Running multiple deep neural networks (DNNs) simultaneously on mobile devices introduces challenges due to constrained computing resources. Previous research has explored the use of heterogeneous processors for accelerating DNN inference but often overlooks thermal issues, which can degrade computing power. In this article, we propose Phoenix, a system specifically designed to enhance the performance of multi-instance DNNs in video applications by maximizing accuracy and ensuring the achievement of a required frame rate. Phoenix allocates DNN tasks to the most suitable hardware processors, understanding complex thermal dynamics through reinforcement learning, and postpones the onset of thermal throttling. Despite optimized task allocation, continuous inference of multiple DNNs can still lead to thermal throttling. To manage performance degradation, Phoenix employs a multi-exit network, adaptively executing inference tasks to ensure consistent frame rates. Phoenix minimizes accuracy loss from early exits by optimally generating and operating multi-exit networks. We evaluated Phoenix using two different benchmarks and Virtual Youtuber streaming application. The results demonstrated that Phoenix effectively enhances device performance by delaying thermal throttling and achieving optimal accuracy while maintaining a consistent frame rate.

Although on-device video super-resolution enables high-quality live 360-degree streaming on mobile devices, existing methods often waste energy by overlooking perceived visual quality. In this paper, we present EOS, an energy-efficient on-device super-resolution system for mobile omnidirectional video (ODV) live streaming. EOS reduces energy waste by dynamically adjusting super-resolution complexity based on the predicted visual quality of super-resolved frames. This approach raises two challenges: (1) designing an adaptive inference policy that maximizes energy savings while minimizing degradation in Quality-of-Experience (QoE), and (2) developing a method to predict visual quality under the constraints of mobile ODV live streaming. To tackle these challenges, EOS introduces EOS SR and a No-Reference Up-scaling Quality Prediction scheme. EOS SR employs a device-agnostic, scalable deep neural network optimized for mobile devices, with an energy-aware scheduler that jointly selects the optimal super-resolution model and GPU frequency. The No-Reference Upscaling Quality Prediction scheme estimates visual quality across arbitrary viewpoints in real time without requiring high-resolution reference videos. Experiments on commodity smartphones show that EOS reduces average power consumption by 34.6%–49.9% compared to baseline methods, while preserving high visual quality and frame rates.

For highly immersive mobile volumetric video streaming, it is essential to deliver photo-realistic full-scene content with smooth playback. Unlike traditional representations such as point clouds, 3D Gaussian Splatting (3DGS) has gained attention for its ability to represent high-quality full-scene 3D content. However, our preliminary experiments show that existing methods for 3DGS-based videos fail to achieve smooth playback on mobile devices. In this paper, we propose Vega, a 3DGS-based photo-realistic full-scene volumetric video streaming system that ensures real-time playback on mobile devices. The core idea behind Vega's real-time rendering is object-level selective computation, which allocates computational resources to visually important objects to meet strict rendering deadlines. To enable mobile streaming based on the selective computation, Vega addresses two challenges: (1) designing an encoding scheme that optimizes the data size of videos while being compatible with object-level prioritization, and (2) developing a rendering pipeline that efficiently operates on resource-constrained mobile devices. We implemented an end-to-end Vega system, consisting of a streaming server and an Android application. Experimental results on commodity smartphones show that Vega achieves 30 frames per second (FPS) for full-scene volumetric video streaming while maintaining competitive data size and visual quality compared to existing baselines.

Deep neural networks (DNNs) have enabled a wide range of artificial intelligence of things (AIoT) applications, but their increasing complexity poses challenges for deployment on resource-constrained devices. Model compression techniques such as pruning and quantization have been widely adopted to address these challenges; however, they inevitably incur accuracy loss due to information loss. Recently, class-aware pruning has emerged as a promising approach, but existing methods often lack flexibility, as they are typically tailored to fixed target class sets and fail to generalize well to dynamic or broad class distributions. To address this limitation, we propose Mixture of Class-aware Experts (MoCE), a novel framework that combines class-aware pruning with a Mixture of Experts (MoE) architecture. MoCE constructs multiple lightweight experts using class-aware pruning, each specialized for a subset of classes, and employs a shared encoder and a lightweight router to dynamically select the appropriate expert at runtime. Our preliminary results demonstrate the potential of combining class-aware pruning and expert selection to enable accurate and efficient inference on resource-limited AIoT devices.

Mobile Augmented Reality (AR) applications demand high-quality, real-time visual prediction, including pixel-level depth and semantics, to enable immersive and context-aware user experiences. Recently, Vision Foundation Models (VFMs) offer strong generalization capabilities on diverse and unseen data, supporting scalable mobile AR experiences. However, deploying VFMs on mobile devices is challenging due to computational limitations, particularly in maintaining both prediction accuracy and real-time performance. In this paper, we present ARIA, the first system that enables on-device inference acceleration of a VFM. ARIA employs the heterogeneity of mobile processors through a parallel and selective inference scheme: full-frame prediction is periodically offloaded to a processor with high parallelism capability like GPU, while low-latency updates on dynamic regions are conducted via a specialized accelerator like NPU. Implemented and evaluated using mobile devices, ARIA achieved significant improvements in accuracy and deadline success rate on diverse real-world mobile AR scenarios.

Existing Android systems exhibit energy inefficiency during mobile web browsing due to the lack of awareness of application-level context. Inferring such context from system-level data alone is challenging, but one promising opportunity is using the sequence of task wakeup events, where one task activates another. These sequences show correlation with the type of webpage being used. In this article, we present Ember, a lightweight and responsive power management system for mobile web browsing using only task wakeup sequences. Ember introduces a neural network–based approach to predict optimal CPU clamping values by addressing three key challenges: (1) embedding task names, given as natural-language strings, into meaningful vectors using a Word2Vec-based embedding scheme tailored for task wakeup sequences; (2) minimizing inference overhead with a touch-driven hierarchical inference method that combines lightweight logistic regression with high-accuracy neural networks to balance responsiveness and efficiency; and (3) adapting to within-page interaction dynamics through an interaction-adaptive clamping mechanism that adjusts constraints across different user interaction phases. Implemented on commercial Android smartphones, Ember reduced power consumption by 6.2%–31.2% across a wide range of webpages while maintaining user-perceived quality of experience (QoE).

Reckless usage of electric (e-) scooters causes many injury accidents, raising critical safety concerns. Despite newly introduced regulations, specifically, speed limits and sidewalk driving prohibitions, the number of accidents increases due to the challenges in enforcement. Therefore, a reliable method to detect safety-threatening illegal behaviors of e-scooters is essential to mitigate this growing problem. In this article, we propose SecureRide, a system that accurately detects illegal e-scooter behaviors, i.e., speeding violation and sidewalk riding, at runtime using only battery information, without the need for additional sensors. To this end, we first design a neural network-based illegal behavior predictor that takes sequences of three battery factors, i.e., voltage, current, and capacity, as inputs. The model architecture is optimized based on time constraints, target accuracy, and resource constraints of the target devices. Next, we devise a runtime detection strategy to achieve both high accuracy and low detection time. SecureRide operates in two modes with different predictors– lightweight-quick and complex-accurate models–depending on the driving situation, ensuring both high accuracy and low detection time. We extensively validate SecureRide based on actual driving experiments. Our results show that SecureRide detects illegal behaviors with an accuracy of up to 99.77% within 1.01 seconds.

Sound event detection (SED) based on on-device machine learning (ML) presents considerable energy challenges for ultra-low-power sensor devices. In this paper, we propose DASH, a duration-aware SED system designed for energy-constrained sensor devices in domestic environments. As repeated inferences for continuous sound events lead to unnecessary energy consumption, DASH aims to minimize unnecessary inferences by predicting the duration of sound events. However, the variability of sound event durations across different environments and scenarios poses a major challenge in developing a responsive yet energy-efficient duration-aware SED system. To address this, DASH introduces three key solutions: (1) N-probability distribution-based event duration prediction, which identifies checkpoints where new inferences are likely needed; (2) Affinity-guided event classification, which performs low-energy affinity matching at checkpoints to determine whether ML inference is necessary; and (3) Interrupt blocking-enabling cycle-based device state control, which periodically checks for event presence with minimal energy consumption at non-checkpoint times. We implemented DASH on MSP430-based sensor devices deployed in real home environments. Experimental results demonstrate that DASH reduced energy consumption by approximately 97–98% compared to evaluation baselines, with only a 4.7% error rate.

We propose a cross-device web solution, called Vulture, which distributes graphical user interface (GUI) elements of apps across multiple devices without requiring modifications of web apps or browsers. Several challenges should be resolved to achieve the goals. First, the peer–server configuration should be efficiently established to distribute web resources in cross-device web environments. Vulture exploits an in-browser virtual proxy that runs the web server’s functionality in web browsers using a virtual HTTP scheme and a relevant API. Second, the functional consistency of web apps must be ensured in GUI-distributed environments. Vulture solves this challenge by providing a single-browser illusion with a two-tier document object models (DOM) architecture, which handles view state changes and user input seamlessly in cross-device environments. We implemented Vulture and extensively evaluated the system under various combinations of operating platforms, devices, and network capabilities while running 50 real web apps. The experiment results show that the proposed scheme provides functionally consistent cross-device web experiences by allowing fine-grained GUI distribution. We also confirmed that the in-browser virtual proxy reduces the GUI distribution time and the view change reproduction time by averages of 38.47% and 20.46%, respectively.

Sound event detection (SED) using lightweight sensor device has recently gained attention as a practical means to capture context and activities especially in domestic environments. However, SED applications running on sensor device are severely constrained by device’s energy capacity. One solution is to offload a portion of inference to server for reducing runtime complexity, i.e., energy consumption, of sensor device. Offloading should consider the trade-off between computation and data transmission costs adequately; more computation on sensor device reduces data to be transmitted and vice versa. To address this challenge, we propose SEDAC (Sound Event Detection with Attention-based audio Compression), a novel technique for split learning in SED that compresses data from sensor device to offload less data. SEDAC compresses the input of SED models, or Mel spectrograms, with minimal computation in sensor device. Rather than directly compressing the input, SEDAC achieves data compression by selectively capturing the key parts of sound events using an attention mechanism. The scheme also modifies an existing loss function and employs knowledge distillation to mitigate potential loss of SED accuracy due to data compression. Our evaluation shows that SEDAC outperforms the state-of-the-art data compressive split learning schemes, up to about 30%. Furthermore, our real-world deployment demonstrates that sensor devices with SEDAC successfully operate with minimal energy and memory overhead.

3D object detection with omnidirectional views enables safety-critical applications such as mobile robot navigation. Such applications increasingly operate on resource-constrained edge devices, facilitating reliable processing without privacy concerns or network delays. To enable cost-effective deployment, cameras have been widely adopted as a low-cost alternative to LiDAR sensors. However, the compute-intensive workload to achieve high performance of camera-based solutions remains challenging due to the computational limitations of edge devices. In this paper, we present Panopticus, a carefully designed system for omnidirectional and camera-based 3D detection on edge devices. Panopticus employs an adaptive multi-branch detection scheme that accounts for spatial complexities. To optimize the accuracy within latency limits, Panopticus dynamically adjusts the model's architecture and operations based on available edge resources and spatial characteristics. We implemented Panopticus on three edge devices and conducted experiments across real-world environments based on the public self-driving dataset and our mobile 360° camera dataset. Experiment results showed that Panopticus improves accuracy by 62% on average given the strict latency objective of 33ms. Also, Panopticus achieves a 2.1× latency reduction on average compared to baselines.

In e-scooter sharing systems, users randomly select and use e-scooters based on inaccurate battery information. This simple rental policy leads to low profitability on two fronts. First, inaccurate battery information causes unexpected device shutdowns, causing negative user experiences and refunds. Second, randomly selected e-scooters increase operation costs for battery management. In this paper, we propose e-scooter recommendation system, EcoRide, which provides accurate battery estimation and profitable e-scooter selection to maximize profitability of sharing systems. To this end, we propose a battery estimation considering four factors, i.e., battery state, temperature, user weight, and road slope, that affect the available battery energy in e-scooter applications. We define a parameter, dynamic voltage threshold (DVT), to represent dynamically changing battery energy, and use it to estimate battery availability. Next, to achieve cost-effective e-scooter selection, we introduce a multi-agent reinforcement learning (MARL)-based technique to learn policies that minimize operation costs. We define sharing system operation as a MARL problem with an objective function based on battery management costs. To cope with unstable training due to a wide service area and multiple requests, a centralized training technique is adopted. The proposed battery estimation and e-scooter selection technique are validated through actual driving tests and a sharing system simulator, respectively. Additionally, our case study using open data from Washington D.C. demonstrates a profit gain of up to 68% with EcoRide.

The capability of energy harvesting application powered by indoor photovoltaic energy is severely affected by dynamic light environments. Accordingly, accurate understanding of the target environment and deploying energy harvesting sensors is practically very hard. In this article, we propose HarvAR, which manages photovoltaic energy harvesting sensors with mobile augmented reality (AR)-empowered techniques. HarvAR utilizes the error-prone RGBD data of mobile device to construct a digital twin (DT), performing depth error compensation and estimating the optical properties of the target space. Using the DT, the proposed system predicts the harvesting capability with low overhead, and recommends adequate locations for installing or relocating harvesting sensors. We implemented the HarvAR system and evaluated its accuracy and efficiency in three indoor environments. Our experiments show that DT configuration and harvesting prediction can be performed in minutes, compared to over 10 h using existing techniques, and harvesting prediction is provided with less than 20% error.

Optimizing deep neural networks (DNNs) running on resource-constrained devices, such as energy harvesting sensor devices, poses unique challenges due to the limited memory and varying energy conditions. Existing efforts have shown that deploying a multi-exit network mitigates the problem by allowing tradeoffs between accuracy and computational complexity. However, previous works did not fully consider two essential requirements: optimized neural architecture and optimized inference policy. In this paper, we present HarvNet, which comprises two complementary techniques for generating and operating a multi-exit network for energy harvesting devices. First, we provide a neural architecture search scheme, HarvNAS, which configures the best multi-exit architecture while meeting memory and energy constraints. Second, HarvSched learns and constructs the best progressive inference policy with different energy constraints by considering runtime factors, such as the harvesting status and the energy storage level. We implemented HarvNAS using the TensorFlow framework and then implemented and evaluated HarvSched on an MSP430-based sensor device. The evaluation showed that HarvNAS generated a model with up to 2.6%p higher accuracy while saving up to 70% of memory compared to the existing technique, and HarvSched enabled zero-downtime operation of the generated model.

The live streaming of omnidirectional video (ODV) on mobile devices demands considerable network resources; thus, current mobile networks are incapable of providing users with high-quality ODV equivalent to conventional flat videos. We observe that mobile devices, in fact, underutilize graphics processing units (GPUs) while processing ODVs; hence, we envisage an opportunity exists in exploiting video super-resolution (VSR) for improved ODV quality. However, the device-specific discrepancy in GPU capability and dynamic behavior of GPU frequency in mobile devices create a challenge in providing VSR-enhanced ODV streaming. In this paper, we propose OmniLive, an on-device VSR system for mobile ODV live streaming. OmniLive addresses the dynamicity of GPU capability with an anytime inference-based VSR technique called Omni SR. For Omni SR, we design a VSR deep neural network (DNN) model with multiple exits and an inference scheduler that decides on the exit of the model at runtime. OmniLive also solves the performance heterogeneity of mobile GPUs using the Omni neural architecture search (NAS) scheme. Omni NAS finds an appropriate DNN model for each mobile device with Omni SR-specific neural architecture search techniques. We implemented OmniLive as a fully functioning system encompassing a streaming server and Android application. The experiment results show that our anytime VSR model provides four times upscaled videos while saving up to 57.15% of inference time compared with the previous super-resolution model showing the lowest inference time on mobile devices. Moreover, OmniLive can maintain 30 frames per second while fully utilizing GPUs on various mobile devices.

Energy-neutral operation (ENO) aims to provide near-perpetual device operation using energy harvested from ambient environments. Existing ENO techniques, however, have two key problems. The batteries used in harvesting devices have inherently limited lifespans, and the device experiences a long cold-start time when charging the battery. In this paper, we propose a long-lasting and highly responsive batteryless system, called RENO, to solve the problems that occur in energy harvesting devices. Using a supercapacitor to store energy, RENO maximizes the responsiveness in ENO especially running in dynamic harvesting environments such as indoor light energy harvesting. Combining the intermittent characteristics of power-neutral operation (PNO), RENO allows dual-mode operation of PNO and ENO, depending on the current harvesting capability. The device works as a PNO device when charging the energy storage, solving the ENO cold-start issue, while the harvested energy is efficiently managed with ENO. For this dual-mode operation, RENO provides hardware and software that handle the switch between PNO and ENO effectively. Application developers are provided with a well-defined API, which enables energy-efficient development of applications without detailed knowledge of the target hardware. Using the API, developers simply declare a task to be executed as either PNO or ENO, and the rest is handled by the system. The prototype system is implemented, and its functionality is evaluated in controlled environments. We also validate the proposed system with two real-world applications, proving the efficacy of dual-mode batteryless operation.

Duty cycle management is critical for the energy-neutral operation of batteryless devices. Many efforts have been made to develop an effective duty cycling method, including machine-learning-based approaches, but existing methods can barely handle the dynamic harvesting environments of batteryless devices. Specifically, most machine-learning-based methods require the harvesting patterns to be collected in advance, as well as manual configuration of the duty-cycle boundaries. In this article, we propose a configuration-free duty cycling scheme for batteryless devices, called CTRL, with which energy harvesting nodes tune the duty cycle themselves adapting to the surrounding environment without user intervention. This approach combines reinforcement learning (RL) with a control system to allow the learning algorithm to explore all possible search space automatically. The learning algorithm sets the target State of Charge (SoC) of the energy storage, instead of explicitly setting the target task frequency at a given time. The control system then satisfies the target SoC by controlling the duty cycle. An evaluation based on the real implementation of the system using publicly available trace data shows that CTRL outperforms state-of-the-art approaches, resulting in 40% less frequent power failures in energy-scarce environments while achieving more than ten times the task frequency in energy-rich environments.

Drones are commonly used in mission-critical applications, and the accurate estimation of available battery capacity before flight is critical for reliable and efficient mission planning. To this end, the battery voltage should be predicted accurately prior to launching a drone. However, in drone applications, a rise in the battery's internal temperature changes the voltage significantly and leads to challenges in voltage prediction. In this paper, we propose a battery voltage prediction method that takes into account the battery's internal temperature to accurately estimate the available capacity of the drone battery. To this end, we devise a temporal temperature factor (TTF) metric that is calculated by accumulating time series data about the battery's discharge history. We employ a machine learning-based prediction model, reflecting the TTF metric, to achieve high prediction accuracy and low complexity. We validated the accuracy and complexity of our model with extensive evaluation. The results show that the proposed model is accurate with less than 1.5% error and readily operates on resource-constrained embedded devices.

Transiently-powered devices rely solely on energy harvesters. Such devices typically use capacitors to store the harvested energy, but recent systems employ supercapacitors to store energy for extended operations. In a supercapacitor-based transiently-powered system, the energy-efficient estimation of the state-of-charge (SoC) of a supercapacitor is critical for practical use of the system, due to its tight energy budget. Conventional voltage-based schemes for capacitors do not provide accuracy in SoC estimation for supercapacitors. Also, the supercapacitor-specific SoC estimation which exploits its charge redistribution characteristics provides an accuracy, but the scheme demands significant overhead, and thus, is not applicable to transiently-powered systems, such as wireless sensor nodes. In this article, we specify three requirements for estimating supercapacitor SoC that should be met to function in transiently-powered systems. We then propose a scheme that meets those requirements. The proposed scheme does not require additional hardware, has a low computation cost, and operates when the system is intermittently powered. Thus, the scheme fits the energy-efficient operation of transiently-powered systems. We implemented the proposed scheme in real hardware and evaluated its functionality and accuracy. The proposed scheme estimated the SoC of supercapacitors with high accuracy, in various configurations and use scenarios, while guaranteeing operations in a typical energy harvesting environment.

Battery-powered devices commonly use Li-ion batteries due to their high energy density. Unfortunately, low ambient temperature and high discharge current significantly affect the available capacity of Li-ion batteries at runtime. One solution to handle this issue is to construct a hybrid energy storage (HES) system, taking advantage of the different performance characteristics of various Li-ion batteries. Conventional HES techniques often configure systems without considering the hardware overhead and capacity loss of batteries. Moreover, dynamic energy management in HES was not considered adequately due to severe power leakage. In this article, we propose DynLiB, a reconfigurable HES architecture that provides high energy availability in various runtime environments. DynLiB constructs HES with only two types of batteries, the main battery and the auxiliary battery, to reduce the operational cost of the hardware. The capacity of each battery is determined based on the application profile that characterizes the runtime factors. DynLiB supports energy transfer between batteries so that the available energy is maximized by dynamically transferring energy between the batteries, depending on the runtime environment. We implemented a prototype system to validate the efficacy of DynLiB. The experimental results showed that DynLiB achieved an average energy gain of 27.2% compared to a single-type battery system in various runtime environments.

Multiple photovoltaic (PV) modules are often used to provide enhanced harvesting capability for light energy-based Internet of Things (IoT) devices. PV modules facing multiple directions can lead to a situational energy loss when parts of the modules are shaded. To address this issue, existing solutions exploit reconfigurable PV arrays to acquire the optimal configuration in a given situation. However, conventional techniques are not energy efficient in estimating the harvesting capability of PV modules, and require high computation to find the optimal PV array at runtime. In this article, we propose PVoT, an energy-efficient reconfigurable PV array, which maximizes the harvesting energy for indoor IoT devices. To this end, we propose the use of photoresistors to estimate the harvesting capability with minimal energy overhead. We also provide hardware and software schemes, which perform event-driven light change detection in an energy-efficient way. Furthermore, we develop a power imbalance threshold metric to quickly find the optimal PV array at runtime. We implemented a prototype PVoT with off-the-shelf components and accompanying software. Experiments with the prototype hardware showed that PVoT achieves a gain of up to 23.9% in harvested energy compared to the existing directly connected PV array scheme.

Games are energy-intensive applications on mobile devices. Optimizing the energy efficiency of games is hence critical for battery-limited mobile devices. Although the advent of energy-aware scheduling (EAS) integrated in recent devices has provided opportunities for improved energy management, the framework is not specifically tuned for game applications. In this paper, we aim to improve the energy efficiency of game applications running on EAS-enabled mobile devices. To this end, we first analyze the functional characteristics of games, and investigate the source of the energy inefficiency. We then propose a scheme, called System-level Energy-optimization for Game Applications (SEGA), to improve the energy efficiency of games. SEGA governs CPU and GPU power consumption in a tightly coupled manner by employing three key techniques: (1) Lsync-aware GPU DVFS governor, (2) adaptive capacity clamping, and (3) on-demand touch boosting. We implemented SEGA on the latest Android-based smartphones. The evaluation results for 23 popular games showed that SEGA reduced the energy consumption of the Google Pixel 2 XL and Samsung Galaxy S9 Plus smartphones, at the device level, by 6.1–22.3 and 4.0–11.7 percent, respectively, with a quality of service (QoS) degradation of 1.1 and 0.5 percent, on average.

The quality of experience (QoE) is an important issue for users when accessing the web. Although many metrics have been designed to estimate the QoE in the desktop environment, few studies have confirmed whether the QoE metrics are valid in the mobile environment. In this paper, we ask questions regarding the validity of using desktop-based QoE metrics for the mobile web and find answers. We first classify the existing QoE metrics into several groups according to three criteria and then identify the differences between the mobile and desktop environments. Based on the analysis, we ask three research questions and develop a system, called WebMythBusters, for collecting and analyzing mobile web experiences. Through an extensive analysis of the collected user data, we find that (1) the metrics focusing on fast completion or fast initiation of the page loading process cannot estimate the actual QoE, (2) the conventional scheme of calculating visual progress is not appropriate, and (3) focusing only on the above-the-fold area is not sufficient in the mobile environment. The findings indicate that QoE metrics designed for the desktop environment are not necessarily adequate for the mobile environment, and appropriate metrics should be devised to reflect the mobile web experience.

Although work has been conducted on smartphone gaze tracking, the existing techniques are not pervasively used because of their heavy weight and low accuracy. Our preliminary analysis shows that these techniques would work better if their models were trained with data from tablets which have large screens. In this paper, we propose GAZEL, a runtime smartphone gaze-tracking scheme that achieves high accuracy on real devices. The key idea of GAZEL, a tablet-to-smartphone transfer learning, is to train a CNN model with data collected from tablets and then transplant the model to a smartphone. To achieve the goal, we designed a new CNN-based model architecture that is head pose resilient and light enough to operate at runtime. We also exploit implicit calibration to alleviate errors caused by differences in users' visual and device characteristics. The experiment results with commercial smartphones show that GAZEL achieves 27.5% better accuracy on smartphones compared to the state-of-the-art techniques and provides gaze tracking at up to 18 fps which is practically usable at runtime.

Cell imbalance in a multicell battery occurs over time due to varying operating environments. This imbalance leads to overall inefficiency in battery discharging due to the relatively weak cells in the battery. Reconfiguring the cells in the battery is one option for addressing the problem, but relevant circuits may lead to severe safety issues. In this article, we aim to optimize the discharge efficiency of a multicell battery using safety-supplemented hardware. To this end, we first design a cell string-level reconfiguration scheme that is safe in hardware operations and also provides scalability due to the low switching complexity. Second, we propose a machine learning-based run-time switch control that considers various battery-related factors, such as the state of charge, state of health, temperature, and current distributions. Specifically, by exploiting the deep reinforcement learning (DRL) technique, we train the complex relationship among the battery factors and derive the best switch configuration in run-time. We implemented a hardware prototype, validated its functionalities, and evaluated the efficacy of the DRL-based control policy. The experimental results showed that the proposed scheme, along with the optimization method, improves the discharge efficiency of multicell batteries. In particular, the discharge efficiency gain is maximized when the cells constituting the battery are unevenly distributed in terms of cell health and exposed temperature.

Unmanned aerial vehicles (UAVs) are often used in mission-critical applications, requiring a critical criterion in flight time. Unfortunately, severe power fluctuations, caused by specific flight patterns, degrade the deliverable capacity of the battery and hamper the flight time. A common approach to mitigating power fluctuations is to employ a hybrid energy storage system using a Li-ion battery with an ultracapacitor (UC). However, the conventional scheme poses inherent problems of low-energy density and power leakage due to the use of the UC and the supplementary hardware required for hybrid storage. In this article, we propose Hydrone, a reconfigurable battery architecture that maximizes the flight time of UAVs, overcoming the previous limitations. Hydrone addresses two key challenges that arise when hybrid energy storage is utilized in UAVs: 1) capacity loss and 2) power leakage. First, the proposed scheme compromises the capacity loss of hybrid storage by using a minimal capacity UC for use as a buffer to counteract the power fluctuations. Second, the power leakage of the hybrid battery is minimized by draining power from the UC only when it is necessary. To this end, the Hydrone architecture provides reconfigurability in hardware and offers two modes of battery operation, i.e., a battery-only mode and a hybrid mode. An appropriate operation is then selected at runtime depending on the flight situation and battery status. To switch modes, we employed a reinforcement learning-based switch control, reflecting the power fluctuation adequately on the flight and battery states. We implemented a hardware prototype to demonstrate the efficiency of Hydrone. Our extensive evaluation shows that the flight time of a UAV is prolonged up to 39% in our experiment setup.

The optimal use of batteries in drones is a critical issue for achieving both reliable operation and maximum flight time. The key is to acquire accurate information about the state of charge (SoC) of the battery in runtime. Drones typically employ series-connected lithium-ion polymer (Li-Po) battery cells, whose SoC is affected by many environmental factors as well as flight patterns. In this article, we propose a scheme, called Optrone, which maximizes the flight time of a drone while safely using the battery. Understanding the implications of the factors affecting the SoC of the drone's battery pack, we propose a three-level SoC, which is a metric for representing the SoC of a battery in runtime. We also provide various operating policies to users to improve the safety and efficiency of operating the drone. We implemented the prototype hardware and software for Optrone, and validated its operation in controlled and real environments. The experimental results in a controlled environment showed that the proposed three-level SoC poses less than 3% error and the operating policies achieved a flight time gain of 19.4%, while guaranteeing battery safety. We also observed a flight time gain of about 10% in real outdoor experiments, where the user rightly adheres to the advised Optrone policy.

Web browsing, previously optimized for the desktop environment, is being fine-tuned for energy-efficient use on mobile devices. Although active attempts have been made to reduce energy consumption, the advent of energy-aware scheduling (EAS) integrated in the recent devices suggests the possibility of a new approach for optimizing energy use by browsers. Our preliminary analysis showed that the existing EAS-enabled system is overly optimized for performance, leading to energy inefficiencies while a web browser is running. In this paper, we analyze the characteristics of web browsers, and investigate the cause of energy inefficiency in EAS-enabled mobile devices. We then propose a system, called WebTune, to improve the energy efficiency of mobile browsers. Exploiting the reinforcement learning technique, WebTune learns the optimal execution speed of the web browser's processes, and adjusts the speed at runtime, thus saving energy and ensuring the quality of service (QoS). WebTune is implemented on the latest Android-based smartphones, and evaluated with Alexa's top 200 websites. The experimental results show that WebTune reduced the device-level energy consumption of the Google Pixel 2 XL and Samsung Galaxy S9 Plus smartphones by 18.7-22.0% and 13.7-16.1%, respectively, without degrading the QoS.

Graphics increasingly play a key role in modern mobile devices. The graphics pipeline requires a close relationship between the CPU and the GPU to ensure energy efficiency and the user's quality of experience (QoE). Our preliminary analysis showed that the current techniques employed to achieve energy efficiency in the Android graphics pipeline are not optimized especially in the frame generation process. In this paper, we aim to improve the energy efficiency of the Android graphics pipeline without degrading the user's QoE. To achieve this goal, we studied the internals of the Android graphics pipeline and observed the energy inefficiency in the existing governing framework of the CPU and GPU. Based on the findings, we propose three techniques for addressing energy inefficiency: (1) aggressively capping the maximum CPU frequency, (2) lowering the CPU frequency by raising the GPU minimum frequency, and (3) allocating the frame rendering-related threads in the energy-efficient CPU cores. These techniques are integrated into a single governing framework, called the GFX Governor, and implemented in the newest Android-based smartphones. Experimental results show that without hampering the user's QoE the average energy consumption of Nexus 6P, Pixel XL, and Pixel 2 XL is reduced at the device level by 24.2%, 18.6%, and 13.7%, respectively, for the 60 chosen applications. We also analyzed the efficacy of the proposed technique in comparison with the state-of-the-art Energy-Aware Scheduling (EAS) implemented in the latest smartphone.

Mobile users are always demanding extended availability in their battery use. Together with enlarged battery capacity, fast charging is one approach that provides an improved user experience in battery use. Recently, device manufacturers have been developing a variety of fast charging techniques for mobile devices. However, the existing techniques severely reduce the charging power when the device is in use while charging. We experimentally demonstrate that the primary cause of the reduction in the charge speed during device use is to cope with the performance degradation incurred by heat generation. We then propose an adaptive charging scheme, called Always-on Quick Charging, which enables fast charging especially when the device is in use. The key idea of our approach is to adjust the charging power while ensuring that the heat generated by the charging does not affect the performance. The proposed scheme is implemented in Google's Pixel 2XL smartphone. The experiment with a real-world usage scenario shows that the charging speed of the proposed scheme is up to 2.4 times faster than the default scheme, while preserving device performance.

Predicting a click sequence in mobile applications improves the user experience in various ways. By predicting which button will be clicked next, one can predict how the application will work and how the device will operate. However, predicting the click sequence is difficult because of the problems involved in collecting click sequences in real application usage. More importantly, accurate predictions are extremely challenging. In this paper, we address these issues. We propose PathFinder, a scheme for collecting click events and based on them predicting the next click in the application. The clicks are collected with the Android Accessibility Service and the next click is predicted via long shortterm memory (LSTM). For the prediction, the base click sequence model is first generated from all users' data; then, a personalized model is trained with an individual click sequence. As training data considerably influences the performance of LSTM, several techniques are developed to enhance the quality of the training data. The experimental results for 100 popular applications showed that the coverage and accuracy of click sequence tracing were 95% and 96%, respectively. Furthermore, PathFinder predicted the top three buttons that would be clicked next with a 0.76 F-measure for 1775043 real click data.

Managing the power consumption of display-related components in mobile devices is difficult because of performance degradation. Therefore, eliminating hidden workloads, such as redundant frames, is preferable, as it directly reduces power without affecting the user experience. Our preliminary study shows that the default launcher of the Android Open Source Project (AOSP) and popular applications, such as Instagram and Pinterest, generate redundant frames. In this paper, we propose a scheme to optimize the power consumption of the smartphone's display-related components by preventing redundant frames generation. By analyzing the frame-generation process, we observe that redundant frame generation is possible in the current Android framework. We then propose a scheme that recognizes and prevents redundant frame generation before actual frame generation (i.e., frame rendering in the GPU). The proposed scheme utilizes a display list, which was introduced in recent Android smartphones for efficient frame generation. We implemented the proposed scheme on Nexus smartphones. On the Nexus 5, the proposed solution reduced the energy of the AOSP default launcher, Instagram, and Pinterest by 40, 35.4, and 39.6 percent, respectively. Furthermore, the experimental results with a general usage scenario showed that our scheme prevented about 35 percent of redundant frame generation with a false-positive rate of 1.8 percent.

Smartphones' surface temperature, also called skin temperature, can rapidly heat up in certain cases, and this causes a variety of safety problems. Therefore, the thermal management of smartphones should consider the skin temperature, and its accurate prediction is important. However, due to the complicated relationship among the many exothermic components in the device, predicting skin temperature is extremely difficult. In this paper, we develop a thermal prediction model that accurately predicts the skin temperature of a mobile device. In an experiment with smartphones, we show that the proposed model achieves an accuracy of 98%, with a ±0.4 °C margin of error. To the best of our knowledge, our work is the first to reveal the complex relationship between the various components inside of a smartphone and its skin temperature.

The thermal issue for mobile devices becomes critical as the devices' performance increases to handle complicated applications. Conventional thermal management limits the performance of the entire device, degrading the quality of both foreground and background applications. This is not desirable because the quality of the foreground application, i.e., the frames per second (FPS), is directly affected, whereas users are generally not aware of the performance of background applications. In this paper, we propose an app-oriented thermal management scheme that specifically restricts background applications to preserve the FPS of foreground applications. For efficient thermal management, we developed a model that predicts the heat contribution of individual applications based on hardware utilization. The proposed system gradually limits system resources for each background application according to its heat contribution. The scheme was implemented on a Galaxy S8+ smartphone, and its usefulness was validated with a thorough evaluation.

DVFS is a widely used methodology for reducing the power consumption of mobile devices. This scheme involves frequency scaling in accordance with a specific governor and the establishment of an operating voltage to be paired with frequency. Incorporated into the settings for operating voltage is a guardband that ensures safe processor operation even at the worst conditions of on-chip temperature. Typically, the processor temperature remains at a normal range (i.e., not the worst), hence the voltage guardband set to guarantee safe operation is overly protected. In this paper, we propose a temperature-aware DVS (T-DVS) that aggressively reduces voltage guardband. We explore the opportunity to provide minimum operating voltages for frequencies at different temperatures and realize a dynamic voltage control scheme that reduces power consumption. The T-DVS manages temperature so that it remains in the “green zone” where maximum voltage gain is enabled for power-efficient operation. We validate the effectiveness of the T-DVS under various thermal conditions by using mobile application processors and different operating scenarios. Experimental results show that the T-DVS leads to power gain without degrading performance regardless of thermal conditions and chip characteristics. By examining the real-world applications of and off-the-shelf smartphone, we show that the voltage gains generated by the T-DVS results in battery lifetime increment.

Extending battery lifetime is an important issue for mobile devices. While extensive attempts have been made at the software level, optimization often risks hampering user experience. One fundamental method to increase battery lifetime is to improve the efficiency of the battery itself. We argue that the multi-cell battery system, which is widely used for enhancing battery efficiency in the electric vehicle (EV) field, can solve this issue. However, due to the hardware constraints and device usage characteristics, battery advancements in the EV field are not directly applicable to mobile devices. In this paper, we propose BattMan, a multi-cell battery management system for mobile devices, for the enhancement of battery efficiency. We develop an accurate battery cell model to estimate the expected battery lifetime considering the recovery effect, the rate capacity effect, and battery aging. We also propose a multi-cell scheduling algorithm to maximize the overall battery lifetime. We implemented BattMan on recent smartphones and evaluated its impact on battery lifetime. The experimental results show that a two-cell configuration of the proposed system increases battery lifetime by an average of between 14-19%, depending on cell aging, in real usage scenarios over a single-cell battery of the same overall capacity. We hope the proposed multi-cell battery scheme opens up a new direction towards battery lifetime improvement in mobile devices.

With the increasing popularity of smartphones, battery life is among the most crucial issues for mobile users. This paper presents a crowdsensing-based use guide to extend the lifetime of smartphones. The system answers a question raised by phone usage: Why is my phone battery draining quickly compared to others phones despite running the same applications? The proposed system pinpoints the major causes of battery drain in terms of both hardware and software aspects. In relation to the hardware aspect, the system quantifies degree of battery aging as a ratio metric; an estimate of 50% indicates that the battery is at half of full capacity, meaning that battery usage time is approximately half that of a new battery. The system automatically profiles battery age based on charging duration data collected by crowdsensing. In its software aspect, the system guides phone configuration to extend application usage times. The system mines large-scale usage data to infer the major energy holes in a user's phone usage. The scheme works autonomously without user intervention and does not require any external equipment. Extensive evaluation with 3,000 users demonstrated that the proposed scheme successfully extends battery life for typical mobile users.

Dynamic Voltage and Frequency Scaling (DVFS) is a widely used methodology to reduce the power consumption of mobile devices. This scheme performs frequency scaling in accordance with a specific governor and sets an operating voltage to be paired with the frequency. Temperature is one of the critical parameters affecting device operation. Practically, a guard-band exists in the operating voltage to ensure safe processor operation even at the worst temperature. DVFS can be optimized in terms of operating voltage under nominal conditions. In this paper, we propose a Temperature-aware DVS (T-DVS) that aggressively reduces the voltage guard-band. We explore the opportunity of providing the minimum operating voltages for frequencies at different temperatures and realize a dynamic voltage control scheme to optimize power consumption. The effectiveness of T-DVS is validated under various thermal conditions by using multi-core application processor. We experimentally observe that T-DVS leads to voltage gain without performance degradation regardless of both thermal conditions and chip characteristics. We show by using off-the-shelf smartphones that the voltage gain achieved by the scheme results in battery lifetime increment.

Mobile data offloading through WiFi is an essential requirement to reduce cellular network traffic. While extensive attempts have been made at mobile data offloading, previous studies have rarely addressed practical issues, such as dealing with diverse user contexts. In this paper, we propose a personalized data offloading scheme to provide maximum throughput within the cellular budget in daily life. We propose an adaptive policy that considers a user's mobility patterns, cellular budget, and network usage for applications. The proposed system employs an adaptive model to predict the throughput of WiFi APs and the network usage of smartphones. Among the three types of predictor model (i.e., spatial, temporal, and spatio-temporal), the system automatically chooses the optimal model for each mobile user without user intervention. The experimental results from 10 mobile users show that the proposed system provides 29% higher throughput than previous systems and minimizes extra data charges.

Research of daily activity recognition has been extensively conducted in the field of ubiquitous computing. However, previous daily activity recognition schemes are either obtrusive or inaccurate since they use just special-purpose devices. In this paper, we propose the collaborative classification for recognizing daily activities with a smartwatch. We exploit a single off-the-shelf smartwatch to distinguish 5 different daily activities such as eating, vacuuming, sleeping, showering, and TV watching. More precisely, we conduct experiments for collecting sensor data from accelerometer and acoustic sensor which are embedded in a smartwatch. However, the simple combination of the raw acceleration and acoustic data does not deliver accurate recognition accuracy. In order to achieve high accuracy, we propose a collaborative classification algorithm which integrates sensor data and ground-truth label for improving recognition accuracy by constructing a mapping table. We evaluate accuracies using single-sensor based approach, multi-sensor based approach, and our collaborative classification approach. The results from activity recognition for about 20 hours data collected by subjects show reliable accuracies for all 5 activities, and the overall accuracy of our collaborative approach is about 91.5%. Experimental results reveal that our approach improves the recall rate of each activity by up to 21.5% as compared to that of the simply combined multi-sensor based approach.

Energy consumption in mobile devices is an important issue for both system developers and users. Users are aware of the battery-related information of their mobile devices and tend to take appropriate actions to increase the battery life. In this article, we propose a framework that accurately estimates the remaining battery time of applications at runtime. The framework profiles the power behavior of applications tied with activated hardware components and estimates the remaining battery budget utilizing the battery-related data provided by the device. The experiments validate that our method predicts the remaining battery time for applications with approximately 93% of accuracy.

Despite numerous studies to reduce the power consumption of the display-related components of mobile devices, previous works have led to a deterioration in user experience due to compromised graphic quality. In this paper, we propose an effective scheme to reduce the energy consumption of the display subsystems of mobile devices without compromising user experience. In preliminary experiments, we noticed that mobile devices typically perform redundant display updates even if the display content does not change. Based on this observation, we first propose a metric called the content rate, which is defined as the number of meaningful frame changes in a second. Our scheme then estimates an optimal refresh rate based on the content rate in order to eliminate redundant display updates. Also proposed is the flicker compensation technique, which prevents the flickering problem caused by the reduced refresh rate. Extensive experiments conducted on the latest smartphones demonstrated that our system effectively reduces the overall power consumption of mobile devices by 35 percent while simultaneously maintaining satisfactory display quality.

Battery-related problems in mobile devices have been extensively investigated in both industry and literature. In particular, battery aging is a critical issue, since battery lifetime decreases as usage time increases. Battery aging primarily causes inconvenience to users by necessitating frequent recharging, and also affects the accuracy of power estimations for mobile devices. Evaluating battery aging and its effects has rarely been addressed in prior works. In this paper, we propose an online scheme to quantify the battery aging of mobile devices. Specifically, we estimate the degree of battery aging as a ratio metric based on patterns of charging time. For example, an estimate of 50% indicates that the battery capacity is only half of full capacity, meaning that the battery usage time is only approximately half that of the new battery's. Our scheme works autonomously on mobile devices and does not require any external equipment. The extensive experiments demonstrated that the proposed scheme quantifies battery aging accurately.

Energy accounting is an essential requirement for optimizing energy consumption on mobile devices. State-of-the-art approaches consider application processes and threads as the sole components of energy consumption. In this framework, the energy consumption of system services is unclear and has not been comprehensively studied. In this paper, we suggest that the energy consumption of system services should be investigated to understand the behavior of applications. We propose a fine-grained energy tracing scheme, EnTrack, to enhance the accuracy of energy tracing by identifying and incorporating the energy portions consumed by system services. We implemented EnTrack on the Android platform and validated its functionality and usefulness. In addition, practical usage cases of EnTrack, which uses it as an energy behavior analysis tool, were introduced. The case studies demonstrated that EnTrack enables an understanding of fine-grained energy consumption, especially in system services, which have previously been concealed.

Many smart home applications, such as monitoring for the elderly and home automation, require location information for individual occupants. Several techniques have been proposed for tracking occupants in a home environment. However, the current techniques do not provide a seamless in-home locating system owing to the occupants' device-free movement and the lack of cost-effective infrastructure for home location tracking. In this paper, we propose a home occupant tracking system that uses a smartphone and an off-the-shelf smartwatch without additional infrastructure. In our system, activity fingerprints are automatically generated from the microphone and the inertial sensors of the smartwatch, and location information is periodically obtained from the smartphone. We designed a hidden Markov model using the relationship between home activities and the room's location. Extensive experiments showed that our system tracks the location of users with 87% accuracy, even when there is no manual training for activities.

Location estimation methods using radio fingerprint have been studied extensively. The approach constructs a database that associates ambient radio signals with physical locations in training phase, and then estimates the location by finding the most similar signal pattern within the database. To achieve robust and accurate location estimation, the training phase should be conducted across the entire target space. In practice, however, a user may only access limited or authorized places in a building, that causes degradation in accuracy. In this paper, we present a smartphone-based autonomous indoor war-walking scheme, which automatically constructs the location fingerprint database, even covering unvisited locations. While a smartphone user explores the target area, the proposed system tracks the user's trajectory and simultaneously trains the location fingerprint database. Furthermore, our scheme interpolates radio signals in the database with an appropriate radio propagation model, and supplements fingerprints for unvisited places. As a result, although a user may sparsely explore the target site, the scheme returns the complete database. We implemented our solution and demonstrated the feasibility of the solution.

Abstract The commercial success of smartphones increases the feasibility of mobile ad hoc networking in daily life; we define such networks as spontaneous smartphone networks (SSNs). Efficient data delivery in SSNs is challenging because of the low node density, ambiguous contact opportunities, and short message lifetime. The existing schemes attempt to select optimal relays via various cumulative metrics (e.g., encounter history, social centrality, or contact distribution), whose effectiveness is ambiguous and suboptimal. In this paper, we introduce a Markov predictor‐based transient delivery scheme that quantifies the regularity of small time scale movement for forwarding decisions. Unlike previous works, we utilized fine‐grained mobility data to reduce errors of estimating contact opportunities and contact duration. On the basis of this forwarding strategy, we developed a multi‐copy routing scheme. The evaluation using real traces indicates that the proposed approach outperforms compared alternatives in terms of delivery rate and cost. Copyright © 2013 John Wiley & Sons, Ltd.

While active studies have been conducted to reduce the power consumption of display-related components of mobile devices, previous work has rarely approached the issues without having to deteriorate graphical quality. In this paper, we propose an effective scheme to reduce display energy consumption without compromising user experience. We first define a metric called the content rate from which an appropriate refresh rate is determined for displaying content. The proposed system then sets an optimal refresh rate based on the content rate. Extensive experiments demonstrate that our system effectively reduces the total power in commercial smartphones, yet the display quality is satisfactorily maintained.

Mobile sensing systems employ various sensors in smartphones to extract human-related information. As the demand for sensing systems increases, a more effective mechanism is required to sense information about human life. In this paper, we present a systematic study on the feasibility and gaining properties of a crowdsensing system that primarily concerns sensing WiFi packets in the air. We propose that this method is effective for estimating urban mobility by using only a small number of participants. During a seven-week deployment, we collected smartphone sensor data, including approximately four million WiFi packets from more than 130,000 unique devices in a city. Our analysis of this dataset examines core issues in urban mobility monitoring, including feasibility, spatio-temporal coverage, scalability, and threats to privacy. Collectively, our findings provide valuable insights to guide the development of new mobile sensing systems for urban life monitoring.

A smartphone battery interface should provide energy-related information efficiently with high accuracy because it is the basis for user actions on battery consumption. Previous studies have mainly focused on battery information per se, but the effectiveness of user interaction with the interface has barely been studied. In this paper, we first discuss the results of a survey on different types of energy-related information and features to investigate what users want in terms of battery consumption for smartphones. We then introduce Powerlet, which is a new battery interface that attempts to actively interact with users to provide battery usage information. We validated the efficiency of Powerlet with real users' experiences derived over a seven-week study period. We found that the users actively managed and changed their phone usage based on the energy-related information provided by Powerlet. With the Powerlet active battery interface, participants reduced daily energy consumption by 8.2% on average.

Quality improvement in mobile applications should be based on the consideration of several factors, such as users' diversity in spatio-temporal usage, as well as the device's resource usage, including battery life. Although application tuning should consider this practical issue, it is difficult to ensure the success of this process during the development stage due to the lack of information about application usage. This paper proposes a user interaction-based profiling system to overcome the limitations of development-level application debugging. In our system, the analysis of both device behavior and energy consumption is possible with fine-grained process-level application monitoring. By providing fine-grained information, including user interaction, system behavior, and power consumption, our system provides meaningful analysis for application tuning. The proposed method does not require the source code of the application and uses a web-based framework so that users can easily provide their usage data. Our case study with a few popular applications demonstrates that the proposed system is practical and useful for application tuning.

Finding a missing child is an important problem concerning not only parents but also our society. It is essential and natural to use serendipitous clues from neighbors for finding a missing child. In this paper, we explore a new architecture of crowd collaboration to expedite this mission-critical process and propose a crowd-sourced cooperative mobile application, CoSMiC. It helps parents find their missing child quickly on the spot before he or she completely disappears. A key idea lies in constructing the location history of a child via crowd participation, thereby leading parents to their child easily and quickly. We implement a prototype application and conduct extensive user studies to assess the design of the application and investigate its potential for practical use.

The goal of energy monitoring and eco-feedback systems is to induce energy consumers to change their behaviors to achieve a more sustainable way of life. Comprehensive research and commercial solutions provide energy consumers with information about overall energy costs or appliance-level energy usage. However, in order to better promote spontaneous energy saving with an eco-feedback system, personalized information is required. The conventional solutions cannot identify the energy usage of an individual user in a shared residential environment. In this paper, we propose the Personalized Energy Auditor, which estimates personal energy usage at home. Our system monitors and analyzes appliance usage, as well as the energy cost of the daily activities of residents. The system then estimates personal energy usage automatically, by linking appliance usage data with the individual user. Our system was installed in residential homes, and the experimental results indicate that it accurately estimates personal energy usage.

Welcome to Chongqing, China, and the 20th IEEE International Conference on Embedded and Real- Time Computing Systems and Applications (RTCSA 2014). RTCSA has been a long-running technical conference sponsored by IEEE. The objective of the conference is to bring together academic researchers and industry developers for intensive discussion of recent advancing in the field of embedded systems, real-time systems, system design practice and emerging applications.

Smartphones enable the collection of mobility data using various sensors. The key challenge in the collection of continuous data is to overcome the limited battery capacity of the device. While extensive research has been conducted to solve energy issues in continuous mobility learning, we argue that previous works have not reached optimal performance. In this paper, we propose an energy-efficient mobility monitoring system, FreeTrack, to collect place-centric mobility data with minimum energy consumption in everyday life. We first analyzed the regularity of life patterns, cellular connection patterns, and battery charging behaviors of 94 smartphone users to examine important features related to human mobility. Based on our findings, we design an adaptive duty cycling scheme that uses zero-cost information (i.e., regular mobility, cell connection, and battery state) as low-level sensing to infer location change without the need to activate sensors. We model the location inference on the Hidden Markov Model and optimize the sensing schedule of individual smartphones for real-time operation. Our extensive experiment with 48 smartphone users shows that the proposed system achieves an energy saving of about 68% over previous works, yet still correctly traces 97% of mobility with 0.2±0.5 places misses in a day.

Monitoring a user's mobility during daily life is an essential requirement in providing advanced mobile services. While extensive attempts have been made to monitor user mobility, previous work has rarely addressed issues with predictions of temporal behavior in real deployment. In this paper, we introduce SmartDC, a mobility prediction-based adaptive duty cycling scheme to provide contextual information about a user's mobility: time-resolved places and paths. Unlike previous approaches that focused on minimizing energy consumption for tracking raw coordinates, we propose efficient techniques to maximize the accuracy of monitoring meaningful places with a given energy constraint. SmartDC comprises unsupervised mobility learner, mobility predictor, and Markov decision process-based adaptive duty cycling. SmartDC estimates the regularity of individual mobility and predicts residence time at places to determine efficient sensing schedules. Our experiment results show that SmartDC consumes 81 percent less energy than the periodic sensing schemes, and 87 percent less energy than a scheme employing context-aware sensing, yet it still correctly monitors 90 percent of a user's location changes within a 160-second delay.

Modeling and estimating power consumption of OLED displays are necessary to understand the energy behavior of emerging mobile devices. Although previous study exists to model and estimate the power consumption of stationary display images, to the best of our knowledge, no prior work is found to deal with runtime power behavior of OLED display running real applications. This paper proposes a runtime power estimation scheme for OLED displays that involves monitoring kernel activities that capture the screen change events of running applications. The experiment results show that the proposed scheme estimates the display energy consumption of running applications with reasonable accuracy.

Android provides a WakeLock mechanism for application developers to ensure the proper execution of applications without having to enter the sleep state of a device. When using the WakeLock mechanism, application developers should bear the responsibility of adequately releasing the acquired lock. Otherwise, the energy will unnecessarily be wasted due to a locked application. This paper presents a scheme, called WakeScope, to handle WakeLock misuse. The scheme is designed to detect and notify of a misuse case of WakeLock handling, which may arise with an application and even with an Android runtime system, and thus provides a practical tool to prevent energy waste in mobile devices. Our experiments with real applications show that WakeScope accurately detects the misused case, with runtime overhead of approximately 1.2% in CPU usage.

Crowd-enabled place-centric systems gather and reason over large mobile sensor datasets and target everyday user locations (such as stores, workplaces, and restaurants). Such systems are transforming various consumer services (for example, local search) and data-driven organizations (city planning). As the demand for these systems increases, our understanding of how to design and deploy successful crowdsensing systems must improve. In this paper, we present a systematic study of the coverage and scaling properties of place-centric crowdsensing. During a two-month deployment, we collected smartphone sensor data from 85 participants using a representative crowdsensing system that captures 48,000 different place visits. Our analysis of this dataset examines issues of core interest to place-centric crowdsensing, including place-temporal coverage, the relationship between the user population and coverage, privacy concerns, and the characterization of the collected data. Collectively, our findings provide valuable insights to guide the building of future place-centric crowdsensing systems and applications.

To prolong the battery lifetime of modern mobile devices, the energy management policy should be developed in a personalized way, adequately reflecting user context or the energy behavior of the user. The first step toward this personalization is to collect the relevant information, accurately and efficiently, from the device. This paper presents a fine-grained and low-overhead framework, called UserScope, which is designed to collect energy-related user contexts in Android smartphones. We classified energy-related smart phone usage and designed an appropriate set of monitoring parameters to collect from the system. The UserScope core is then implemented as a kernel module to collect all the necessary information in an event-driven manner. This kernel-level implementation ensures monitoring accuracy and low system overhead. UserScope also provides a data-sharing mechanism with which other software components in the system can easily interface. Our experiments show that User Scope accurately extracts energy related system information with 0.8% CPU overhead. The practicality of UserScope is also validated with real deployment and subsequent analysis of the collected data.

A user's location information is commonly used in diverse mobile services, yet providing the actual name or semantic meaning of a place is challenging. Previous works required manual user interventions for place naming, such as searching by additional keywords and/or selecting place in a list. We believe that applying mobile sensing techniques to this problem can greatly reduce user intervention. In this paper, we present an autonomous place naming system using opportunistic crowdsensing and knowledge from crowdsourcing. Our goal is to provide a place name from a person's perspective: that is, functional name (e.g., food place, shopping place), business name (e.g., Starbucks, Apple Store), or personal name (e.g., my home, my workplace). The main idea is to bridge the gap between crowdsensing data from smartphone users and location information in social network services. The proposed system automatically extracts a wide range of semantic features about the places from both crowdsensing data and social networks to model a place name. We then infer the place name by linking the crowdsensing data with knowledge in social networks. Extensive evaluations with real deployments show that the proposed system outperforms the related approaches and greatly reduces user intervention for place naming.

Fueled by the widespread adoption of sensor-enabled smartphones, mobile crowdsourcing is an area of rapid innovation. Many crowd-powered sensor systems are now part of our daily life -- for example, providing highway congestion information. However, participation in these systems can easily expose users to a significant drain on already limited mobile battery resources. For instance, the energy burden of sampling certain sensors (such as WiFi or GPS) can quickly accumulate to levels users are unwilling to bear. Crowd system designers must minimize the negative energy side-effects of participation if they are to acquire and maintain large-scale user populations.

Reducing the standby power used by home appliances is critical in a household energy management system. Although significant effort has been made to minimize the standby power use of appliances, manual operation is still required to eliminate standby power usage. Additionally, the current regulation strategy of standby power typically focuses on real-power consumption, and it does not consider the apparent power and power factors. We propose an automatic standby power reduction system that is based on user-context profiling. Our system profiles and analyzes the occupancy pattern, as well as the appliance usage. The system then actively manages standby power utilization by predicting the probabilities of future appliance usage. We built a prototype smart meter to monitor and control the power lines. We also developed software that implements the proposed scheme. Our experiments, conducted for three to five weeks in four households, show that power consumption in standby mode can be reduced.

With the growing number of smartphone users, peer-to-peer ad hoc content sharing is expected to occur more often. Thus, new content sharing mechanisms should be developed as traditional data delivery schemes are not efficient for content sharing due to the sporadic connectivity between smartphones. To accomplish data delivery in such challenging environments, researchers have proposed the use of store-carry-forward protocols, in which a node stores a message and carries it until a forwarding opportunity arises through an encounter with other nodes. Most previous works in this field have focused on the prediction of whether two nodes would encounter each other, without considering the place and time of the encounter. In this paper, we propose discover-predict-deliver as an efficient content sharing scheme for delay-tolerant smartphone networks. In our proposed scheme, contents are shared using the mobility information of individuals. Specifically, our approach employs a mobility learning algorithm to identify places indoors and outdoors. A hidden Markov model is used to predict an individual's future mobility information. Evaluation based on real traces indicates that with the proposed approach, 87 percent of contents can be correctly discovered and delivered within 2 hours when the content is available only in 30 percent of nodes in the network. We implement a sample application on commercial smartphones, and we validate its efficiency to analyze the practical feasibility of the content sharing application. Our system approximately results in a 2 percent CPU overhead and reduces the battery lifetime of a smartphone by 15 percent at most.

Automated and scalable approaches for understanding the semantics of places are critical to improving both existing and emerging mobile services. In this paper, we present [email protected] (CSP), a framework that exploits a previously untapped resource -- opportunistically captured images and audio clips from smartphones -- to link place visits with place categories (e.g., store, restaurant). CSP combines signals based on location and user trajectories (using WiFi/GPS) along with various visual and audio place "hints" mined from opportunistic sensor data. Place hints include words spoken by people, text written on signs or objects recognized in the environment. We evaluate CSP with a seven-week, 36-user experiment involving 1,241 places in five locations around the world. Our results show that CSP can classify places into a variety of categories with an overall accuracy of 69%, outperforming currently available alternative solutions.

A mobility model is an essential requirement in accurately predicting an individual's future location. While extensive studies have been conducted to predict human mobility, previous work used coarse-grained mobility data with limited ability to capture human movements at a fine-grained level. In this paper, we empirically analyze several mobility models for predicting temporal behavior of an individual user. Unlike previous approaches, which employed coarse-grained mobility data with partial temporal-coverage, we use fine-grained and continuous mobility data for the evaluation of mobility models.We explore the regularity and predictability of human mobility, and evaluate location-dependent and location-independent models with several feature-aided schemes. Our experimental results show that a location-dependent predictor is better than a location-independent predictor for predicting temporal behavior of individual user. The duration of stay at a location is strongly correlated to the arrival time at the current location and the return-tendency to the next location, rather than recent k location sequences.We also find that false-positive predictions can be reduced by adaptive use of mobility models.

The Wi-Fi fingerprinting (WF) technique normally suffers from the RSS (Received Signal Strength) variance problem caused by environmental changes that are inherent in both the training and localization phases. Several calibration algorithms have been proposed but they only focus on the hardware variance problem. Moreover, smartphones were not evaluated and these are now widely used in WF systems. In this paper, we analyze various aspect of the RSS variance problem when using smartphones for WF: device type, device placement, user direction, and environmental changes over time. To overcome the RSS variance problem, we also propose a smartphone-based, indoor pedestrian-tracking system. The scheme uses the location where the maximum RSS is observed, which is preserved even though RSS varies significantly. We experimentally validate that the proposed system is robust to the RSS variance problem.

This paper presents experiment results showing the relationship between operating temperature and the lifetime of mobile devices. Conventional knowledge is that energy-efficiency is higher at lower operating temperatures. However, we show that this assumption is not always true, because the energy-efficiency of mobile device is governed not only by static leakage power but also by the chemical characteristics of the battery. We believe that future research on thermal management of mobile devices should consider the results of these observations.

Context-aware service provisioning for mobile phones is challenging because of diverse user contexts and mobile applications. The context recognition process generally reduces the device performance due to the competitive use of limited resources in the mobile phone. While extensive attempts have been made to provide appropriate services based on user context, previous work is limited to supporting diverse user contexts and various services. In this paper, we introduce PION, a framework for personalized service provisioning to manage diverse user contexts and provide appropriate mobile services in daily life. PION comprises the Service Hub and Pioneer. The Service Hub is a service agent server between a smartphone user and a service provider that defines the properties of mobile services. The Pioneer collects cognitive context data, classifies user contexts and their relations, and predicts essential mobile services based on a user's mobility data. We have implemented PION on the Android framework, and our evaluation demonstrates its efficiency in managing diverse user contexts and providing mobile services in real deployments. We believe that the PION framework is a viable context-aware system for smartphone users.

Stateless address autoconfiguration protocols allow nodes to select addresses and validate the uniqueness of a selected address by duplicate address detection (DAD). The considerable cost of DAD results from the message complexity increase in multihop network topologies, such as wireless sensor networks. This paper proposes a lightweight, hybrid address autoconfiguration protocol, called Spectrum, that deploys IPv6-compatible addresses into 6LoWPAN networks in a distributed manner. Spectrum creates the virtual coordinate system on the network and deploys addresses based on the location of the nodes. The deployment policy based on the virtual locations of the nodes reduces the DAD cost in the initial configuration as well as the cost for additional configurations of newly arrived nodes. The authors implemented and tested the proposed scheme in a real environment. Simulations and experiments confirmed a reasonable message cost for both stateful and stateless autoconfigurations.

Currently available location technologies such as the global positioning system (GPS) or Wi-Fi fingerprinting are limited, respectively, to outdoor applications or require offline signal learning. In this paper, we present a smart phone-based autonomous construction and management of a personalized location provider in indoor and outdoor environments. Our system makes use of electronic compass and accelerometer, specifically for indoor user tracking. We mainly focus on providing point of interest (POI) locations with room-level accuracy in everyday life. We present a practical tracking model to handle noisy sensors and complicated human movements with unconstrained placement. We also employ a room-level fingerprint-based place-learning technique to generate logical location from the properties of pervasive Wi-Fi radio signals. The key concept is to track the physical location of a user by employing inertial sensors in the smartphone and to aggregate identical POIs by matching logical location. The proposed system does not require a priori signal training since each user incrementally constructs his/her own radio map into their daily lives. We implemented the system on Android phones and validated its practical usage in everyday life through real deployment. The extensive experimental results show that our system is indeed acceptable as a fundamental system for various mobile services on a smartphone.

Although active research has recently been conducted on received signal strength (RSS) fingerprint-based indoor localization, most of the current systems hardly overcome the costly and time-consuming offline training phase. In this paper, we propose an autonomous and collaborative RSS fingerprint collection and localization system. Mobile users track their position with inertial sensors and measure RSS from the surrounding access points. In this scenario, anonymous mobile users automatically collect data in daily life without purposefully surveying an entire building. The server progressively builds up a precise radio map as more users interact with their fingerprint data. The time drift error of inertial sensors is also compromised at run-time with the fingerprint-based localization, which runs with the collective fingerprints being currently built by the server. The proposed system has been implemented on a recent Android smartphone. The experiment results show that reasonable location accuracy is obtained with automatic fingerprinting in indoor environments.

WiFi positioning systems require radio maps in the form of either RF fingerprints or positions of WiFi access points (APs). In particular, knowledge of the AP positions is essential to enable a locating mechanism as well as to understand the nature of underlying WiFi networks, such as density, connectivity, interference characteristics, and so on. In this paper, we propose an approach called Serendipity, which locates WiFi APs in an unsupervised manner using radio scans collected by ordinary smartphone users. From the radio scans, we extract dissimilarities between all pairs of WiFi APs and estimate relative positions of APs by analyzing the dissimilarities based on a multidimensional scaling technique. We then find the absolute positions with additional radio scans whose positions are known. The discovered positions of WiFi APs are used for the positioning of smartphones or the management of the WiFi networks. To validate the proposed approach, we conducted experiments on several indoor locations.

Indoor pedestrian tracking extends location-based services to indoor environments. Typical indoor positioning systems employ a training/positioning model using Wi-Fi fingerprints. While these approaches have practical results in terms of accuracy and coverage, they require an indoor map, which is typically not available to the average user and involves significant training costs. A practical indoor pedestrian tracking approach should consider the indoor environment without a pretrained database or floor plan. In this paper, we present an indoor pedestrian tracking system, called SmartSLAM, which automatically constructs an indoor floor plan and radio fingerprint map for anonymous buildings using a smartphone. The scheme employs odometry tracing using inertial sensors, an observation model using Wi-Fi signals, and a Bayesian estimation for floor-plan construction. SmartSLAM is a true simultaneous localization and mapping implementation that does not necessitate additional devices, such as laser rangefinders or wheel encoders. We implemented the scheme on off-the-shelf smartphones and evaluated the performance in our university buildings. Despite inherent tracking errors from noisy sensors, SmartSLAM successfully constructed indoor floor plans.

With the widespread use of smartphones, the loss of a device is critical, both in disrupting daily communications, and in losing valuable property. When a mobile device is missing, localization techniques may assist in finding the device. Current techniques, however, hardly provide a complete solution because of inaccurate position estimation, especially in indoor environments. In this paper, we describe a software architecture called FindingMiMo, which tracks and locates a missing mobile device in indoor environments. The system consists of a missing mobile which logs diverse environmental features on a daily basis, and a chaser which traces the trail of the device using the observation log. During daily operation, the mobile device does not perform location estimation; it only observes the ambient features such as radio signals to minimize its operation cost. Instead, the chaser determines where the missing device measured the observations. This research implemented the scheme on Android-based smartphones. Real experiments with carefully designed, missing-and-tracking scenarios show that the participants successfully approached their lost phones within four meters distance, on average.

Monitoring a user's mobility during daily life is an essential requirement in providing advanced mobile services. While extensive attempts have been made to monitor user mobility, previous work has rarely addressed issues with battery lifetime in real deployment. In this paper, we introduce SmartDC, a mobility prediction-based adaptive duty cycling scheme to provide contextual information about a user's mobility: time-resolved places and paths. Unlike previous approaches that focused on minimizing energy consumption for tracking raw coordinates, we propose efficient techniques to maximize the accuracy of monitoring meaningful places with a given energy constraint. SmartDC comprises unsupervised mobility learner, mobility predictor, and Markov decision process-based adaptive duty cycling. SmartDC estimates the regularity of individual mobility and predicts residence time at places to determine efficient sensing schedules. Our experiment results show that SmartDC consumes 81% less energy than the periodic sensing schemes, and 87% less energy than a scheme employing context-aware sensing, yet it still correctly monitors 80% of a user's location changes within a 160-second delay.

Available techniques for indoor object locating systems, such as inertial sensor-based system or radio fingerprinting, hardly satisfy both cost-effectiveness and accuracy. In particular, inertial sensor-based locating systems are often supplemented with radio signals to improve localization accuracy. A radio-assisted localization system is still costly due to the infrastructure requirements and management overheads. In this paper, we propose a low-cost and yet accurate indoor pedestrian localization scheme with a small number of radio beacons whose location information is unknown. Our scheme applies the Simultaneous Location and Mapping (SLAM) technique used in robotics to mobile device, which is equipped with both inertial sensors and the IEEE802.15.4a Chirp Spread Spectrum (CSS) radio, to obtain accurate locations of pedestrians in indoor environment. The proposed system is validated with real implementations. The experiment results show approximately 1.5 m mean error observed during 276 m of pedestrian moving in a 380 m2indoor environment with five position-unknown beacons.

Estimating the geographical position of mobile device such as Smart phone in an indoor environment is not easy without the use of specific infrastructures. In this article, we introduce an indoor user tracking system. The system constructs a topological map with Wi-Fi signal calibrations, assigns semantically meaningful labels into the map, and estimates the semantic location of the user based on the current Wi-Fi observation. The system does not require a geometric map, or costly radio map building process. We implemented the system with the off-the-shelf Smart phone and experimentally validated the scheme.

This paper presents a run-time monitoring framework to detect end-to-end timing constraint violations of event flows in distributed real-time systems. The framework analyzes every event on possible event flow paths and automatically inserts timing fault checks for run-time detection. When the framework detects a timing violation, it provides users with the event flow's run-time path and the time consumption of each participating software module. In addition, it invokes a timing fault handler according to the timing fault specification, which allows our approach to aid the monitoring and management of the deployed systems. The experimental results show that the framework correctly detects timing constraint with insignificant overhead and provides related diagnostic information.

Processor power management in handheld devices is the primary technique for exploiting power reduction while ensuring performance. Modern mobile devices require high performance at the system level to decode high-bitrate multimedia. For this reason, processor offloading using off-chip controllers is commonly exercised in this field. However, current power management techniques do not fully consider the offloading architecture. We propose a scheme to achieve power reduction through an empirical method, which detects and classifies off-chip usages, in addition to combining dynamic voltage scaling (DVS) with dynamic power management (DPM). We experimented with the proposed technique in a real hardware environment and achieved up to a 37% power reduction compared with previous schemes.

Abstract Network reprogramming is a process used to update program codes of sensor nodes that are already deployed. To deal with potentially unstable link conditions of wireless sensor networks, the epidemic approach based on 3‐way advertise‐request‐data handshaking is preferred for network reprogramming. Existing epidemic protocols, however, require a long completion period and high traffic overhead in high‐density networks, mainly due to the hidden terminal problem. In this paper, we address this problem by dynamically adjusting the frequency of advertisement messages in terms of the density of sensor nodes, which is the number of sensor nodes in a certain area. We compare the performance of the proposed scheme, called DANP (Density‐Adaptive Network Reprogramming Protocol), with a well‐known epidemic protocol, Deluge. Simulations indicate that, in the grid topologies, DANP outperforms Deluge by about 30% in terms of the completion time and about 50% in terms of the traffic overhead. Significant performance gain is observed in random topologies as well. The performance of DANP is further confirmed via measurements in an experimental test bed. Copyright © 2009 John Wiley & Sons, Ltd.

We propose a 6LoWPAN-SNMP that enables transmission of SNMP messages over IPv6-enabled low-power wireless personal area networks (6LoWPAN). The 6LoWPAN-SNMP is an extended modification of the Simple Network Management Protocol (SNMP). Compared to traditional IP-networks, 6LoWPAN is a severely resource-constrained network; hence, existing SNMP protocols need to be modified to meet the goals in RFC 4919, ldquo6LoWPAN Problems and Goalsrdquo. The proposed 6LoWPAN-SNMP provides for native communication of SNMP messages on LoWPAN networks. The proposed mechanism is resource-efficient and fully compatible with the standard SNMP. It utilizes SNMP header compression and provides extended protocol operations to reduce the number of SNMP messages generated among the SNMP entities. Compatibility with the standard SNMP is achieved by a proxy forwarder on the 6LoWPAN gateway. The proposed mechanism is implemented on actual hardware platforms using the open-source Net-SNMP library and the Berkeley 6LoWPAN on the TinyOS 2.1. The feasibility and effectiveness of 6LoWPAN-SNMP is evaluated. The experimental results show that 6LoWPAN can be effectively supported with network management functionality based on SNMP.

As resource-constrained network technology develops, such as wireless sensor networks, connectivity to an IP-based network has become an important requirement. Assigning the global unique address to network nodes is a prerequisite to the connectivity to the IP-based networks. Since conventional address auto-configuration protocols require high network bandwidth and management cost, they are not suitable for wireless sensor networks. In this paper, we propose a lightweight address auto-configuration mechanism for resource-constrained networks. The proposed algorithm uses three coordinators that assign geometric information to the network to remove the assumption that each node has location information. Each node gathers the hop distance from the coordinators and generates a unique address based on the location information. The proposed algorithm is implemented with real hardware, and the performance is evaluated. The result shows that the mechanism efficiently assigns unique addresses to sensor nodes.

Discovering the location of the mobile nodes carried by people is important issue for many sensor applications. Several localization techniques have been proposed, but human mobility patterns and collaboration between mobile nodes have been seldom considered. In this paper, we propose a mobile node localization system based on collaboration and route information that characterizes human mobility. To validate the feasibility of our approach, the proposed system is implemented and experiments are conducted on real routes and to evaluate various scenarios, simulation experiment was also conducted.

We show a mobile monitoring application, called Sensible Doctor, for sensor network diagnosis. The solution is applied to a restricted region of the network, which surrounds the mobile monitoring device. The network traffic is reduced to cover only the monitoring part of the network by adopting algorithms tailored to mobile monitoring. Sensible Doctor is implemented on the RETOS operating system and provides system information, system profiling, remote application installation and network topology monitoring.

As the use of wireless sensor networks (WSNs) becomes widespread, node density tends to increase. This poses a new challenge for medium access control (MAC) protocol design. Although traditional MAC protocols achieve low-power operation, they use only a single channel which limits their performance. Several multi-channel MAC protocols for WSNs have been recently proposed. One of the key observations is that these protocols are less energy efficient than single-channel MAC protocols under light traffic conditions. In this paper, we propose an energy efficient multichannel MAC protocol, Y-MAC, for WSNs. Our goal is to achieve both high performance and energy efficiency under diverse traffic conditions. In contrast to most of previous multi-channel MAC protocols for WSNs, we implemented Y-MAC on a real sensor node platform and conducted extensive experiments to evaluate its performance. Experimental results show that Y-MAC is energy efficient and maintains high performance under high-traffic conditions.

Power management of sensor nodes is essential to maximize the lifetime of wireless sensor networks. Energy consumption of individual sensor node differs greatly depending on its type. Existing operating systems for sensor networks typically leave the power management of sensors to applications. Unless specific features of the sensors and batteries are considered in the application design, energy consumption may increase or the application may malfunction. In this paper, we define factors to be considered for sensor-specific power management, and then propose an automated sensor-specific power management system. The system is implemented on the multi-threaded RETOS operating system for sensor networks. The experiment results with various applications show that the proposed system reduced the energy consumption by up to 29% without modification of the application code.

Recently the demand on high graphic ability has increased in handheld systems due to multimedia or game application. Conventional DVS (dynamic voltage scaling) methods focus on reducing energy consumption of CPU. They didn't consider the influence of multimedia accelerators to make DVS policies. From the preliminary experiment, we found that the conventional DVS algorithm is not optimal to a system with a multimedia accelerator in context of reducing energy consumption of the system. We propose a power management mechanism that guarantees both QoS and reduction of energy consumption by considering the relation between the frequencies of the CPU and the frequencies of the multimedia accelerator. The proposed mechanism is a DVS technique in context of multimedia accelerator. We experimented the proposed mechanism on a testbed equipped with Intel XScale processor and Intel 2700 G multimedia accelerator. The experimental result shows that the proposed method reduces energy consumption of the system as much as by 33% compared to conventional CPU-only DVS algorithms.

One interesting issue in recent sensor network research is the application of sensor nodes to mobile objects, such as people or animals, to gather non-stationary information. Apparently, locating mobile nodes is essential in many of the mobile sensor applications. Although several localization techniques have been proposed to locate mobile nodes, the mechanisms are usually based on many impractical assumptions, and mobility pattern of the mobile object is seldom considered. In this paper, we propose a practical and yet simple mobile node localization system using nearby anchor information which contains absolute time and positions. A history of anchor information is used to characterize the mobility of mobile devices. The unknown node then calculates its position with the archived anchor using a regression model. The simulation results show that the proposed algorithm outperforms previous methods under realistic conditions. We have also implemented the algorithm on real hardware, and diverse experiments were conducted to validate the feasibility of the proposed approach.

Abstract To support Quality of service (QoS)‐sensitive applications like real‐time video streaming in IEEE 802.11 networks, a MAC layer extension for QoS, IEEE 802.11e, has been recently ratified as a standard. This MAC layer solution, however, addresses only the issue of prioritized access to the wireless medium and leaves such issues as QoS guarantee and admission control to the traffic control systems at the higher layers. This paper presents an IP‐layer traffic control system for IEEE 802.11 networks based on available bandwidth estimation. We build an analytical model for estimating the available bandwidth by extending an existing throughput computation model, and implement a traffic control system that provides QoS guarantees and admission control by utilizing the estimated available bandwidth information. We have conducted extensive performance evaluation of the proposed scheme via both simulations and measurements in the real test‐bed. The experiment results show that our estimation model and traffic control system work accurately and effectively in various network load conditions without IEEE 802.11e. The presence of IEEE 802.11e will allow even more efficient QoS provision, as the proposed scheme and the MAC layer QoS support will complement each other. Copyright © 2006 John Wiley & Sons, Ltd.

This paper introduces a component-based network management system for wireless sensor networks. The research is motivated by a variety of problems that occur due to the unpredictable behavior of wireless sensor networks. Once sensor nodes are deployed, the inner behavior of a sensor field can only be analyzed by monitoring incoming data packets at the sink node or base station, if only such a framework exists. In this paper, we present a component-based management system, RMTool, which allows developers to easily monitor and analyze the network status and interactively configure the network over unexpected problems while running applications over it. The system has been implemented on a multi-threaded sensor network operating system RETOS, which supports run-time loadable kernel modules. The preliminary evaluation shows that RMTool provides management functionalities as designed, and the implementation is efficient due to the component-based module architecture.

Previous work for maximizing system lifetime in sensor networks has proposed mechanism such on clustered networks or energy-aware routing protocols. In those works, prediction of lifetime and measurement of energy are dependent on predefined data such as transmission energy, data rate and so on. In common sensor networks, packets are unevenly generated and transmission energy is difficult to measure. Previous work cannot reflect the attributes of real networks. In this paper, we propose a routing algorithm that considers hardware-dependent battery consumption behavior of underlying sensor hardware. Each node reads its current voltage level and estimates its lifetime. Routing is then decided upon the predicted lifetime. Simulation results show that the proposed protocol extends the network lifetime by more than 12% over traditional methods.

This paper presents the design principles, implementation, and evaluation of the RETOS operating system which is specifically developed for micro sensor nodes. RETOS has four distinct objectives, which are to provide (1) a multithreaded programming interface, (2) system resiliency, (3) kernel extensibility with dynamic reconfiguration, and (4) WSN-oriented network abstraction. RETOS is a multithreaded operating system, hence it provides the commonly used thread model of programming interface to developers. We have used various implementation techniques to optimize the performance and resource usage of multithreading. RETOS also provides software solutions to separate kernel from user applications, and supports their robust execution on MMU-less hardware. The RETOS kernel can be dynamically reconfigured, via loadable kernel framework, so a application-optimized and resource-efficient kernel is constructed. Finally, the networking architecture in RETOS is designed with a layering concept to provide WSN-specific network abstraction. RETOS currently supports Atmel ATmega128, TI MSP430, and Chipcon CC2430 family of microcontrollers. Several real-world WSN applications are developed for RETOS and the overall evaluation of the systems is described in the paper.

This demonstration shows the programming development suite of the RETOS operating system for sensor networks, which provides a robust and multithreaded programming interface to application programmers. We first demonstrate how to build the RETOS kernel on the TI MSP430, Atmel ATmega 128 and Chipcon CC2430 family of microcontrollers. The application or a kernel module is then compiled and disseminated, via wireless channel, to the target motes. The GUI-based RMon network management tool for RETOS is also demonstrated to monitor the networked sensors, and even to control the system's parameters or applications running on them via a remote shell. The system is demonstrated to run on a mixed set of MSP430, ATmega128 and CC2430-based motes. Overall, our demonstration will convince attendees of the programming convenience of developing sensor network applications using RETOS, which is, indeed, a mature and practical system that can be used to develop real-world applications.

Recently power management of display modules has become important for handheld systems. In this paper, we observe the interaction between the LCD, main consumer of the power in handheld systems, and the CPU. We propose an efficient method which reduces power consumption of the handheld system by controlling the voltage and frequency of the CPU to reduce display power. The experimental result on Mainstone II with PXA270 shows that the proposed mechanism reduces power consumption of the system by 41% and 15% for multimedia applications and web applications, respectively, when compared to the conventional DVS (Dynamic Voltage Scaling) scheme.

Sensor nodes are typically battery powered and operate in unattended environments. Minimizing the energy consumption of sensor nodes is, therefore, important to prolong the network life time. Since the radio is a main energy consumer, most of the Medium Access Control (MAC) protocols for wireless sensor networks (WSNs) have focused on energy efficiency. With widespread use of sensor network applications, network bandwidth required by the applications tends to increase and even multiple applications may run in a sensor node. For these reasons, maximizing the bandwidth of WSNs has become a major consideration in MAC design.

This paper discusses the limitations of implementing low power listening (LPL) on the contention-based IEEE 802.15.4 MAC protocol, and proposes an implementation technique called virtual preamble cross-checking (VPCC), which enables LPL. The proposed technique uses the concept of virtual preamble and a cross-checking method to meet the requirements of the relationship between preamble length and listening interval. These methods lead to reliable data communication, as well as the reduction of the idle listening. We implemented the mechanism on IEEE 802.15.4-compatible RF hardware and validated the performance. The experiment results show that increasing the listening interval within a tolerable latency can optimize the duty cycle of the RF transceiver for low power consumption.

Compared to a centralized system for acoustic source localization using WSN, a distributed system has many advantages in terms of scalability, power consumption, and response time. Complexity is an important factor to consider when implementing a distributed system because of the limited resource of WSN devices. The complexity of the source localization algorithm is related to the area required for location estimation. Since reducing the size of the estimation area influences the accuracy of the source localization system, a policy should be carefully designed to reduce the size. In this paper, the authors use a Voronoi diagram to define an estimation area. The Voronoi diagram-based mechanism may yield an optimized estimation area and accuracy, but creating a Voronoi diagram is not suitable for WSN devices due to the high implementation overhead. This paper proposes an efficient mechanism to create the estimation area, which can be implemented in WSN devices and achieve reasonable accuracy, compared to the ideal Voronoi diagram-based one. The proposed algorithm has been applied to a distributed source localization algorithm, and compared with a Voronoi diagram through extensive simulations.

Abstract In recent years, there has been wide‐spread use of large and high‐resolution liquid‐crystal displays (LCDs) on handheld devices. The portion of LCD power consumption in the overall system has gradually increased. While most of the previous research on LCD power management has focused on the hardware level, practical mechanisms at the software level are hardly known. This paper presents a power‐aware LCD management mechanism, based on dynamic refresh‐rate scaling and frame buffer monitoring. The proposed mechanism guarantees the display quality of service, which is inherently specified by content types. The mechanism does not require additional hardware or modifications to applications. The experiment results—on a commercial PDA with a $320\times240$ resolution—show that the proposed mechanisms effectively reduce the power consumption by up to 10%, while satisfying the display quality requirements for the LCD screen. Copyright © 2006 John Wiley & Sons, Ltd.

Wireless sensor network technology has found diverse applications in numerous fields. As the networking technology is refined in many ways, the need for system modulation with effective performance becomes essential. A multitude of architectures, which includes system abstraction and layering, has been proposed to solve the need at the operating system level. However, previous efforts do not qualify for networking architecture required by sensor networking, since they are aimed at hardware abstraction or protocol-based layering. In this paper, we classify developers into kernel, network and application developers and propose a network architecture that enables those developers to program independently. Network stack is separated into three different layers; MLL, NSL, DNL. This three-layered architecture provides an effective programming environment to sensor network developers by minimizing modification of other layers and maximizing reusability of the networking module. To validate the proposed mechanism, we implemented and assessed the performance with a few network algorithms and applications, based on the RETOS, which supports a dynamic loadable kernel module

Current tracking applications in sensor networks lack bi-directional communications with mobile objects. They have a passive communication relationship with unidirectional location data messages routed from mobile objects to the central processing node. In order for an object tracking system to be an "active" tracking system, where the user of the system is able to control the mobile objects in various ways, the system first needs an efficient way to deliver control messages to a mobile object, other than a simple flooding of the control messages. This paper presents an efficient detection method for multiple mobile nodes, which yields a reduction in the number of flooding messages by adopting the hybrid solution of combining a fill-scale flooding and a constrained flooding for the delivery of control messages. Simulation shows that our solution provides about 90% to 10% of reduction on both energy consumption and network traffic comparing to full-scale flooding when the number of mobile nodes increases by 1 to 9.

A popular web site is simultaneously approached by many clients and therefore it is vital to schedule the connections efficiently in or- der to improve the system performance. SRPT (Shortest Remaining Pro- cessing Time first) improves the response time of each connection. The new feature of HTTP/1.1 is to enable the clients to request multiple inlined- documents in a HTML document. When multiple inlined-documents are requested per connection, the response time of each inlined-document may be improved by the connection scheduling based on SRPT. It cannot, how- ever, guarantee that the response time of each connection with multiple inlined-documents is improved. In this paper we propose a window-based connection scheduling technique which provides a better response time in HTTP/1.1. The experimental results show that the performance with the proposed approach is improved.

This paper presents the implementation details and experiment results of the IPME QoS network management system which is specifically designed to accommodate time-critical multimedia applications in multicast network environments. As the multimedia and computer communication techniques develop rapidly, there has been growing research interest in effective management of multicast services. When applying a multicast network to multimedia applications, an efficient handling of time-critical media and network resources becomes crucial to the success of multicast services since the underlying communication structure may get complicated. Managing a multimedia communication system generally requires a sophisticated model which accommodates many techniques on efficient media transport, adequate QoS provision and effective network management. In order to provide QoS for multimedia applications in a multicast network environment, in particular, we believe that it is important to establish a QoS-oriented network management model which should specifically be designed for multicast networks. Although the necessity of such a model has been acknowledged for some time, there has been little research effort to integrate multimedia QoS from the aspect of multicast network management.

The paper presents a parallel computing model, called H-BSP, which adds a hierarchical concept to the BSP (Bulk Synchronous Parallel) computing model. A H-BSP program consists of a number of BSP groups which are dynamically created at run time and executed in a hierarchical fashion. H-BSP allows the algorithm designer to develop a more efficient algorithm by utilizing processor locality in the program. The paper describes the structure of the H-BSP model, complexity analysis and an example of the H-BSP algorithm. Also presented are the performance characteristics of the H-BSP algorithm based on simulation analysis. Simulation results show that H-BSP model takes advantages of processor locality and performs well in low bandwidth networks or in a constant valence architecture such as a 2 dimensional mesh. It is also proved that H-BSP model can predict algorithm performance better than the BSP model due to its locality preserving nature.

The performance of message passing architecture machines is often ameliorated by devising alternative message passing strategies.A number have been proposed for incorporation into the design of architectures for massively parallel machines.Among these approaches are (i) static machine topologies supporting packet switched message routing (store-and-forward)(ii) dynamic machine topologies supporting circuit switched routing and (iii) shared memory systems.This paper reports the results of analytical modelling of these strategies -verifying the results by experimentation on a large reconfigurable transputer machine (the ParSiFal T-Rack).A family of hybrid architectures called the Hyper-Dynavnic Architecture is then proposed which incorporates both packet switched and circuit switched features.Extrapolation of the verified analytical models and various simulation results obtained under realistic environments then allow some prediction of the potential performance of such architectures. 1