Past decades have witnessed the rapid development of battery-free backscatter-based cameras, opening up promising long-term and sustainable surveillance applications by harvesting wireless energy. However, limited by the fluctuations of the ambient energy, existing backscatter systems suffer from unreliability issues, e.g., stalls and frame losses in video transmission. To tackle this problem, we propose EATag, the first energy-aware backscatter system that can provide reliable battery-free video frame transmission through adaptation. EATag provides hardware and software co-designs to achieve energy budgeting, prediction, capacity estimation, and adaptation for reliable video transmission. By leveraging the reader to monitor the operation duty cycles and estimate the channel fluctuation, the tag only needs to consume a little energy to achieve efficient adaptation. We implement the EATag system with off-the-shelf components. We conduct extensive experiments to evaluate the performance of EATag. Results demonstrate that EATag can achieve a stalling rate reduction of 57.3%-89.2% when compared to the existing battery-free video systems. We also present an IC schematic design of the tag with TSMC 90 nm standard cells. Circuit simulation results show that it can potentially consume only 58 μ W to provide reliable battery-free video transmission.
The rapid growth of IoT-powered smart communities has created valuable, but siloed data, limited by interoperability and data-sovereignty concerns. We present ComDeX , a context-aware federated IoT middleware that enables selective, intelligent collaboration across communities. ComDeX uses a property-graph data model to encode IoT entities as context-rich pub/sub messages, discovered and filtered via an advertisement-based federation, so communities share only what is relevant while retaining control. To support collaborative intelligence, ComDeX adds a cross-organization Discovery → Eligibility → Enrollment (DEE) layer that enables discovery of FL tasks and clients, independent of the chosen FL runtime. We prototype ComDeX on MQTT and NGSI-LD and evaluate in a smart-port scenario, demonstrating faster and more efficient cross-community exchange than existing NGSI-LD federation approaches. In non-IID FL experiments, we showcase cross-community enrollment capabilities and expose different DEE coordination tradeoffs. Overall, ComDeX enables scalable, adaptive collaboration for data sharing and federated model training.
Autonomous Driving (AD) has garnered significant attention in recent years across multiple domains. Despite notable advancements in algorithms and hardware, large-scale deployment of autonomous vehicles remains constrained by the lack of adequate real-time guarantees. This article comprehensively investigates real-time assurance challenges in AD systems from theoretical and practical perspectives. First, we introduce foundational concepts of real-time systems and analyze common modeling approaches, including the multi-rate DAG and processing chain DAG models. We then delve into the task scheduling and communication mechanisms of three representative middleware systems–ROS2, Cyber, and ERDOS–and the seL4 operating system. Our analysis reveals their underlying design philosophies and optimization strategies for real-time performance. The findings highlight the critical difficulty of meeting stringent real-time requirements in autonomous systems. By offering insights into current limitations and opportunities for improvement, this article aims to establish a deeper understanding of real-time assurance issues and encourage greater focus on system-level guarantees within the AD community.
As an important technology of computer vision, semantic segmentation is widely used in the Internet of Things (IoT), such as autonomous driving, intelligent healthcare, and Electronic Travel Aids (ETAs). It can segment the images captured by the IoT devices into different semantic areas, thereby helping them to more accurately understand image content and achieve more intelligent control and application. The security of semantic segmentation models directly affects the security of the IoT applications, especially when used for important security tasks. If attackers can control the model’s output, it will bring catastrophic consequences, even casualties. We propose a backdoor attack method against semantic segmentation models, the Weather Backdoor Attack (WBA). In our method, we innovatively propose to use weather features as triggers, which not only improves the stealthiness of triggers, but also achieves a high attack success rate. The experimental results show that our proposed WBA method can attack the semantic segmentation model with a high success rate by poisoning only a small portion of the training data. Under the settings of different victim classes, target classes, and various weather conditions, our proposed WBA method can also maintain the effectiveness of attacks. Meanwhile, due to the fact that the trigger we designed in this article simulates real-world weather conditions, it also has higher stealthiness.
The use of low-cost sensors (LCS) in Internet of Things (IoT) networks offers a promising way to improve air quality monitoring. However, there is a major concern regarding their long-term accuracy due to continuous data drift, which requires frequent data recalibration. To address this, we present window-based uncertainty drift detection and recalibration (W-UDDR), a unified system that automates the entire process. Our system uses a Bayesian approach with Gaussian processes (GP) to automatically and accurately detect when sensor output needs to be corrected. To achieve this, the uncertainty of the estimates is quantified using predictive confidence intervals alongside a window system that detects the need for recalibration in real time. We validate our approach using an air quality real-world sensor deployment, systematically assessing key performance metrics such as the frequency of recalibration and the required sample size. Our results show that W-UDDR successfully triggers automatic events when it detects drifts, achieving significant long-term accuracy improvements ranging from 40% to 95%. Hence, this tool provides an automated real-time mechanism to facilitate long-term sensor deployment maintenance.
We consider piece-wise linear lossy compression schemes for wireless sensor network applications, involving streaming sensor data, as small-size wireless packet payloads. Comparisons of existing schemes can mislead if no due care is paid to the exact cost, that we call representation cost , which counts how many bits are required to represent the sequence of the produced line segments. We rectify the situation in the presented work by presenting a fair means to evaluate the representation cost of competing schemes. Schemes that produce a small number of line segments can result in poor compression ratio when one considers the representation cost involved. Additionally, we outline an extremely competitive scheme which, at the cost of more computation, satisfies a given size for the compressed data output, while minimizing L ∞ error. It improves the compression ratio, reducing the representation cost by up to 140% over previous schemes while also practically achieving a lower maximum reconstruction error than the reconstruction error bound requested. We address several technicalities for fairly mapping, and evaluating, existing schemes that, when originally proposed, were indifferent to data representation concerns. We evaluate the compression schemes using a number of datasets, composed of different sensor types, compressed under a range of reconstruction error bounds and payload size restrictions, including some data that are challenging for any piece-wise linear compression scheme.
In last-mile logistics, couriers are assigned to deliver orders to customers and conduct real-time generated pick-up orders. Meanwhile, salespeople are hired to conduct customer expansion tasks, i.e., visiting and attracting potential customers, to increase the market share. Effectively scheduling heterogeneous workers (i.e., couriers and salespeople) for heterogeneous tasks (i.e., delivery, pick-up, and customer expansion tasks) is crucial for enhancing operational efficiency and driving revenue growth. In the state-of-the-practice, couriers and salespeople are scheduled separately, leading to a sub-optimal task completion rate. Therefore, we explore a cooperation-based worker scheduling scheme, where couriers and salespeople can also conduct customer expansion and pick-up tasks, respectively. We formulate the problem as a Markov Decision Process and design a symbolic efficiency-aware cooperative decision system called ECOS for heterogeneous worker scheduling to optimize task completion rate. Firstly, a Customized Task-Worker Representation module is designed to predict the spatial-temporal distribution of pick-up tasks and capture workers’ task-specific efficiencies and heterogeneous task-worker distributions. Then, an Efficiency-aware Multi-agent Actor-Critic module is proposed to achieve long-term optimization of the sequential worker scheduling in complicated environments. We have implemented ECOS in one leading logistics company in China, and evaluation results demonstrate that ECOS effectively improves completion rates of heterogeneous tasks.
The Internet of Things (IoT) is rapidly expanding in agriculture, healthcare, and industry, where secure real-time communication is essential. IoT devices often use lightweight protocols such as MQTT due to their limited resources. However, weak security exposes them to cyberattacks, increasing the need for effective Intrusion Detection Systems (IDS). Traditional heuristic methods, including signature-based IDSs, face challenges in MQTT environments due to limited protocol awareness, encrypted payloads, and elusive traffic patterns, leading to high false-positive rates. However, AI-based IDSs improve detection but often suffer from high computational costs and adapt poorly to zero-day attacks. To address these challenges, we propose MGIDS , a hybrid IDS for MQTT IoT networks that handles both encrypted and unencrypted traffic. MGIDS uses unified covariance-distance encoding ( CovDist ) maps flows into quantised bins, which are processed using a dynamic grid-based graph structure ( DYNGrid ). While MGIDS defines the overall framework, DYNGrid serves as the core learning and inference engine. The system adopts a two-stage hierarchical classification strategy, where Stage-1 performs coarse-grained separation, and Stage-2 refines decisions into specific attack types. The proposed method achieves 97.82%, 93.08%, and 97.26% accuracy for known attacks, and 92.80%, 84.86%, and 94.80% for unknown attacks, outperforming existing approaches while demonstrating scalability and robustness.
Wearable devices are becoming more prevalent in people’s daily lives, particularly in applications such as activity recognition, health monitoring and fitness management. However, the majority of existing wearable devices remain heavily dependent on battery-based power sources, which introduces several practical and sustainability challenges. Frequent battery replacement or recharging imposes inconvenience on users, increases long-term operational costs, and contributes to environmental concerns associated with battery disposal and resource consumption. To address these issues, we present KineticWear, the first battery-free wearable system that utilises kinetic energy harvested from human activities both as the sole energy source and as a sensing signal for on-device human activity recognition (HAR). Based on a careful end-to-end design of all hardware and software components, KineticWear achieves real-time HAR on an ultra low-power microcontroller unit (MCU) including on-board classification and transmission of the inferred activity over a wireless link. Using empirical data, we find that decision tree (DT) and convolutional neural network (CNN) models offer activity recognition accuracies of 87 % and 99.5 %, respectively. Systematic real-world experiments demonstrate that KineticWear harvests sufficient energy to operate the wearable device up to 95.2 % of the time, and that the device can infer and report an ongoing activity within 8 seconds using DT classification algorithm, taking three orders of magnitude shorter classification time than CNN. Thus, KineticWear offers significantly enhanced performance compared with state-of-the-art off-device activity recognition systems powered by kinetic energy harvesting.
Multi-modality machine learning (ML) systems are increasingly deployed in edge environments, such as smart building and autonomous robotic applications. These systems enable analysis of complex subjects by performing ML inferences on multiple data sources provided by Internet of Things (IoT)-enabled devices. However, operating such ML systems for multi-tenant applications presents a runtime orchestration challenge to optimize the execution of multiple inference tasks across modalities. Analysis requests from different tenants often rely on different data sources with distinct quality requirements, resulting in a wide range of conflicting optimization objectives for runtime inference orchestration. The diversity of ML models and their performance variations further complicate the orchestration, particularly in scheduling inference tasks and distributing inference workloads on a heterogeneous edge system. This article addresses these challenges by introducing an adaptive orchestration that manages inference task scheduling for multi-tenant applications in a multi-modality ML system. Our orchestration supports a complex tradeoff between quality refinements and inference capability of inference services while scheduling multiple inference tasks under dynamic time constraints. The orchestration employs an efficient mechanism for selecting instances of inference services to distribute inference workloads across a heterogeneous edge cluster. This mechanism allows the orchestration to leverage cross-modal information to refine inference quality following the tenant objectives. Furthermore, our orchestration considers the performance variation caused by resource contentions, runtime failures, and explainability overheads while distributing workloads. Extensive experiments on two real-world edge applications in a common heterogeneous edge testbed demonstrate that our orchestration can improve inference accuracy by up to 10% and reduce the late response ratio by up to 18% compared to existing approaches within our multi-tenant edge scenarios.
Losing track of reading progress when switching lines can be frustrating. Eye gaze tracking technology offers a potential solution by highlighting read paragraphs, aiding users in avoiding wrong line switches. However, the gap between gaze tracking accuracy (2–3 cm) and text line spacing (3–5 mm) makes direct application impractical. Existing methods leverage the linear reading pattern but fail during jump reading. This article presents a reading tracking and highlighting system that supports both linear and jump reading. The system leverages the large language model’s contextual perception capability in aiding reading tracking. A reading tracking domain-specific line-gaze alignment opportunity is also exploited to enable dynamic and frequent calibration of the gaze results. Controlled experiments demonstrate reliable linear reading tracking with performance comparable to that of the state-of-the-art linear reading tracking solution, while supporting jump reading tracking with 84% accuracy. Furthermore, real-world field tests with 24 volunteers demonstrated the system’s effectiveness in tracking and highlighting read paragraphs, improving reading efficiency, and enhancing user experience.
Below-ground biomass (BGB) of root tubers is an important phenotypic trait in crop monitoring and other agricultural applications. This article proposes a novel tuber biomass sensing (TBS) framework that uses internet of things (IoT) devices to enable non-destructive estimation of below-ground root tuber biomass. Specifically, we perform extensive experiments to build a new BGB dataset with more than 700,000 received signal strength (RSS) measurements collected by our low-cost wireless network. Then, we propose a novel data-driven model that integrates convolution neural networks, residual connections, and attention mechanisms to facilitate discriminative feature extraction from RSS data and achieve state-of-the-art (SOTA) performance in biomass estimation. In addition, to mitigate performance degradation caused by imbalanced training data, we propose a contrastive learning method that aligns feature representations of samples with similar biomass values while increasing the separation between those with significantly different values. This method reduces estimation bias toward high-frequency biomass labels, thereby improving the performance and generalizability of the data-driven model. Experimental results demonstrate the efficacy of the proposed TBS framework. Our dataset and pre-trained models are publicly available on https://zenodo.org/records/15000852 .
ACM TIOT launched a special issue on the theme of LLM Empowered Internet of Things, exploring the intersection of Large Language Models (LLMs) and the Internet of Things (IoT). As IoT continues to expand, advanced computational models are increasingly essential for processing and analyzing the massive data generated by interconnected devices. This special issue focuses on how LLMs can enhance IoT systems in several key areas. The second part of this special issue introduces the remaining six accepted papers that spans a board range of IoT scenarios from embedded and cyber-physical systems, human-centered applications, to IoT security.
Fully decentralized model training for on-road vehicles enables leveraging crowdsourced data without relying on central servers, infrastructure, or persistent Internet connectivity. However, real-world vehicular scenarios pose fundamental challenges to decentralized learning, including highly dynamic network topology, unreliable wireless links, and heterogeneous radio capability constraints. To address these challenges, we propose RoADTrain , a route-assisted decentralized peer model training framework with formal convergence guarantees. In RoADTrain, vehicles share short-horizon route information to estimate inter-vehicle contact duration and link reliability, which are embedded into a base graph. Vehicles then select communication partners by maximizing the algebraic connectivity of the resulting communication graph, thereby accelerating information mixing and model convergence. We further investigate subcarrier reuse and extend the algorithm to three representative radio capability cases. Extensive evaluations show that RoADTrain achieves convergence comparable to communication-unconstrained state-of-the-art approaches while reducing communication overhead by up to 4.5×, and significantly outperforms communication-constrained decentralized baselines by up to 29% higher driving success rates in online evaluations. We further empirically demonstrate the robustness of RoADTrain under a range of unexpected effects encountered in complex real-world applications, as well as in a large-scale traffic scenario.
The growth of capabilities of mobile devices allows them to host increasingly sophisticated applications. These capabilities have piqued interest in domains such as automotive, gaming, or 5G-based telecommunications, in use cases where hosting services in mobile devices is no longer a mere option, but has become a necessity due to stringent Quality of Service requirements. In this context, the emerging Hybrid Edge Cloud (HEC) paradigm provides an infrastructure for service execution in end devices. However, the highly dynamic and distributed nature of end devices makes the orchestration of services executed in them an open challenge; paramount to the operation of the system, as devices may often need to delegate services to other devices. We present the Collaborative Edge Cloud (CEC), an extension of HEC in which service orchestration and delegation are decoupled from a service’s business logic. Moreover, as a proof of concept of CEC, we also present the Pervasive Delegation of Services (PODS) platform, an implementation of CEC that allows services to be executed in end devices and orchestrated independently of their logic in an entirely decentralized way. Our evaluation shows that PODS, and thus CEC, improves the lifetime of collaborative services up to 4.9 × relative to other approaches.
As a core component of intelligent transportation systems, vehicle localization technology enables accurate positioning, supporting comprehensive insights into traffic flow, vehicle status, and environmental changes. Current vehicle position localization technologies primarily rely on visual sensors and the Global Positioning System, while their performance can be affected by extreme weather conditions and signal stability. However, vehicle-generated sound, as a stable data source unaffected by environmental conditions and free from signal limitations, is often underutilized by existing studies. In this paper, we construct a vehicle localization dataset based on sound signals and further propose a combined filtering strategy that integrates adaptive filtering with spectral subtraction filtering, dynamically adjusting the filter parameters to suppress time-correlated noise within the signal. We also remove broadband noise in the frequency domain while preserving high-frequency signal details, offering a significant advantage over existing methods in terms of signal-to-noise ratio improvement. The proposed dataset and filtering strategy are validated using the EfficientNet-1D Fusion model. Experimental results demonstrate that the proposed combined filtering method excels in recognition accuracy and computational efficiency.
The Internet of Things (IoT) enables large-scale distributed sensing, improving automation and real-time decision-making across domains such as smart cities and environmental monitoring. However, as IoT deployments increase in scale and complexity, ensuring the reliability of sensor nodes becomes increasingly challenging. A key issue is the timely and resource-efficient detection of faulty behavior in IoT nodes. Most existing solutions focus on a limited set of fault types, such as communication failures or sensor data anomalies, leaving many faults undetected. In this paper, we propose VarLogger, a lightweight runtime anomaly detection method. Our key observation is that many IoT faults manifest as sub-normal software behavior, such as delayed function responses or repeated hardware access. VarLogger therefore monitors device-internal software event traces and their temporal and spatial properties. We design and evaluate two unsupervised learning-based methods to capture normal event traces and detect sub-normal behavior directly on the edge. Evaluation using real-world use cases and diverse fault types (software, communication, and others) demonstrates that the time-based method outperforms state-of-the-art approaches. VarLogger achieves an F1 score above 0.9, while memory and CPU analyses confirm its suitability for resource-constrained devices. Most importantly, VarLogger can detect previously unknown faults in a real-time and resource-efficient manner.
Detecting earthquakes in seismological time series is a core task in observational seismology, supporting a range of applications from early warning systems to tectonic research. Typically, seismic sensors passively record data and send it to the cloud or edge for integration, storage, and analysis. While this cloud-based approach is effective in urban or well-connected areas, it is impractical in remote, underwater, or underground environments where network infrastructure is unreliable. In such settings, the sensors must operate independently for extended periods while coping with strict constraints on power, memory, and connectivity. To address these challenges, we present LightEQ, a system that combines an efficient data processing pipeline and a lightweight deep-learning model specifically designed for seismic event detection in such environments. LightEQ runs on ultra-low-power microcontrollers with just 100 kB of RAM, enabling real-time, on-device earthquake detection without the need for continuous streaming of raw data to a central location. We evaluate LightEQ against a traditional STA/LTA approach and state-of-the-art (SOTA) machine learning models, using the Stanford Earthquake Dataset. Unlike existing neural network (NN) models, which are too large for microcontrollers, LightEQ is over ten times smaller than most of the SOTA models. Our results demonstrate that communication is the most energy-intensive task in this setting, and that traditional model-driven filters like STA/LTA are inefficient due to their high false positive rate. In contrast, LightEQ improves detection accuracy with NN, providing a more energy-efficient solution by reducing the number of false positives before transmission. Compared to the STA/LTA method alone, LightEQ extends battery life by at least 3-fold by minimizing energy consumption associated with transmitting false positives to the cloud.
The rapid expansion of Internet of Things (IoT) systems has transformed industries through real-time monitoring and automation, generating vast and heterogeneous data streams. As IoT networks expand, the increasing volume and diversity of data, spanning real-time telemetry, device logs, and historical records, complicate the management of IoT systems, including system monitoring, analysis, and reasoning. To address this challenge, we introduce LUMEN (Large Language Models as Unified Multi-Agent Systems for IoT ENhancement), a novel approach combining multi-agent Large Language Models (LLMs), knowledge graphs, and heterogeneous databases to enable cognitive digital twins for IoT observability. LUMEN models IoT systems as knowledge graphs, capturing device relationships and metadata while monitoring data is stored in time-series or object databases. Specialized LLM-based agents collaborate dynamically to analyze IoT systems and explain the findings in natural language, generating and executing analysis code when necessary. Integrated with off-the-shelf network monitoring tools, LUMEN facilitates semantic reasoning and human-in-the-loop collaboration, delivering adaptive insights across diverse data contexts. Two industrial case studies demonstrate the ability of LUMEN to automate analysis workflows, enhance system adaptability, and provide interpretable analytics. This work advances IoT observability by integrating LLMs, semantic intelligence, and explainable analytics into a scalable and adaptive solution using a multi-agent architecture for complex IoT systems.
