These constraints allow us to improve tracking accuracy, particularly in scenarios involving occluded UAVs or those with weak appearance features. By integrating the enhanced UAV detector with the formation-aware swarm tracking framework, our approach achieves notable advancements in both detection and tracking performance.
Routing through large urban areas is a daily problem for the vast majority of the commuters. In this paper, we consider the extension of the routing problem where we want to find the optimal path for different objectives. Specifically, we consider the problem for finding paths that are quick, safe or efficient in the use of resources. Multi-objective routing path optimization is an active research field in the area of path optimization. We present an efficient technique for finding the optimal path which satisfies user defined constraints. Experimental evaluation leads us to the conclusion that multi-objective routing path optimization can be deployed efficiently in terms of computational cost and solution accuracy. We perform experiments comparing our work with a state of the art technique which concerns the multi-objective shortest path optimization problem.
Walking is a fundamental mode of human movement and an essential component of urban mobility. However, traditional pedestrian navigation systems lack real-time sidewalk accessibility data, relying primarily on static maps that fail to reflect dynamic hazards, such as construction zones, obstructions, or uneven pavement. To address this gap, we present a sidewalk navigation system that integrates crowdsourced reports, adaptive routing, and AI-powered assistance. The system leverages OpenStreetMap (OSM) maps and real-time user contributions to dynamically adjust pedestrian routes based on reported obstacles, ensuring a safer and more efficient walking experience. In addition, a user-driven rating system validates sidewalk conditions, while an AI assistant powered by Large Language Models (LLMs) provides context-aware guidance, interactive navigation, and additional insights. The application also features turn-by-turn navigation with audio support, enhancing accessibility for many users. By combining real-time crowdsourced data, personalized routing, and AIenhanced navigation, our system addresses critical limitations in existing pedestrian navigation applications and provides a more interactive, adaptive, and user-driven approach to sidewalk accessibility.
This Advanced Seminar provides a comprehensive overview of the research landscape of employing Large Language Models (LLMs) for Urban Mobility applications. The presented work in this seminar is categorized based on how LLMs are employed to serve various urban mobility applications. This goes from employing LLMs as a black box with a bit of prompt engineering, to fine-tuning LLMs to fit urban mobility applications, to completely retrain a vanilla LLM architecture with urban mobility data, to modifying the internal LLM loss function to fit urban mobility applications. The seminar concludes by presenting a set of benchmarking and evaluation work while pointing out to research gaps, open problems, and future research directions for employing LLMs to urban mobility applications.
Data-driven approaches to urban management and planning have attracted significant attention within interdisciplinary research spanning civil engineering and spatial data mining. Urban computing methods, particularly those leveraging graph representation learning techniques developed over the past decade, provide great potential in advancing urban management planning. In this paper, we adopt the perspective that cities can be effectively represented as network-structured graphs, wherein locations such as buildings and amenities are interconnected through road networks. These urban networks are further partitioned into subgraphs, and we introduce a novel model designed to learn effective representations of these subgraphs as distinct urban areas by utilizing heterogeneous graph neural networks. Specifically, our contributions include: firstly, developing a graph representation learning model that captures the geospatial characteristics of urban locations within dynamically-sized graphs; secondly, implementing an innovative graph pooling-based Graph Convolution Network; and finally, establishing a new benchmark for urban location representation learning tailored to urban management and planning tasks. Experimental results on various graph types demonstrate that our proposed model significantly outperforms existing methods in graph representation learning, particularly in applications involving urban location analysis for public safety and dense, heterogeneous social media networks.
Mobile Crowd Sensing (MCS) leverages the widespread availability of smart devices to collect and analyze environmental and social data. While MCS has been widely applied in smart cities to optimize vehicle traffic and safety, the needs of pedestrians remain largely unaddressed. To address this need, we propose CrossTime, a sensing application designed to estimate waiting times at pedestrian crossings. Using GPS data, accelerometer readings, and open-source intersection location data, CrossTime autonomously detects when a user is waiting at an intersection. To evaluate its feasibility and accuracy, we conducted a test case on three routes in an urban environment, comparing system-detected waiting times with manually recorded values. Our results show that CrossTime effectively captures pedestrian waiting behavior, although there are some discrepancies due to sensor limitations and environmental factors.
When a fleet of mobile units is used for surveillance and response to potential events, the geographic area of interest is often partitioned into smaller regions, and a particular (subset of) unit(s) is assigned to each region. One of the main reasons is to put a bound on the travel time for a unit in charge of responding to a new event/request from the unit's current location to the location of the occurrence of that event. In practice, the area may contain obstacles (e.g., buildings that have to be circumvented by ground mobile units or no-fly zones in case of drones). Although the problems of navigating among obstacles and spatial partitioning have been studied in the past, in this work we take a step towards tackling the setting of partitioning a geographic area of interest with obstacles in it, for the purpose of bounding the worst-case response time to an event by a member of a fleet of mobile units. To this end, we introduce a novel data structure and present an algorithmic solution for its construction, enabling a distribution of a fleet of mobile units to disjoint regions of the area of interest in a manner that will ensure a bound on the worst case response time in each region. Our experiments over real and synthetic datasets demonstrate the benefits of the proposed methodology over adaptation of existing spatial partitioning techniques.
Mobility is a fundamental aspect of human life, and mobility data offers valuable insights into user behavior. Yet, this data also exposes users to privacy risks given pattern unicity in their trajectories, i.e., the singularity in the displacements made by users. Existing strategies to quantify such user exposure either focus only on the sequences of places visited by each user, as the widely used uniqueness measure, or are tied to specific attack models. We here introduce MoBES, a novel, scalable, customizable and highly interpretable measure of user exposure in mobility data. MoBES leverages multiple existing metrics to build a multi-dimensional space, which in turn is used to capture each user's mobility signature behavior. MoBES quantifies user exposure based on how distinct a user's signature is from her neighbors in the defined metric space. As such, MoBES is designed to be a fundamental expression of user behavior, and not tied to any specific attack model. We evaluate MoBES on a real mobility dataset, showing that it effectively captures user exposure within the behavioral metric space. We also compare MoBES with the uniqueness measure, showing that MoBES is able to uncover users who, even though visiting the same places as others in the crowd, are still at risk of exposure due to the unicity of their mobility behavior.
One important “fact of life” when modeling motion, especially when it comes from real location-in-time data sources, is the uncertainty of the objects' whereabouts at a given time instant. Different models have been proposed in the literature and their impacts have been considered on algorithms for processing various spatio-temporal (continuous as well as snapshot) queries like range, k-nearest neighbor (kNN), contact, etc. However, having the available models and solutions does not bring them closer for use by domain scientist who may want to explore the impact of the uncertainty on realistic domains such as tracking of animals. In such scenarios, a domain scientist may be interested in getting an idea as to what is the possibility of an animal being within certain distance from a water supply (i.e., range query). Similarly, the domain scientists may be interested in the probability of two atoms being within a certain distance from each other (i.e., alibi query). To enable such analysis, we developed a prototype system that offers: (1) selection of a dataset consisting of ($x, y, t$) points corresponding to discrete locations of the objects' motion; (2) visualization of the uncertainty of that motion under the model of space-time prisms (also referred to as beads); (3) Visualization of the answers to range and alibi queries (and, where applicable, the map of the given area). In this demo paper we describe the basic architecture of our prototype system and discuss its functionalities.
Today it is simple to collect and transmit GNSS data from vehicles with a high frequency. However, there is a storage and processing cost related to handling the data. Further, some data has limited value, e.g., redundant GNSS data from a vehicle stopped at an intersection. In this paper, sampling methods for GNSS data focusing on time, distance, speed, and heading changes are systematically analyzed. The goal is to retain only valuable data. A set of metrics is proposed to quantify the value of the data, e.g., no redundancy and retention of the spatial and temporal distributions. An existing commercial approach to GNSS-based travel time computation in road networks is used to measure if the sampled GNSS is accurate for this important purpose. The results show that sampling methods using individual properties, such as time, space, or speed, have their own strengths and weaknesses. However, with hybrid methods, it is possible to retain the strengths and eliminate most weaknesses. Using a large, real-world GNSS dataset, we show that a hybrid method that retains only 20% of the original data can achieve travel time estimation with an error of just 1.0-1.3%.
This paper presents a framework for remote monitoring of unmanned aerial vehicles (UAVs) developed specifically for urban environments through crowd sensing. This system is built around a centralized control mechanism that manages UAV traffic to reduce the risks of airspace congestion and sudden accidents. The UAVs periodically broadcast their remote identifications (RIDs), which are received by ground observers on their mobile devices through dedicated applications. These RIDs contain multiple information, including the drone's ID, location, altitude, velocity, and a timestamp. The ground observers then forward the received RIDs to the surveillance station after appending their coordinates, ID, and a time stamp. A key component of our approach is the incorporation of signal quality metrics - path loss, shadowing, and received signal strength (RSS) - along with distance considerations in the activation process of ground observers for RID message forwarding. To address potential data congestion and incentive overuse in densely populated areas, our model uses a predictive strategy for dynamic observer activation, ensuring that the surveillance system processes an optimal number of RID reports efficiently. This comprehensive consideration of both signal integrity and spatial proximity significantly improves the detection and monitoring precision of UAVs, increasing the system's ability to effectively monitor UAV activities while conserving resources in urban environments.
The emergence of Vision Transformer (ViT) models and their variants (e.g., Swin Transformers) are prevalent in recent years due to their higher accuracy in vision AI applications. However, their efficient execution in edge computing environments (e.g., mobile phones/embedded platforms) remains a challenge due to the heavy computational demands (both GPU cycles and GPU memory) of these large-sized models. To serve these models efficiently at the edge, we introduce a novel approach combining model slicing and smart batching to distribute workloads between resource-constrained client devices and powerful edge servers. Model slicing allows breaking a large model into smaller segments, called slices, and let the client execute a few initial but a variable number of slices (head slices) and a nearby edge server runs the rest of the slices (tail slices), smart batching enables the server to queue several requests from multiple clients batch together for inference leading to better GPU resource utilization. We propose two batching strategies at the server: one runs faster but requires higher GPU memory and the other one demands less memory with slight overhead of internal data movement. Experimental results show that our approach achieves inference time reductions of up to 67 % while maintaining high GPU utilization and demonstrates significant improvements in inference speed, showcasing the viability of this approach for distributed AI systems.
A challenge overlooked in prior blockchain algorithms is that they do not consider large-scale network outages and rely on the assumption of a reliable global network connectivity. In the event of a large-scale network partition, forks may occur between partitioned regions. After the partition ends, forks will be discarded, leading to the loss of many blocks and a considerable amount of wasted work. In this paper, we propose a sharding mechanism to improve blockchain's resilience to the possibility of a global internet outage. We form consensus groups dynamically and consider the partitioning of the group as a hint to split the blockchain into branches and guarantee that all of them will be merged after the network is recovered. We indicate different methodologies to ensure blockchain security while partitioning occurs. Our experiments use simulations to show how this approach can improve the performance of blockchain algorithms and prevent wasted computational power during partitioning.
The ship trajectories collected by the Automatic Identification System (AIS) are widely used in maritime applications. However, a significant issue with AIS data is that large AIS gaps occur. Existing trajectory imputation methods for AIS data have three main limitations: (1) the temporal aspect is ignored; (2) the methods fall short when dealing with complex ship movements; (3) the common-route assumption does not always hold. To overcome these limitations, we propose TrajImpMC, a tracking-based framework that uses polygon-based ship location estimates from multiple cameras to impute large AIS gaps. TrajImpMC combines speed constraints and Kalman filters, and can return imputed trajectories that contain both spatial and temporal information. Extensive experiments are conducted on real datasets. In terms of the quality of the imputed trajectories, TrajImpMC improves the RMSE errors by at least one order of magnitude over two existing state-of-the-art AIS imputation methods. In addition, a visual comparison shows that the imputed trajectories of TrajImpMC align very well with the real ship trajectories during AIS gaps. The code for this paper is available at: https://github.com/songwu0001/TrajImpMC.
We investigate the challenge of generating OverpassQL from natural language in the Text-to-OverpassQL task and explore the data in the existing OverpassNL dataset. To address the structural mismatch between natural language and OverpassQL, we propose a task decomposition-based multi-step prompting approach that generates auxiliary information to help align natural language with OverpassQL structures, thereby enhancing model performance. Furthermore, we introduce a Key-Value Correction Module specifically targeting key-value pair matching difficulties in Text-to-OverpassQL tasks, designed to rectify potential syntactic errors and key-value mismatches in generated queries. Our experiments on GPT-3.5 Turbo and GPT-4 demonstrate absolute performance gains of 1.4% and 0.6% respectively. Under retrieval-augmented setting ablation, we achieve a more significant 3.5% improvement with GPT-3.5 Turbo. Experimental results confirm that our method consistently improves performance across various models and configurations, particularly showing enhanced effectiveness in medium and small-scale models.
Schistosomiasis remains a significant public health challenge in Africa, where it is considered endemic, particularly in areas with inadequate access to clean water. This study focuses on mitigating the spread of schistosomiasis by monitoring the physicochemical parameters of water sources to detect the proliferation of infected snails, which serve as the intermediate host for the parasite. We propose a system that leverages sensing technology to monitor water physicochemical parameters, e.g., temperature (Temp), pH, and electrical conductivity (EC) and uses AI to predict their future values, alongside the dynamics of the infected snail population, which are crucial for schistosomiasis transmission. By forecasting these trends, the system facilitates proactive interventions, such as water treatment and sanitation improvements. To accomplish this, we leverage FNet, a model that effectively captures temporal features through discrete Fourier transforms (DFT), delivering Transformer-level accuracy while being more resource-efficient. While FNet is originally tailored for language processing tasks, we adapt it for time series analysis by employing a quantization technique. This approach is particularly advantageous in contexts where computational and operational efficiency is critical, making it well-suited for environments with limited infrastructure or financial flexibility. Our evaluation demonstrates that the FNet model outperforms transformers in terms of resource efficiency while maintaining similar performance, making it a promising tool for real-time monitoring and early detection of schistosomiasis risk. The results highlight the potential of this approach to enhance disease control efforts and contribute to improved public health outcomes in endemic regions.
IoT deployments in smart spaces can enable the development of useful services for their inhabitants. However, the diversity of smart spaces and their sensor infrastructures makes it challenging to develop space-agnostic applications. Moreover, existing schemas addressing interoperability challenges often lack the vocabulary needed to represent the integration of smart space systems and their inhabitants. We present a schema to annotate inhabited smart spaces in support of inhabitant-oriented applications. Our schema integrates well-known ontologies to represent inhabitants, events/activities, and the space itself, along with their interconnections. It also supports the representation of uncertain information from IoT and mobile sensors (e.g., a person's location or occupancy/attendance at an event). Additionally, we introduce an annotation tool that uses an easy-to-use GUI to describe a smart space based on our schema. We demonstrate the potential of our approach through a series of SPARQL queries and a system deployed at the UCI campus that annotates sensor data to support a space-agnostic occupancy monitoring application.
Understanding human patterns-of-life (PoL) is essential towards ensuring safe and secure indoor facility environment as well as outdoor urban environment. Prediction of human movement in between places of interest is vital in understanding human PoL. Movement between spaces maybe represented and detected in one of the two forms: 1) trajectories: locations measured at regular time intervals by mobile sensors, bluetooth or GPS sensors; or 2) stay transitions: semantic PoI (points of interest) and stay duration data measurable by eventbased sensors that collect data when a check-in or check-out event is detected. Stay transition data provides a more compressed data format compared to trajectories data, especially in situations with longer stay durations, while preserving the information necessary for PoL analysis. Now as introduced briefly in the paper, our deployed end application (Digital Twin of a facility with non-player characters, besides the interactive user in virtual reality) needed a well-performing and validated AI/ML model for simulating high quality stay transitions behavior. In this study we thus primarily present our findings with developing and validating that model, which is a multi-task neural network for stay transition prediction. The neural network consists of two heads, for corresponding two tasks of stay category prediction and stay duration prediction. We evaluated gated recurrent units and multi-layer perceptrons of varying network sizes for stay category prediction; while mixture density networks, noisy generator-only networks, and generative adversarial networks of varying network sizes for stay duration prediction. We have then evaluated four multi-task models, constructed by combining these specialized models, on their ability to predict stay transition data. We tested our models on datasets from two different cases: 1) a simulation-generated dataset of indoor movement within the HFIR (high flux isotope reactor) nuclear reactor facility at Oak Ridge National Laboratory (ORNL); and 2) the GeoLife human mobility dataset of outdoor urban movement available in literature. Our results indicate that GRUMDN, which combines gated recurrent units (GRU) for stay category prediction task, and mixture density networks (MDN) for stay duration prediction task, did overall outperform other multitask models and the current state-of-the-art.
