Deep neural networks (DNNs) deployed on edge devices often suffer from severe performance degradation when exposed to dynamic and continually shifting environments. Test-time adaptation (TTA) has emerged as a promising solution by updating models online with incoming test data. However, edge deployment poses unique challenges: limited computational resources, latency caused by adaptation delays, and knowledge isolation across devices. The situation becomes even more complex in open-world scenarios, where the presence of unknown categories further disrupts adaptation. To overcome these limitations, we propose TeamTTA, a cloud-integrated framework designed for efficient multi-device collaboration open-set test-time adaptation. Specifically, TeamTTA aggregates reliable samples from multiple edge devices through crowdsourcing, uploads them to the cloud, and maintains a memory buffer for continual adaptation. A large vision model (LVM) in the cloud leverages its zero-shot generalization ability to filter out open-set samples and acts as a teacher model, distilling its knowledge into a replicated student edge model stored in the cloud. The adapted model parameters, or alternatively global statistics under poor network conditions, are then transmitted back to the edge devices for efficient inference. Extensive experiments on standard public TTA benchmarks, including corrupted and open-set datasets, show that TeamTTA achieves superior adaptation accuracy, robustness to distribution shifts, and communication efficiency, outperforming state-of-the-art TTA baselines. These results validate the effectiveness of integrating cloud-edge collaboration and LVM-driven knowledge distillation for real-world edge intelligence.

The dynamics of multivariate time series (MTS) data are jointly characterized by its nonlinear temporal dependencies and complex variable dependencies, making unsupervised time series anomaly detection a challenging task. Existing methods primarily rely on prediction or reconstruction errors, neglecting the valuable information within the variable dependencies. In this paper, we propose a variable dependency discrepancy-based Transformer (VDDFormer) for unsupervised MTS anomaly detection. VDDFormer comprises a variable correlation encoder, a temporal dependency encoder, and a reconstruction decoder. The variable correlation encoder capitalizes on a variable dependency attention mechanism, which employs self-attention to learn the global variable dependencies; meanwhile, the local variable dependencies are captured by the adaptive correlation matrix. The global and local variable dependencies are then used to compute the variable dependency discrepancy as a new intrinsic property to distinguish between normal and abnormal patterns. By integrating this new discrepancy with the reconstruction error, the model effectively enhances its anomaly differentiation capability. Extensive experiments on five real-world anomaly detection datasets demonstrate that VDDFormer effectively and robustly detects group anomaly patterns by leveraging the variable dependency discrepancy and achieves state-of-the-art performance on four out of the five datasets.

Accurate classification of healthcare time series is critical for clinical decision-making. However, existing models often struggle under real-world data shifts and lack interpretability- two key requirements for reliable medical deployment. To address these challenges, we propose SHINE, a novel endto-end framework that learns disentangled and shift-invariant representations by modeling the generative process of multivariate healthcare signals. Specifically, SHINE first introduces a genuine data representation learning that disentangles healthcare signals into trend, seasonality, and noise components, reflecting distinct temporal dynamics of healthcare series. Then, we inject several inductive biases into each component to encourage latent representations to be invariant to data shifts and aligned with their corresponding semantic units. Extensive experiments on six healthcare benchmarks spanning ECG, EEG, and continuous glucose monitoring (CGM) domains-under a variety of simulated real-world shift scenarios-demonstrate that SHINE consistently outperforms state-of-the-art baselines, providing robust performance and clinically meaningful interpretations grounded in the estimated components.

Efficient data stream processing, particularly for persistence estimation, is crucial in handling high-velocity data streams characterized by skewed distributions of item frequencies. Unlike more straightforward frequency metrics, persistence captures items' recurrence across multiple time windows, posing a significant challenge to existing single-structure sketches where high-persistence and low-persistence items collide. To address this, we introduce the Hypersistent Sketch, a unified framework for high-performance estimation built on two decoupled mechanisms: filtering and accelerating. The filtering component, a Cold Filter, directly addresses the skewed nature of data streams. It separates hot items from the majority of cold ones, which allows for differential treatment. The accelerating component, a Burst Filter, then optimizes the processing of hot items. It significantly improves throughput by preventing repeated insertions within a single window. We demonstrate its generality by applying it to various state-of-the-art sketches (e.g., On-Off, Waving, P-Sketch), showing it consistently enhances their original performance. We also deploy our framework on Redis platforms, demonstrating the framework’s broad applicability and scalability.

Autonomous farm vehicles (AFVs) encounter significant challenges in large-scale networking and massive data transmission. The rapid development of global low Earth orbit (LEO) satellite networks provides reliable support for AFVs. However, the time-varying characteristics of the satellite-terrestrial channel and large-scale collaborative scheduling among AFVs pose challenges for joint computation offloading between satellites and AFVs. This paper proposes a satellite and autonomous farm vehicle integrated network (SAFVIN) architecture. We formulate the joint satellite and AFVs computation offloading problem as a Markov decision process (MDP). We propose a deep rein forcement computation offloading (DRCO) method that adapts to satellite networks. Unlike traditional computation offloading methods, the proposed DRCO takes into account the time varying satellite network channel states. The DRCO can rapidly converge to high-quality decisions in satellite network with strong randomness, thereby adapting to dynamic environments more quickly and achieving superior performance. We compare the proposed DRCO with the heuristic coordinate descent (CD), and with deep Q-network (DQN) and deep deterministic policy gradient (DDPG) algorithms. The DRCO achieves a 2% lower latency loss while only incurring 21% of the time overhead required by the CD. Furthermore, unlike DQN and DDPG algorithms, which rely on continuous time frame input and output for network updates, the proposed DRCO can directly leverage past experience to adapt to dynamic satellite network. Compared with other deep reinforcement learning algorithms including DQN and DDPG, the DRCO achieves an average energy consumption reduction of approximately 10%.

Human motion prediction from observed 3D skeleton sequences is a challenging task in computer vision, primarily due to the difficulty of learning robust and universal feature representations. This paper introduces a novel self-supervised learning framework for human motion prediction, pioneering the use of masked reconstruction pretraining in this domain. Our framework comprises two phases: pretraining and formal training. Both phases segment motion sequences into temporal patches to enable efficient encoding, where pretraining randomly masks patch subsets to reconstruct the full sequence using unmasked patches, and formal training is formulated as an extreme special case of pretraining, where the entire future motion sequence to be predicted is masked as a reconstruction task. To address the loss of fine-grained features caused by the patch-based encoding strategy, we propose a Transformer model with masked skip-connections designed to complement our learning framework. Extensive evaluations on the Human3.6M and 3DPW datasets show that our approach surpasses state-of-the-art methods, achieving reductions in Mean Per Joint Position Error (MPJPE) of 4.3% and 4.1%, respectively.

Efficient data stream processing, particularly for persistence estimation, is crucial in handling high-velocity data streams characterized by skewed distributions of item frequencies. Unlike more straightforward frequency metrics, persistence captures items' recurrence across multiple time windows, requiring nuanced processing approaches. In response, we introduce the Hypersistent Sketch, an algorithm that significantly enhances persistence estimation through innovative filtering techniques. Our design incorporates a Cold Filter to address the skewed nature of data streams where a few high-frequency (hot) items dominate. This filter allows for differential treatment by using smaller counters for most low-frequency (cold) items, thus conservatively allocating memory resources that would otherwise be sized uniformly based on hot items. However, the Cold Filter can reduce throughput due to its segregative processing. To mitigate this, we implement a Burst Filter, which optimizes the processing of hot items. The Burst Filter significantly improves throughput by preventing repeated insertions within a single window—where persistence increases by at most one—and deferring the insertion until the window's end. Comparative evaluations demonstrate that the Hypersistent Sketch outperforms existing solutions like the On-Off Sketch, offering up to 3 times improved throughput while maintaining competitive accuracy and substantially reducing memory usage in handling large-scale data streams.

Continual adaptation to domain shifts at test time (CTTA) is crucial for enhancing the intelligence of deep learning enabled IoT applications. However, prevailing CTTA methods, which typically update all batch normalization (BN) layers, exhibit two memory inefficiencies. First, the reliance on BN layers for adaptation necessitates large batch sizes, leading to high memory usage. Second, updating all BN layers requires storing the activations of all BN layers for backpropagation, exacerbating the memory demand. Both factors lead to substantial memory costs, making existing solutions impractical for IoT devices. In this paper, we present FoCTTA, a low-memory CTTA strategy. The key is to automatically identify and adapt a few drift-sensitive representation layers, rather than blindly update all BN layers. The shift from BN to representation layers eliminates the need for large batch sizes. Also, by updating adaptation-critical layers only, FoCTTA avoids storing excessive activations. This focused adaptation approach ensures that FoCTTA is not only memory-efficient but also maintains effective adaptation. Evaluations show that FoCTTA improves the adaptation accuracy over the state-of-the-arts by 4.5%, 4.9%, and 14.8% on CIFAR10-C, CIFAR100-C, and ImageNet-C under the same memory constraints. Across various batch sizes, FoCTTA reduces the memory usage by 3-fold on average, while improving the accuracy by 8.1%, 3.6%, and 0.2%, respectively, on the three datasets.

With the increasing popularity of image and video analysis on mobile devices, high-throughput image inference has become essential. However, current mobile deep learning frameworks face key bottlenecks: high computational load in JPEG image recognition and low processor efficiency, which limit overall image processing throughput. To address these issues, this paper proposes the FCG framework (Frequency Domain model for CPU and GPU), a mobile JPEG inference framework based on frequency domain data and a hybrid parallel architecture that enables high-throughput inference for JPEG-encoded images on mobile devices. FCG decouples JPEG decoding from model inference by discarding the traditional RGB decoding process and retaining only the Huffman decoding. This decoding step is further accelerated through multi-core processing, significantly reducing the computational burden and latency during preprocessing. In light of the characteristics of frequency domain data and the heterogeneous CPU/GPU processors on mobile devices, FCG reconstructs the deep learning model to ensure recognition accuracy while optimizing resource utilization. By effectively allocating tasks and combining parallel and sequential execution, FCG optimizes processor resource utilization to achieve high throughput and low latency. FCG outperforms the state-of-the-art NN-Stretch by reducing latency by 36%. It also achieves significant throughput improvements—3.6x, 3.3x, and 2.8x—on CPU, GPU, and CPU+GPU configurations, respectively, compared to sequential inference systems. Additionally, FCG reduces power consumption by 56%, 35%, and 43% in these configurations.

Bluetooth audio, as a common application in our daily life, faces a major challenge due to its limited transmission range in meeting users' demands. Traditional solutions, such as using high-power Bluetooth transmitters, require hardware upgrades that are neither cost-effective nor energy-efficient. This work introduces WiLE-Audio, a novel approach that extends Low Energy Audio (LE Audio) coverage through physical-layer cross-technology communication (CTC) from WiFi to Bluetooth Low Energy (BLE). We first present a novel symbol mapping technique from WiFi DQPSK to BLE GFSK symbols, which enables all-channel and reliable CTC to support Bluetooth channel hopping. Then, to implement CTC to commodity WiFi Network Interface Card (NIC), we present a real-time reverse scrambling method that dynamically calculates the payload of WiFi packets at the NIC driver. Finally, to align with the strict time window requirements of the BLE receiver, we design a precise timing strategy and a priority scheduling mechanism at the WiFi transmitter, effectively mitigating timing offsets due to Carrier Sense Multiple Access with Collision Avoidance (CSMA/CA) and queue management. These systematic innovations allow WiLE-Audio to be easily implemented into existing commercial WiFi and BLE devices with only a simple software upgrade on the WiFi side. Furthermore, using existing WiFi infrastructures, WiLE-Audio enables the relaying of Bluetooth Low Energy Audio (LE-Audio) and the roaming of BLE receiver at any WiFi-covered location. We implement WiLE-Audio on commercial WiFi and BLE devices, and conduct extensive evaluations across various scenarios. Experimental results demonstrate that WiLE-Audio extends the transmission distance of LE Audio by 2× in single-hop mode and at least 2.6× in two-hop roaming mode. This work provides a cost-effective and scalable solution for enhancing Bluetooth audio coverage, providing a promising prospect for whole-house or even whole-building LE Audio listening.

Bluetooth audio faces a major challenge due to its limited transmission range in meeting users' demands. This work introduces WiLE-Audio, a novel approach that extends Low Energy Audio (LE Audio) coverage through cross-technology communication (CTC) from WiFi to BLE. We first present a novel symbol mapping technique to enable all-channel and reliable CTC. Then, we present a real-time reverse scrambling method to implement CTC to commodity WiFi devices. Finally, we design a precise timing strategy and a priority scheduling to align with the strict time window requirements of the BLE receiver. These systematic innovations allow WiLE-Audio to be easily implemented into existing commercial devices with only a simple software upgrade on the WiFi side. Furthermore, we achieve seamless switching between Bluetooth classic audio and WiLE-Audio, thus supporting whole-house audio roaming. We implement our work on commercial WiFi and BLE devices and demonstrate that WiLE-Audio extends the transmission distance of LE Audio more than 2×.

Object detection, a fundamental task in computer vision, is crucial for various intelligent edge computing applications. However, object detection algorithms are usually heavy in computation, hindering their deployments on resource-constrained edge devices. Traditional edge-cloud collaboration schemes, like deep neural network (DNN) partitioning across edge and cloud, are unfit for object detection due to the significant communication costs incurred by the large size of intermediate results. To this end, we propose a Difficult-Case based Small-Big model (DCSB) framework. It employs a difficult-case discriminator on the edge device to control data transfer between the small model on the edge and the large model in the cloud. We also adopt regional sampling to further reduce the bandwidth consumption and create a discriminator zoo to accommodate the varying networking conditions. Additionally, we extend DCSB to video tasks by developing an adaptive sampling rate update algorithm, aiming to minimize computational demands without sacrificing detection accuracy. Extensive experiments show that DCSB can detect 97.26%-97.96% objects while saving 74.37%-82.23% network bandwidth, compared to cloud-only methods. Furthermore, DCSB significantly outperforms the latest DNN partitioning methods, reducing inference time by 92.60%-95.10% given an 8Mbps transmission bandwidth. In video tasks, DCSB matches the detection accuracy of leading video analysis methods while cutting the computational overhead by 40%.

Satellite-terrestrial integrated networks based on edge computing can provide computation offloading service to terminal devices in remote areas. However, it faces various limitations, including satellite energy consumption, computation delay, and environmental dynamics, etc. In this paper, we propose a satellite-terrestrial integrated cloud and edge computing network (STCECN) architecture, including satellite layer, terrestrial layer and cloud center, where computing resources exist in multi-layer heterogeneous edge computing clusters. Optimization of system delay and energy consumption is defined as a mixed-integer programming problem. Moreover, we present a deep reinforcement learning-based computation offloading decision algorithm that can adapt to the dynamics and variability of satellite networks. A dual network computation offloading decision method is proposed for delay and energy consumption based on deep reinforcement learning offloading (DRLO), including deep convolutional network update method, quantization strategy, and bandwidth resource allocation. Meanwhile, the proposed method is based on previous experience and integrates deviation adjustment strategies for decision making to solve the problem of pseudo-patch loss caused by satellite network switching. The simulation results indicate that the proposed method performs almost consistently with traditional heuristic algorithms, with only 20% of the time consumption of the latter, and the number of pseudo packet loss also decreases to the original 10–20%.

Advancements in Earth observation sensors on low Earth orbit (LEO) satellites have significantly increased the volume of remote sensing images. This growth has led to challenges such as higher storage demands, downlink bandwidth stress, and transmission delays, particularly for real-time remote sensing image scene classification (RSISC). To address this, we propose a novel Satellite-Terrestrial Collaborative Scene Classification (STCSC) framework that integrates transmission and computation. The framework employs an attention-aware policy on the satellite, which adaptively determines the sequence of images and selection of image blocks for transmission, as well as these blocks' sampling rates. This policy is based on image complexity and the real-time data transmission rate, prioritizing blocks crucial for downstream tasks. On the ground, a classification model processes the received image blocks, balancing classification accuracy and transmission delay. Moreover, we have developed a comprehensive simulation system to validate the performance of our framework, including simulations of the satellite, transmission, and ground modules. Simulation results demonstrate that our STCSC framework can reduce transmission delay by 76.6% while enhancing classification accuracy on the ground by 0.6%. Additionally, our attention-aware policy is compatible with any ground classification model.

Accurately detecting objects and their interrelationships for Video Scene Graph Generation (VidSGG) confronts two primary challenges. The first involves the identification of active objects interacting with humans from the numerous background objects, while the second challenge is long-tailed distribution among predicate classes. To tackle these challenges, we propose STABILE, a novel framework with a spatial-temporal saliency-guided contrastive learning scheme. For the first challenge, STABILE features an active object retriever that includes an object saliency fusion block for enhancing object embeddings with motion cues alongside an object temporal encoder to capture temporal dependencies. For the second challenge, STABILE introduces an unbiased relationship representation learning module with an Unbiased Multi-Label (UML) contrastive loss to mitigate the effect of long-tailed distribution. With the enhancements in both aspects, STABILE substantially boosts the accuracy of scene graph generation. Extensive experiments demonstrate the superiority of STABILE, setting new benchmarks in the field by offering enhanced accuracy and unbiased scene graph generation.

The Vision Transformer (ViT) excels in accuracy when handling high-resolution images, yet it confronts the challenge of significant spatial redundancy, leading to increased computational and memory requirements. To address this, we present the Localization and Focus Vision Transformer (LF-ViT). This model operates by strategically curtailing computational demands without impinging on performance. In the Localization phase, a reduced-resolution image is processed; if a definitive prediction remains elusive, our pioneering Neighborhood Global Class Attention (NGCA) mechanism is triggered, effectively identifying and spotlighting class-discriminative regions based on initial findings. Subsequently, in the Focus phase, this designated region is used from the original image to enhance recognition. Uniquely, LF-ViT employs consistent parameters across both phases, ensuring seamless end-to-end optimization. Our empirical tests affirm LF-ViT's prowess: it remarkably decreases Deit-S's FLOPs by 63% and concurrently amplifies throughput twofold. Code of this project is at https://github.com/edgeai1/LF-ViT.git.

Multivariate time series forecasting plays an important role in various applications ranging from meteorology study, traffic management to economics planning. In the past decades, many efforts have been made toward accurate and reliable forecasting methods development under the assumption of intact input data. However, the time series data from real-world scenarios is often partially observed due to device malfunction or costly data acquisition, which can seriously impede the performance of the existing approaches. A naive employment of imputation methods unavoidably involves error accumulation and leads to suboptimal solutions. Motivated by this, we propose a Biased Temporal Convolution Graph Network that jointly captures the temporal dependencies and spatial structure. In particular, we inject bias into the two carefully developed modules, the Multi-Scale Instance PartialTCN and Biased GCN, to account for missing patterns. The experimental results show that our proposed model is able to achieve up to $9.93$\% improvements over the existing methods on five real-world benchmark datasets. Our code is available at: https://github.com/chenxiaodanhit/BiTGraph.

In-network caching is one of the most important characteristic of Named Data Networking (NDN). However, while replacing producers in responding to interest requests, caching data packets also shields consumers from perceiving the bottleneck bandwidth of the transmission path between the producer and the consumer. Therefore, when the content source switches from the cache node to the producer due to data exhaustion, the consumer can not adjust the requesting rate accordingly, and may lead to the serious bufferbloat or packet loss - we call it as Cache Trap. We found that Cache Trap occurs commonly in streaming services and the state-of-art NDN congestion control schemes cannot achieve efficient and stable quality of service when it happens. To escape Cache Trap, this paper proposes an explicit rate feedback congestion control algorithm, named as RFCC. RFCC leverages NDN routers' ability of encapsulating customized information in data packets to send link state information to consumers. Specifically, when responding to interest packets, RFCC nodes estimate data throughput received from the producer and insert this information into the returned data packets. The consumer perceives the change of content source according to the hopcount tag in data packet, and then adjusts the sending rate of interest packet based on the explicit rate information. We have implemented RFCC in both real-world NDN live video streaming and NDNsim simulation platforms, and compared it with the state-of-arts congestion control algorithms in a variety of scenarios. The experimental results show that when Cache Trap occurs, RFCC maintains a stable QoE in live video streaming, reduces 50% delay jitters compared with DPCCP and achieves$2.4 \times$throughput compared with PCON.

The rapid growth of deep learning-based services and applications underscores the need for efficient neural network model deployment. Traditional cloud-centric solutions, despite their computational power, face significant challenges such as high energy consumption, network transmission delays, and user privacy concerns. Conversely, performing high-performance inference on resource-constrained mobile devices, especially for tasks like advertising click prediction, presents its own set of difficulties. To address these challenges, we propose a device-cloud collaboration system utilizing a difficult-case discriminator. This system classifies input samples based on semantic information into difficult and simple cases. Difficult cases are processed in the cloud using a large model, while simple cases are handled on the device by a smaller model. This approach maximizes system resources and protects user privacy. Evaluations on public datasets show that our system significantly outperforms other advertisement methods in click prediction accuracy and uploading efficiency. Compared to the device-only approach, our system improves the area under the curve (AUC) by 8.9%, and compared to the cloud-centric approach, it reduces the upload ratio by 34%. Moreover, deploying our system on a specific smartphone demonstrates substantial improvements in private real datasets.

Over the past decade, smartphones have become indispensable personal mobile devices, experiencing a remarkable surge in software apps. These apps empower users to seamlessly connect with various internet services, such as social communication and online shopping. Accurately predicting smartphone app usage can effectively improve user experience and optimize resource utilization. However, existing models often treat app usage prediction as a classification problem, which suffers from issues of app usage imbalance and out-of-distribution (OOD) during deployment. To address these challenges, this paper proposes a novel click-through rate (CTR) ranking-based method for predicting app usage. By transforming the classification problem into a CTR problem, we can eliminate the negative impact of the app usage imbalance issue. To address the OOD issue during deployment, we generate the app click sequence and three types of discriminative features, which enable generalization on unseen apps. The app click sequence and the three types of features serve as inputs for training a CTR estimation model in the cloud, and the trained model is then deployed on the user's smartphone to predict the CTR for each installed app. The decision-making process involves ranking these CTR values and selecting the app with the highest CTR as the final prediction. Our method has been extensively tested with large-scale app usage data. The results demonstrate that our approach is able to outperform state-of-the-art methods, with improvements over 4.93% in top-3 accuracy and 6.64% in top-5 accuracy. It achieves approximately twice the accuracy in predicting apps with low usage frequencies in comparison to baseline methods. Our method has been successfully deployed on the app recommendation system of a leading smartphone manufacturer.

Multivariate time series forecasting is of central importance in modern intelligent decision systems. The dynamics of multivariate time series are jointly characterized by temporal dependencies and spatial correlations. Hence, it is equally important to build the forecasting models from both perspectives. The real-world multivariate time series data often presents spatial correlations that show structures and evolve dynamically. To capture such dynamic spatial structures, the existing forecasting approaches often rely on a two-stage learning process (learning dynamic series representations and then generating spatial structures), which is sensitive to the small time-window input data and has high variance. To address this, we propose a novel forecasting model with a structured matrix basis. At its core is a dynamic spatial structure generation function whose output space is well-constrained and the generated structures have lower variance, meanwhile, it is more expressive and can offer interpretable dynamics. This is achieved via a novel structured parameterization and imposing structure regularization on the matrix basis. The resulting forecasting model can achieve up to $8.5\%$ improvements over the existing methods on six benchmark datasets, and meanwhile, it enables us to gain insights into the dynamics of underlying systems.

Object detection is crucial in video analytics pipelines, but there is a need to optimize deep neural networks (DNNs)-based object detection for resource-constrained Internet of Things (IoT) devices devices. The computational constraints inherent to the IoT device inevitably curtail its precision and real-time efficacy in the domain of object detection, with pronounced challenges arising, particularly when confronted with high-resolution video streams. To overcome these limitations, we propose UPD (Using Physical Dynamics), a novel on-device system that enables real-time and accurate object detection for high-resolution video streams. UPD employs a lightweight tracking algorithm for the detection of the majority of video frames, concurrently executing the object detector in a parallel fashion only in select instances. UPD addresses tracking errors by eliminating inaccurate feature points and correcting tracking results using physical information about the object. Unlike previous approaches that depend solely on the high-latency object detector to offset errors, our method is unaffected by the video resolution level. Extensive experiments demonstrate that UPD facilitates real-time analysis of high-resolution videos on IoT devices and significantly improves the overall accuracy (mIoU) compared to state-of-the-art DBT (Detection-Based-Tracking) frameworks, achieving a 100% accuracy improvement on three commonly used datasets. A video demo can be found at https://youtu.be/gKRQPHJ6gmY.

Deploying compute-intensive deep neural networks (DNNs) on resource-constrained end devices has become a prominent trend, enabling localized intelligence. However, efficiently deploying these DNNs at scale poses challenges. To address this, extensive research has focused on the early exit architecture based on convolutional neural networks (CNNs), which dynamically adapt network depth to reduce inference computation. Nevertheless, the sequential execution of all internal classifiers (ICs) and subsequent termination based on an exit criterion is inefficient. Motivated by these insights, we introduce a resolution-adaptive prediction network (RAPNet) architecture. RAPNet comprises a lightweight prediction network that captures global image features and an inference network integrated with an early exit architecture. The prediction network accurately determines the optimal IC position conditioned on the input images for efficient image classification. Additionally, we incorporate resolution-adaptive inference and feature fusion mechanisms by computational reuse, to effectively mitigate image spatial redundancy and improve the accuracy of ICs. We conduct extensive experiments across various data sets and architectures to demonstrate that RAPNet achieves a significantly better accuracy versus computational tradeoff than other recently proposed early exit methods. For instance, when using MobileNet as the base network, RAPNet achieves significant accuracy improvements of 12% and 5.7% on the Tiny Imagenet and CIFAR-100 data sets, respectively, surpassing other early exit methods with similar computational constraints.

In order to improve performance and avoid overheating on mobile devices, precise thermal control with low overhead is crucial. To achieve this, we propose incorporating a performance monitoring counter (PMC)-based power model into thermal control, which enables a more accurate evaluation of the CPU’s power consumption. We demonstrate the plausibility of this approach using polynomial regression based on Moore’s Law. Additionally, we introduce a lightweight PMC sampling method that can collect multiple PMCs at once in the kernel space, reducing sampling overhead. By replacing the utilization-based model in the original the intelligent power allocation (IPA) with a PMC-based power model, we realize the PMC-based IPA governor can be ported to real mobile devices. After updating the thermal control governor in the Linux kernel, we perform tests on our PMC-based IPA using a mobile phone device. We compare it with Stepwise and IPA, which are commonly used in current mobile phone systems. We choose the CPU-intensive workbench, I/O-intensive workbench, and CPU and I/O-intensive hybrid workbench as workloads. The results show that PMC-based IPA effectively reduces energy consumption while improving performance. In particular, during the CPU and I/O-intensive hybrid experiment, where CPU-intensive and I/O-intensive tasks are executed alternately, PMC-based IPA reduces the running time by 10.0% and energy consumption by 16.6% compared to the original IPA. In order to verify the benefits of PMC-based IPA, mobile phone testing software AI Bench and Antutu are utilized. The results show that our scheme is able to control temperature more precisely than IPA and achieves a better score while consuming less energy, particularly during AI computing. These experiment results suggest that PMC-based IPA is valuable for practical use.

With human action anticipation becoming an essential tool for many practical applications, there has been an increasing trend in developing more accurate anticipation models in recent years. Most of the existing methods target standard action anticipation datasets, in which they could produce promising results by learning action-level contextual patterns. However, the over-simplified scenarios of standard datasets often do not hold in reality, which hinders them from being applied to real-world applications. To address this, we propose a scene-graph-based novel model SEAD that learns the action anticipation at the high semantic level rather than focusing on the action level. The proposed model is composed of two main modules, 1) the scene prediction module, which predicts future scene graphs using a grammar dictionary, and 2) the action anticipation module, which is responsible for predicting future actions with an LSTM network by taking as input the observed and predicted scene graphs. We evaluate our model on two real-world video datasets (Charades and Home Action Genome) as well as a standard action anticipation dataset (CAD-120) to verify its efficacy. The experimental results show that SEAD is able to outperform existing methods by large margins on the two real-world datasets and can also yield stable predictions on the standard dataset at the same time. In particular, our proposed model surpasses the state-of-the-art methods with mean average precision improvements consistently higher than 65% on the Charades dataset and an average improvement of 40.6% on the Home Action Genome dataset.

The increasing deployment of multiple deep neural networks (DNNs) on edge devices is revolutionizing mobile vision applications, spanning autonomous vehicles, augmented reality, and video surveillance. These applications demand adaptation to contextual and environmental drifts, typically through fine-tuning on edge devices without cloud access, due to increasing data privacy concerns and the urgency for timely responses. However, fine-tuning multiple DNNs on edge devices faces significant challenges due to the substantial computational workload. In this paper, we present PatchLine, a novel framework tailored for efficient on-device training in the form of fine-tuning for multi-DNN vision applications. At the core of PatchLine is an innovative lightweight adapter design called patches coupled with a strategic patch updating approach across models. Specifically, PatchLine adopts drift-adaptive incremental patching, correlation-aware warm patching, and entropy-based sample selection, to holistically reduce the number of trainable parameters, training epochs, and training samples. Experiments on four datasets, three vision tasks, four backbones, and two platforms demonstrate that PatchLine reduces the total computational cost by an average of 55% without sacrificing accuracy compared to the state-of-the-art.

Data uploading applications in wireless networks may suffer from the degrade of Quality of Experiences (QoEs), due to the untimely adjustment of congestion window (cwnd) in face of the rapid change of wireless channel. To mitigate this problem, we analyzed the relationship between the NIC driver queue length at the wireless sender and the end-to-end transmission performances, and found a strong correlation between them, since the bottleneck mostly occurs at the wireless link. Based on this observation, we designed QCC, a congestion control algorithm that adjusts cwnd according to the residual queue length after each round of NIC transmission. Since obtaining congestion information locally at the sender leads to a much shorter feedback path than waiting for the end-to-end ACK feedback, QCC can track the time-varying wireless links much faster and more accurately. In addition, QCC also presents adaptive slow start mechanism and MAC layer-assisted fast recovery mechanism, both of which make efficient use of residual queue length to further improve transmission performances. Experiment results on both real-world Wi-Fi and cellular networks reveal that QCC can achieve at least 2.36X lower delay than that of BBR while ensuring 98.5% throughput of BBR.

As one of the basic tasks of computer vision, object detection has been widely used in many intelligent applications. However, object detection algorithms are usually heavyweight in computation, hindering their implementations on resource-constrained edge devices. Current edge-cloud collaboration methods, such as CNN partition over edge-cloud devices, are not suitable for object detection since the large data size of the intermediate results will introduce extravagant communication costs. To address this challenge, we propose a difficult-case based small-big model (DCSB) framework that deploys a difficult-case discriminator on the edge device to control the data transfer between the small model (edge) and the big model (cloud). Upon receiving data, the edge device operates a difficult-case discriminator to classify images into easy cases and difficult cases according to the specific semantics of the images. The difficult cases will be uploaded to the cloud. To reduce bandwidth consumption, we propose a regional sampling method that adaptively down-samples some regions of the difficult case to reduce the amount of transferred data based on the primary results of the lightweight model. Experimental results on VOC, COCO, and HELMET datasets using two object detection algorithms demonstrate that DCSB can detect 93.77%-97.05% objects but save 77.19% -80.55% of network bandwidth compared with the cloud-only method, while the edge-only method can only detect 54.90%-68.28% objects in the same condition. In addition, compared with the state-of-the-art model partition method - CAS, DCSB saves 95.19%-95.80% of the inference time when the transmission bandwidth is 8Mbps.

Relative Camera Pose Estimation (RCPE) aims to calculate the translation and rotation between two frames with overlapped regions, which is crucial to computer vision and robotics. This paper presents a novel siamese convolutional transformer model, SiTPose, to regress relative camera pose directly. SiTPose is distinguished in three aspects: (1) With a cross-attention feature extractor and a compact transformer encoder, extreme rotation errors (> 150°) are significantly reduced: from 9.7‰ with the state-of-the-art 8-Points to 1‱ on the 7Scenes dataset. (2) SiTPose is also robust to narrow-baseline cases (slight rotation angle and large translation between neighboring frames), while existing RCPE methods mainly focus on wide-baseline cases. (3) SiTPose can be flexibly extended to geometry-based vSLAM (namely SiTSLAM) in a multi-threaded way to prevent tracking lost and scale ambiguity problems. Results on multiple datasets show that SiT-SLAM yields a marked improvement in robustness and localization accuracy in complex scenarios, e.g., RMSE error is reduced from 26.36m with the classic ORBSLAM3 method to 6.94m on the KITTI-09.

With the development of wireless communications (e.g., WiFi 6 and 5G), more and more high-bandwidth networks are emerging in our daily life. However, due to the limited speed of the slow start phase in congestion control algorithms, the high-bandwidth links may not be fully utilized, which will degrade the Quality of Service (QoS). The reason is, since the available link capacity is unknown until the link is fully occupied, the sender has to gradually increase the congestion window (cwnd) from a small initial value, causing the link to be underutilized, until a packet is dropped or a congestion signal is detected. Especially, for a short flow, the transmission is often finished before the link capacity is reached, leading to the waste of available bandwidth. To better exploit the high bandwidth links, this paper proposes FBE (Fast Bandwidth Estimation without Filling up the Bottleneck Link), by leveraging the effective ACK's returning rate to estimate the bottleneck link capacity. More specifically, instead of sending out any additional probe packets, FBE uses the ACK rates from the first two RTT rounds to quickly estimate the bandwidth during slow start phase. Since the original ACK rate is significantly lower than the available link bandwidth due to the exhaustion of send window, and the competing flows also have an impact on the ACK rate, FBE elaborates the ACK interval compensation algorithm to refine the ACK intervals to reflect the link rate, and then updates cwnd to a suitable size. To address the challenge of inaccurate bandwidth estimation, especially for rapidly changing wireless link, FBE dynamically adjusts cwnd according to the feedback of driver queue length after the bandwidth estimation. Experiments in real WiFi and LTE networks show that FBE reduces the slow start convergence time by 54.8% and 53.5% compared to CUBIC and BBR with traditional slow start, respectively. And when transferring short flows (512KB in size), FBE reduces the flow completion time by 40.4% and 43.8% compared to CUBIC and BBR, respectively.

Many intelligent applications based on deep neural networks (DNNs) are increasingly running on Internet of Things (IoT) devices. Unfortunately, the computing resources of these IoT devices are limited, which will seriously hinder the widespread deployment of various smart applications. A popular solution is to offload part of computation tasks from the IoT device to cloud by way of device–cloud collaboration. However, existing collaboration approaches may suffer from long network transmission delay or degraded accuracy due to the large amount of intermediate results, bring enormous challenges to the tasks, such as object detection, that require massive computing resources. In this article, we propose an efficient device–cloud collaborative inference (DCCI) object detection framework, which dynamically adjusts the amount of transferred data according to the content of input images. Specifically, a content-aware hard-case discriminator is proposed to automatically classify the input images as hard-cases or simple-cases, the hard-cases are uploaded to the cloud to be processed by a deployed heavyweight model, and the simple cases are processed by a lightweight model deployed to the IoT device, where the lightweight model is automatically compressed based on reinforcement learning according to the resource constraints of the IoT device. Furthermore, a collaborative scheduler based on the runtime load and network transmission capability of IoT devices is proposed to optimize the collaborative computation between IoT devices and the cloud. Extensive experimental evaluations show that compared to the Device-only approach, DCCI can reduce the memory footprint and compute resources of IoT devices by more than 90.0% and 30.87%, respectively. Compared to Cloud-centric, DCCI can save$2.0\times $of network bandwidth. In addition, compared with the state-of-the-art DNN partitioning method, DCCI can save$1.2\times $of inference latency, and$1.3\times $of IoT device energy consumption with the same accuracy constraint.

Recently, buoyed by advances in the space industry, low Earth orbit (LEO) satellites have become an important part of the Internet of Things (IoT). LEO satellites have entered the era of a big data link with IoT, how to deal with the data from the satellite IoT is a problem worthy of consideration. Conventional object detection method in optical remote sensing simply transmits the raw data to the ground. However, it ignores the properties of the images and the connection with the downstream task. To obtain efficient data transmission and accurate object detection, we propose a task-inspired satellite–terrestrial collaborative object detection framework called STCOD. It detects regions of interest (ROIs) and adopts a block-based adaptive sampling method to compress the background (BG) in optical remote sensing images by introducing satellite edge computing (SEC) on satellites. The STCOD framework also sets the transmission priority of image blocks according to their contributions to the task and uses fountain code to ensure the reliable transmission of important image blocks. We build a whole software simulation framework to validate our method, including the satellite module, the transmission module, and the terrestrial module. Extensive experimental results show that the STCOD framework can reduce the amount of downlink data decreased by 50.04% while losing the detection accuracy by 0.54%. In our simulated satellite–terrestrial link, the STCOD framework can reduce the number of satellite-to-terrestrial transmissions by half. When the packet loss rate is between 5% and 20%, the detection accuracy is lost only 0.05% to 0.5%.

Data upload applications such as streaming of live videos and cloud services bring convenience to our lives. However, the Quality of Experience in wireless networks is often unsatisfactory. One of the reasons is, wireless communication is vulnerable to unpredictable factors such as rapid change of channel and competition of channel resources, leading to hysteresis and inaccuracy when performing a congestion control algorithm. To mitigate this problem, we analyzed the relationship between the real-time length of the NIC driver queue at the sender and the end-to-end transmission performances, and found a strong correlation between them. The reason is, when the wireless link is the first hop of data upload, the bottleneck mostly occurs at this hop, thus causing the accumulation of packets on the NIC driver queue. Based on this observation, we designed QCC, a congestion control algorithm that adjusts the congestion window (cwnd) according to the residual queue length after each round of NIC transmission. Specifically, the cwnd will be quickly reduced when this queue length is large to mitigate congestion, and gradually increased when it is small to increase link utility. By this means, QCC can track the time-varying wireless links quickly and accurately to achieve both high throughput and low latency. We evaluate QCC on both real-world Wi-Fi and cellular network implementations. Our experiment results reveal that QCC can achieve 2.04X lower delays than that of BBR while ensuring the similar link utilization rate as BBR (99% of BBR's throughput).

With the development of embedded systems and Cross-Technology Communication (CTC) techniques, high throughput communication among heterogeneous IoT devices in the same frequency band (ISM band) can be achieved, which provides opportunities to en-hance the coexistence and cooperation for heterogeneous IoT de-vices. However, such improvement on the throughput is limited, since parallel communication has not been considered by most of the existing CTC techniques. There still exists unavoidable distortion in the reliability of the existing CTC techniques because of the heterogeneous properties of the protocols, hardware, and operating systems. Therefore, to enhance the communication throughput among heterogeneous IoT devices as much as possible, we study the parallel physical-layer CTC (PCTC) in this paper. We propose two advanced physical-layer CTCs. The first one improves the communication reliability between two heterogeneous IoT devices by retrieving the candidate emulation frames with high quality, and the other orthogonalizes the above candidate frames to achieve concurrent transmission. PCTC is designed for WiFi to ZigBee communication, and it is also implemented in USRP B210, which can improve the reliability and concurrency of the physical-layer CTC. Both theoretical analysis and experiment results verify its trans-mission reliability and improvement of throughput by comparing our technique with the existing ones.

Recent advances on Physical-layer Cross-Technology Communication (PHY-CTC) have achieved high throughput direct communication across different wireless technologies. These PHY-CTC works are commonly achieved by emulating the target signal waveform of the receiver. However, signal emulation suffers from inherent unreliability due to imperfect emulation, and it only supports few communication channels. When applied in WiFi to Bluetooth Low Energy (BLE) scenario, it will face two challenges: i) a BLE receiver can not tolerate any bit error in a frame, while emulation errors are easy to appear; and ii) the BLE device performs channel hopping while most BLE channels are unavailable for emulation based CTC.To address these challenges, we present WiBle, a high reliable and all-channel supporting PHY-CTC from WiFi to BLE. The key technical insight of WiBle is symbol transition mapping: When a symbol is transmitted by a WiFi sender and flows into a BLE receiver, it will leave some unique signatures which can be leveraged to extract information. More specifically, it is observed that the phase shifts of BLE received signal can be mapped to the transitions of WiFi symbols. Therefore, by carefully selecting the symbols at the WiFi sender, we can generate the desired phase shifts for correct BLE GFSK demodulation and achieve reliable CTC. Evaluation results on both USRP and commodity chip show that WiBle outperforms state-of-the-art CTCs by higher reliability (> 95% frame reception ratio), wider channel coverage (supporting all 40 BLE channels), and higher throughput (974.3Kbps), under a full range of configurations including indoor/outdoor and LoS/NLoS settings.

Recent research on physical layer cross technology communication (PHY-CTC) brings a timely answer for escalated wireless coexistence and open spectrum movement. PHY-CTC achieves direct communication among heterogeneous wireless technologies (e.g.,WiFi, Bluetooth, and ZigBee) in physical layer and thus brings communication support for coexistence service such as spectrum management and IoT device control. To put PHY-CTC into service, however, there still exists a gap due to its transmission failure and asymmetric link (i.e., one-way PHY-CTC) issues. In this paper, we propose NetCTC – the first networking support design for PHY-CTC to establish feedbacks (e.g., ACKs) and thus meet the upper layer networking requirements in heterogeneous unicast, multicast and broadcast. The core design of NetCTC is a real-time interaction mechanism which achieves reliable, transmission efficient and concurrent interactive communication among heterogeneous devices. We implement and evaluate NetCTC on commodity devices and the USRP-N210 platform. Our extensive evaluation demonstrates that NetCTC achieves reliable bidirectional cross technology communication under a full range of wireless configurations including stationary, mobile and duty-cycled settings.

Today's wireless networks have become increasingly heterogenous, mobile and dense. To satisfy the rising demands of ubiquitous connections, billions of multi-radio gateways have to be deployed, inevitably incurring high deployment cost and extra traffic overhead. Recent advances on Cross-Technology Communication (CTC) have shown its ability to avoid these drawbacks. However, the state-of-the-art CTCs from Bluetooth to WiFi, two of the most popular wireless techniques, still suffer from low data-rate (e.g., 3.1Kbps), which severely restricts their applicability. We present BlueFi, the first physical-layer CTC (PHY-CTC) from Bluetooth Low Energy (BLE) to WiFi, which enables high throughput, bidirectional and parallel transmissions between BLE and WiFi via spectral analysis. The key observation is that commodity WiFi chipsets can operate in the spectral analysis mode, in which WiFi can recognize specific BLE signal waveforms in frequency domain at symbol-level granularity. Leveraging this feature, we manufacture desired waveforms by choosing frame payload at BLE side, and observe spectral patterns at WiFi side. To achieve bidirectional links, we design a PHY-CTC method from WiFi to BLE based on signal emulation. We implement our prototype on USRP (with 802.11g PHY) and commodity BLE devices. Extensive evaluations show that BlueFi can achieve 120Kbps per link from BLE to WiFi with more than 95% frame reception ratio, over 38x faster than state-of-the-art CTCs. Moreover, BlueFi can support 9 wireless links in parallel, leading to the total throughput over 1Mbps.

Wireless Personal Area Network (WPAN) technologies (e.g., Bluetooth, ZigBee) have been widely used in our daily life. However, due to their short transmission distances, the delivery of urgent messages (e.g., emergency alarms) over long distance, suffers from long delay, since the messages have to be transported via multi-hop way. Recent studies show that adding low-power wide-area network (LPWAN) radios such as LoRa onto WPAN devices (e.g., Bluetooth) effectively overcomes the limitation. However, the introduce of heterogeneous communication inevitably incurs extra hardware cost, deployment inconvenience and traffic overhead from gateways. In this paper, we present BLE2LoRa, a novel Bluetooth Low Energy (BLE) to LoRaWAN cross-technology communication (CTC) approach, which leverages the frequency shifting ability of BLE device to emulate LoRa's chirp signal. Such emulation is feasible because a chirp is a signal whose frequency increases or decreases over time, while a BLE device can also construct specific signals with ladder-shaped frequencies which is similar to LoRa chirp by carefully selecting BLE payload bits. Therefore, without any hardware modification at both sender and receiver side, a LoRa device can demodulate BLE frames. Moreover, leveraging the high sensitivity of LoRa base station, a long distant CTC can be achieved. We bulid our BLE2LoRa prototype on USRP B210 (with LoRaWAN PHY) and commodity BLE chips (CC1200). Our evaluation reveals that BLE2LoRa can achieve 4.06kbps throughput from BLE to LoRa with more than 80% frame reception rate, leading to over 600 meters communication distance, which is over 20x range extension over native Bluetooth.

Cross-technology Communication (CTC) is a promising paradigm for efficient coordination and cooperation among heterogeneous wireless technologies. Recent advances in physical-layer CTC (PHY-CTC) approaches the standards' maximum transmission rate by exploring PHY-layer signal features. However, due to the lack of reliable feedback, current PHY-CTC technologies can hardly ensure transmission reliability. This paper presents RAP (Reliable Acknowledged PHY-CTC), a bidirectional CTC design with reliable PHY-CTC feedback. First, we present a novel PHY-CTC technique to efficiently establish a reliable feedback channel (e.g., ACKs or NACKs). Then, based on the feedback, we propose a joint intra-packet coding and inter-packet coding scheme to improve the reliability of CTC. Finally, we present an on-demand data (re)transmission scheme to support unicast, multicast and broadcast more efficiently. We implement and evaluate RAP on USRP N210 with IEEE 802.11g PHY (WiFi) and commodity ZigBee devices. The experiment results show RAP achieves reliable data transmission (>99% packet reception rate (PRR)) and high throughput (over 35kbps) under a wide range of scenarios.

Physical-Layer Cross-Technology Communication (PHY-CTC), which achieves direct communication among heterogeneous technologies, brings great opportunities to help diverse IoT devices achieve harmonious coexistence through explicit coordination. The core technique of PHY-CTC is signal emulation which utilizes the signal of one technology (e.g., WiFi) to emulate the signal of another technology (e.g., ZigBee). The signal emulation based approach, however, inevitably introduces emulation errors which further lead to unreliable communication. In this paper, we aim to recover the intrinsic emulation errors and establish reliable PHY-CTC. We propose TwinBee which (i) explores chip-level error patterns and (ii) corrects emulation errors with symbol-level chip-combining coding/decoding and soft mapping. To achieve this, TwinBee dose not require accessing chip information as well as making hardware changes. We implement TwinBee on commodity devices (i.e., Laptops with Atheros AR2425 WiFi card and TelosB motes) and the USRPN210 platform (for physical layer evaluation). Experiment results show that TwinBee significantly improves the Packet Reception Ratio (PRR) of PHY-CTC from 50%-60% to more than 99%. Furthermore, we demonstrate the reliability of TwinBee in a data dissemination application over a network of 20 TelosB nodes, achieving over 42× reduction of data dissemination delay compared to the state-of-the-art.

Low-Power Wide-Area Network (LPWAN) is an emerging platform for Internet-of-Thing (IoT) devices to access the base station far away. However, two of the most popular IoT techniques, Bluetooth and ZigBee, can not be connected to LPWAN directly due to their very short communication distance (e.g., 30 meters). Our work, named as Symphony, implements an universal LPWAN on existing heterogeneous wireless devices by overcoming two challenges. First, Symphony achieves a long-range communication from both Bluetooth Low Energy (BLE) and ZigBee to LoRaWAN, enabling these ubiquitously deployed low-power devices to access a base station from faraway. It is achieved by exploiting Narrow-Band Communication, where the BLE/ZigBee devices generate ultra narrow-band signals (i.e., single-tone sinusoidal signals) through payload manipulation, while the LoRaWAN base station detects these signals via its demodulator, which has a high receiver sensitivity for long range communication. Second, Symphony enables concurrent transmissions from heterogeneous radios (i.e., BLE, ZigBee and LoRa) at a LoRaWAN base station. This is achieved by Cross-Technology Parallel Decoding, which is able to disentangle and decode the interfering transmissions. Our evaluations on USRP and commodity devices reveal that Symphony achieves a concurrent wireless communication from BLE, ZigBee and LoRa commercial chips to a LoRaWAN base station over 500 meters, $16 \times$ range extension over native BLE/ZigBee.

Recent advances on physical-layer Cross-Technology Communication (PHY-CTC) achieve high throughput direct communication across different wireless technologies, by emulating the standard waveform of the receiver. However, this signal emulation method faces the challenges of inherent unreliability due to the imperfect emulation. Therefore, it's not suitable to achieve PHY-CTC from WiFi to BLE, since a BLE receiver can not tolerate any bit error in preamble checking when receiving a BLE frame. We present NBee, the first WiFi to BLE physical-level CTC. The key insight lies in Narrow-Band Decoding, i.e., 22MHz bandwidth WiFi (802.11b) signal can be correctly decoded at the BLE RF front-end with only 1MHz bandwidth, if the WiFi payload bits are selected by a specific pattern. More specifically, NBee leverages the unique signatures in the WiFi signal distorted by 1MHz Low Pass Filter (LPF) at BLE to extract information. Evaluation results on commodity BLE chips show NBee can achieve 1Mbps CTC with 95% packet reception rate (PRR), 3400x faster than the state-of-art CTC from WiFi to BLE.

The cross-technology communication (CTC) is a promising technique proposed recently to bridge heterogeneous wireless technologies in the ISM bands. Existing solutions use only the coarse-grained packet-level information for CTC modulation, suffering from a low throughput (e.g., 10 b/s). Our approach, called BlueBee, explores the dense PHY-layer information for CTC by emulating legitimate ZigBee frames with the Bluetooth radio. Uniquely, BlueBee achieves dual-standard compliance and transparency for its only modifying the payload of Bluetooth frames, requiring neither hardware nor firmware changes at either the Bluetooth sender or the ZigBee receiver. Our implementation on both USRP and commodity devices shows that BlueBee can achieve standard ZigBee bit rate of 250 kb/s at more than 99% accuracy, which is over 10000 x faster than the state-of-the-art packet-level CTC technologies.

Recent research on physical layer cross technology communication (PHY-CTC) brings a timely answer for escalated wireless coexistence and open spectrum movement. PHY-CTC achieves direct communication among heterogeneous wireless technologies (e.g.,WiFi, Bluetooth, and ZigBee) in physical layer and thus brings communication support for coexistence service such as spectrum management and IoT device control. To put PHY-CTC into service, however, there still exists a gap due to its transmission failure and asymmetric link (i.e., one-way PHYCTC) issues. In this paper, we propose NetCTC - the first networking support design for PHY-CTC to establish feedbacks (e.g., ACKs) and thus meet the upper layer networking requirements in heterogeneous unicast, multicast and broadcast. The core design of NetCTC is a real-time interaction mechanism which achieves reliable, transmission efficient and parallel interactive communication among heterogeneous devices. We implement and evaluate NetCTC on commodity devices (Laptops with Atheros AR2425 WiFi NIC, smart phones with Broadcom BCM4330 WiFi chip and MicaZ CC2420) and the USRP-N210 platform. Our extensive evaluation demonstrates that NetCTC achieves reliable bidirectional cross technology communication under a full range of wireless configurations including stationary, mobile and dutycycled settings.

Cross-Technology Communication (CTC) is an enabling technology for efficient coexistence and effective cooperation among heterogeneous wireless devices by exchanging data frames directly without gateways. Recent advances in the physical-layer CTC achieve thousands of times faster speed than that of previous packet-level CTC techniques. However, physical-layer CTC still faces the challenge of inherent unreliability due to the imperfection of physical-layer signal emulation. Our work, named TwinBee, aims to recover the intrinsic errors of physical-layer CTC, by exploring chip-level error patterns. This is achieved interestingly without even observing the chip information and without any hardware modification. System evaluation shows that our key idea, namely symbol-level chip-combining decoding with soft mapping, significantly improves the Packet Reception Ratio (PRR) of the physical-layer CTC from 50%-60% to more than 99%. We also demonstrate the reliability of TwinBee in a data dissemination application over a network of 20 TelosB nodes, achieving over 40x reduction of data dissemination delay compared to Deluge.

Cross-Technology Communication (CTC) supports direct message exchange among heterogeneous wireless technologies (e.g., Wi-Fi, ZigBee, and BlueTooth) under the same ISM band, enabling explicit cross-technology control and coordination. For instance, a Wi-Fi AP can directly control ZigBee-enabled smart light bulbs without an expensive dual-radio gateway. Such CTC capability can be further amplified if we can extend the communication range of CTC to support long-range wide-area IoT applications such as environmental monitoring, smart metering, and precision agriculture. Our work, named LongBee, is the first to extend the range of Cross-Technology Communication. At the transmitter side, LongBee concentrates the effective TX power through down-clocked operations, and at the receiver side, LongBee improves the RX sensitivity with an innovative transition coding to ensure reliable preamble detection and payload reception. All these are achieved without modifying hardware and without introducing extra Wi-Fi RF energy cost. We implemented the LongBee on the USRP platform and commodity ZigBee devices. Our comprehensive evaluation reveals that LongBee with concentrated TX power and higher RX sensitivity achieves reliably over 10x range extension over native ZigBee communication and 2x range extension than the longest distance achieved by existing CTC schemes so far.

Cross-technology Communication (CTC) is a key technique to explore the full capacity of heterogeneous wireless. The latest CTC designs explore the PHY-layer to reach the standards' maximum rate, but leaving a critical gap to practicality -- existing PHY-layer CTCs are commonly transmitter-side techniques requiring a high-end transmitter (with a high degree of freedom in signal manipulation) to emulate the receiver signal closely. This inherently limits the reverse direction (low-end to high-end) communication. We present XBee, a unique receiver-side CTC that fills in the gap and makes a critical step towards achieving CTC bidirectionality. XBee is demonstrated as a ZigBee to BLE communication, where the key innovation lies in the unique mechanism of cross-technology decoding, or cross-decoding in short, which interprets a ZigBee frame only by carefully observing the bit patterns obtained at the BLE receiver. Technically, XBee counterintuitively explores the sampling offset to overcome the intrinsic challenge due to BLE's narrower bandwidth (1MHz) than ZigBee (2MHz). Extensive implementation and evaluation on USRP and commodity devices reach 250 kbps under 85% reliability, a 15,000x improvement over state-of-the-art ZigBee to BLE communication, and comparable with the latest PHY-layer CTCs to achieve CTC bidirectionality.

Under significant coexistence in the ISM band, the impact of cross-technology interference (CTI) has become a major threat to low-power IoT. This paper presents ECC that uniquely enables explicit channel coordination among heterogeneities via cross-technology communication (CTC) introduced in the latest studies, while maintaining full compatibility to commodity devices. Unlike any implicit coordination designs adopting statistical models to probabilistically predict white spaces, ECC generates the white space using WiFi CTS, which is then explicitly notified to ZigBee through CTC for immediate use. Technical highlight of ECC lies in ensuring ZigBee communication under CTI, without disrupting WiFi operation. This is effectively achieved by the dynamic adjustment of CTS duration with respect to traffic amount and spectrum availability, which essentially enables ECC to be generally applied to various scenarios without prior knowledge. Lastly, ECC significantly reduces delay and energy in low duty cycled ZigBee, by waking them up upon channel availability (via CTC). We evaluate ECC on commercial platforms: Atheros AR2425 WiFi card and TelosB motes. Experiment results show that ECC achieves 1.8x ZigBee packet reception ratio, and cuts down delay and energy by 98.6% and 51% under the low duty cycle.

With the wide application of location-based social networks (LBSNs), point-of-interest (POI) recommendation has become one of the major services in LBSNs. The behaviors of users in LBSNs are mainly checking in POIs, and these checking-in behaviors are influenced by user's behavior habits and his/her friends. In social networks, social influence is often used to help businesses to attract more users. Each target user has a different influence on different POI in social networks. This paper selects the list of POIs with the greatest influence for recommending users. Our goals are to satisfy the target user's service need, and simultaneously to promote businesses' locations (POIs). This paper defines a POI recommendation problem for location promotion. Additionally, we use submodular properties to solve the optimization problem. At last, this paper conducted a comprehensive performance evaluation for our method using two real LBSN datasets. Experimental results show that our proposed method achieves significantly superior POI recommendations comparing with other state-of-the-art recommendation approaches in terms of location promotion.

Cross-Technology Communication is an emerging research direction providing a promising solution to the coexistence problem of heterogeneous wireless technologies in the ISM bands. However, existing works use only the coarse-grained packet-level information for cross-technology modulation, suffering from a low throughput (e.g., 10bps). Our approach, called BlueBee, aims at achieving much higher CTC throughput thus extends CTC applications. We pro- poses a new direction by emulating legitimate ZigBee frames using a Bluetooth Low Energy (BLE) radio. Uniquely, BlueBee achieves dual-standard compliance (i.e., BLE and ZigBee) and transparency by selecting only the payload of Bluetooth frames, requiring neither hardware nor firmware changes at the BLE senders and ZigBee receivers. Our implementation on commodity device testbeds shows that BlueBee can achieve a more than 99% accuracy and a through- put 10,000x faster than the state-of-the-art CTC reported so far. In addition, we show a demo of using BlueBee on a smartphone to control several smart light bulbs a ached with ZigBee radio.

Recent advances in Cross-Technology Communication (CTC) have improved efficient coexistence and cooperation among heterogeneous wireless devices (e.g., WiFi, ZigBee, and Bluetooth) operating in the same ISM band. However, until now the effectiveness of existing CTCs, which rely on packet-level modulation, is limited due to their low throughput (e.g., tens of bps). Our work, named WEBee, opens a promising direction for high-throughput CTC via physical-level emulation. WEBee uses a high-speed wireless radio (e.g., WiFi OFDM) to emulate the desired signals of a low-speed radio (e.g., ZigBee). Our unique emulation technique manipulates only the payload of WiFi packets, requiring neither hardware nor firmware changes in commodity technologies -- a feature allowing zero-cost fast deployment on existing WiFi infrastructure. We designed and implemented WEBee with commodity devices (Atheros AR2425 WiFi card and MicaZ CC2420) and the USRP-N210 platform (for PHY layer evaluation). Our comprehensive evaluation reveals that WEBee can achieve a more than 99% reliable parallel CTC between WiFi and ZigBee with 126 Kbps in noisy environments, a throughput about 16,000x faster than current state-of-the-art CTCs.

The applicability of existing Cross-Technology Communication (CTC) methods, which rely on packet-level modulation, is severely limited due to their very low throughput, e.g., tens of bps. Our work, named as WEBee, opens a promising direction for high throughput CTC via physical-level emulation. Specifically, WEBee synthesizes the time-domain signals by choosing appropriate frequency-domain components fed into the subcarriers of WiFi OFDM. WE-Bee can emulate the desired physical-layer ZigBee signals by manipulating only the data bits in WiFi packet payload, requiring neither hardware nor firmware changes in commodity technologies. Moreover, WEBee enables the parallel CTC, where one WiFi frame emulates two ZigBee frames simultaneously. To evaluate the performance, we implemented WEBee on commodity devices (the Atheros AR2425 WiFi card, BCM 4330 WiFi card and CC2420, CC2530 ZigBee devices). Our comprehensive evaluation reveals that WEBee can achieve the CTC between WiFi and ZigBee with a reliable throughput of 126Kbps in noisy environment, 16,000x faster than current state-of-the-art CTC methods.

Cross-Technology Communication is an emerging research direction providing a promising solution to the wireless coexistence problem in the ISM bands. However, the state-of-the-art CTC designs have intrinsic limitations in the throughput due to their use of coarse-grained packet-level information. In contrast, we propose to exploit the fine-grained signal modulation information via a technique called PHY-layer emulation to boost CTC throughput. We can embed a legitimate packet of a target technology, e.g., ZigBee, within the payload of a source technology, e.g., WiFi or Bluetooth Low Energy (BLE). At the mean time, we require no modification at the hardware or firmware at either sender or receiver. We can achieve 8,000x throughput from WiFi to ZigBee and 10,000x throughput from BLE to ZigBee compared to the state of the art. We also have a demo showcasing how our designs can be implemented on off-the-shelf smartphones for smart light bulbs control.

Cross-Technology Communication is a promising solution proposed recently to the coexistence problem of heterogeneous wireless technologies in the ISM bands. The existing works use only the coarse-grained packet-level information for cross-technology modulation, suffering from a low throughput (e.g., 10bps). Our approach, called BlueBee, proposes a new direction by emulating legitimate ZigBee frames using a Bluetooth radio. Uniquely, BlueBee achieves dual-standard compliance and transparency by selecting only the payload of Bluetooth frames, requiring neither hardware nor firmware changes at the Bluetooth senders and ZigBee receivers. Our implementation on both USRP and commodity devices shows that BlueBee can achieve a more than 99% accuracy and a throughput 10,000x faster than the state-of-the-art CTC reported so far.

In this work, we focus on minimizing the overall network energy consumption problem under delay-constrained: N different packets from N time varying channels must be transmitted by a hard deadline of T slots. Each transmitter determines how much power to transmit with, during each time slot based on the current channel quality and the number of un transmitted bits, with the objective of minimizing overall network energy consumption. We transform the non-convex optimization problem into a geometric programming problem which has convex form and propose a distributed approximate optimization algorithm. Moreover, a lazy updating distributed algorithm is also presented for infrequent message passing. Experimental results show that the proposed distributed algorithm converges fast and the results of the distributed algorithm and results of centralized algorithm are very close.

We present the design, implementation, and evaluation of AirCloud -- a novel client-cloud system for pervasive and personal air-quality monitoring at low cost. At the frontend, we create two types of Internet-connected particulate matter (PM2:5) monitors -- AQM and miniAQM, with carefully designed mechanical structures for optimal air-flow. On the cloud-side, we create an air-quality analytics engine that learn and create models of air-quality based on a fusion of sensor data. This engine is used to calibrate AQMs and mini-AQMs in real-time, and infer PM2:5 concentrations. We evaluate AirCloud using 5 months of data and 2 month of continuous deployment, and show that AirCloud is able to achieve good accuracies at much lower cost than previous solutions. We also show three real applications built on top of AirCloud by 3rd party developers to further demonstrate the value of our system.

Traffic routing guidance algorithms which only consider user satisfaction will result in road density unbalance. On the contrary, the route algorithms will not meet user request if they merely focus on the traffic load balance. So it is important to take user satisfaction and traffic load into consideration at the same time. Although a few works focus on this combination, they ignore the difference between the user requests. We propose a traffic dispersion routing algorithm on VANET naming BPR-US which is based on backpressure theory. It tries to satisfy the user demand and also separates the traffic flow. At last, we make simulation experiments to compare BPR-US algorithm and other route guidance algorithms, proving that BPR-US algorithm can have a better effect on both user satisfaction and traffic load.