Emerging IoT applications are transitioning from battery-powered to grid-powered nodes. DRP, a contention-based data dissemination protocol, was developed for these applications. Traditional contention-based protocols resolve collisions through control packet exchanges, significantly reducing goodput. DRP mitigates this issue by employing a distributed delay timer mechanism that assigns transmission-start delays based on the average link quality between a sender and its children, prioritizing highly connected nodes for early transmission. However, our in-field experiments reveal that DRP is unable to accommodate real-world link quality fluctuations, leading to overlapping transmissions from multiple senders. This overlap triggers CSMA’s random back-off delays, ultimately degrading the goodput performance. To address these shortcomings, we first conduct a theoretical analysis that characterizes the design requirements induced by real-world link quality fluctuations and DRP’s passive acknowledgments. Guided by this analysis, we design EDRP, which integrates two novel components: (i) Link-Quality Aware CSMA (LQ-CSMA) and (ii) a Machine Learning-based Block Size Selection (ML-BSS) algorithm for rateless codes. LQ-CSMA dynamically restricts the back-off delay range based on real-time link quality estimates, ensuring that nodes with stronger connectivity experience shorter delays. ML-BSS algorithm predicts future link quality conditions and optimally adjusts the block size for rateless coding, reducing overhead and enhancing goodput. In-field evaluations of EDRP demonstrate an average goodput improvement of 39.43% than the competing protocols.

Managing a vast number of virtual machines (VMs) efficiently is a critical challenge in modern large-scale data centers. The continuous creation and termination of VMs lead to resource fragmentation across physical machines (PMs), necessitating periodic VM rescheduling to optimize resource utilization. Despite its significance, VM rescheduling has received limited attention in the literature. A key challenge is that, unlike conventional combinatorial optimization problems, the efficiency of rescheduling algorithms is heavily impacted by inference time, as VM states evolve dynamically during execution. This scalability bottleneck hampers existing methods. To address this, we propose VMR$^{2}$L, a reinforcement learning framework tailored for VM rescheduling. VMR$^{2}$L integrates a two-stage decision-making process to accommodate complex operational constraints, a feature extraction mechanism that captures critical relational information for rescheduling, and a risk-aware evaluation strategy that enables users to balance execution speed and rescheduling accuracy. Extensive experiments using real-world data from a production-scale data center demonstrate that VMR$^{2}$L achieves near-optimal performance while reducing inference time to a matter of seconds. To facilitate reproducibility, we provide access to our implementation and datasets.

Mesoscale IoT applications, such as P2P energy trade and real-time industrial control systems, demand high throughput and low latency, with a secondary emphasis on energy efficiency as they rely on grid power or large-capacity batteries. MARS, a multi-radio architecture, leverages ML to instantaneously select the optimal radio for transmission, outperforming single-radio systems. However, MARS encounters a significant issue with cost sensitivity, where high-cost errors account for 40% throughput loss. Current cost-sensitive ML algorithms assign a misclassification cost for each class, but not for each data sample. In MARS, each data sample has different costs, making it tedious to employ existing cost-sensitive ML algorithms. First, we address this issue by developing COMNETS, an ML-based radio selector using oblique trees optimized by (TAO). TAO incorporates sample-specific misclassification costs to avert high-cost errors, and achieves a 50% reduction in the decision tree size, making it more suitable for resource-constrained IoT devices. Second, we prove the stability property of TAO and leverage it to understand the critical factors affecting the radio-selection problem. Finally, our real-world evaluation of COMNETS at two different locations shows an average throughput gain of 20.83%, 17.39% than MARS.

IoT is rapidly expanding from traditional small-scale (0–100m) applications like smart homes and large-scale (1–5km) applications like Microsoft’s FarmBeats to emerging mesoscale (0.1–1.5km) applications such as smart-grid NANs and peer-to-peer energy trading in smart homes. These applications demand high throughput and low latency but currently lack dedicated radio technologies. Our qualitative analysis identified Zigbee and LoRa as promising candidates. Further quantitative analysis revealed that a multi-radio architecture combining these radios achieves the best throughput. However, within the 500–1200m range, termed the gray region, it is unpredictable which radio offers higher throughput at any given moment. To address this, we developed MARS, a Multi-radio Architecture with Radio Selection, powered by TAO-optimized decision trees that select the high-throughput radio at the time of transmission. These decision trees require instantaneous path quality estimates, but traditional multi-hop Zigbee networks cannot provide these promptly due to propagation and queuing delays. We overcome this challenge by introducing Decision Tree-based updates to instantaneously estimate end-to-end path quality. Large-scale, real-world experiments with MARS demonstrated average throughput gains of 48.2% and 49.79% at two different locations.

Modern industry-scale data centers need to manage a large number of virtual machines (VMs). Due to the continual creation and release of VMs, many small resource fragments are scattered across physical machines (PMs). To handle these fragments, data centers periodically reschedule some VMs to alternative PMs, a practice commonly referred to as VM rescheduling. Despite the increasing importance of VM rescheduling as data centers grow in size, the problem remains understudied. We first show that, unlike most combinatorial optimization tasks, the inference time of VM rescheduling algorithms significantly influences their performance, due to dynamic VM state changes during this period. This causes existing methods to scale poorly. Therefore, we develop a reinforcement learning system for VM rescheduling, VMR2L, which incorporates a set of customized techniques, such as a two-stage framework that accommodates diverse constraints and workload conditions, a feature extraction module that captures relational information specific to rescheduling, as well as a risk-seeking evaluation enabling users to optimize the trade-off between latency and accuracy. We conduct extensive experiments with data from an industry-scale data center. Our results show that VMR2L can achieve a performance comparable to the optimal solution but with a running time of seconds. Code12 and datasets3 are open-sourced.

The application of reinforcement learning in controlling Heating, Ventilation, and Air Conditioning (HVAC) systems has been extensively researched. Existing studies primarily focus on Model-Free Reinforcement Learning (MFRL), which involves trial-and-error interactions with real buildings to train the agent. However, MFRL encounters a significant challenge: it requires a large amount of training data to achieve satisfactory performance. While simulation models have been used to generate training data and expedite the training process, they necessitate high-fidelity building models that are difficult to calibrate. As a result, Model-Based Reinforcement Learning (MBRL) has been employed for HVAC control. Although MBRL demonstrates remarkable sample efficiency, it often falls short in terms of asymptotic control performance, particularly in achieving substantial energy savings while ensuring occupants’ thermal comfort. In this study, we conduct experiments to analyze the limitations of current MBRL-based HVAC control methods, focusing on model uncertainty and controller effectiveness. Leveraging the insights gained from these experiments, we develop MB2C, an innovative MBRL-based HVAC control system that combines high control performance with exceptional sample efficiency. MB2C learns the dynamics of the building by employing an ensemble of environment-conditioned neural networks and utilizes a novel control method called Model Predictive Path Integral (MPPI) for HVAC control. MPPI generates candidate action sequences using an importance sampling weighted algorithm, which is well-suited for multi-zone buildings with high state and action dimensions. We evaluate MB2C using EnergyPlus simulations in a five-zone office building, and the results demonstrate that MB2C achieves 8.23% higher energy savings compared to the state-of-the-art MBRL solution while maintaining comparable thermal comfort. Moreover, MB2C significantly reduces the required training data set by an order of magnitude ($10.52\times $) while delivering performance on par with MFRL approaches. Note to Practitioners—Our research addresses a critical challenge in HVAC control, offering an innovative solution to enhance the data efficiency of HVAC systems while optimizing energy usage. Traditional approaches, such as Model-Free Reinforcement Learning, often require a large volume of real-world data. Our primary focus is improving the effectiveness of HVAC control, a vital aspect of building management that directly affects energy consumption and occupant well-being. We introduce MB2C, a Model-Based Reinforcement Learning system designed to significantly improve energy savings while maintaining thermal comfort. MB2C achieves remarkable results, offering exceptional sample efficiency and substantially reducing the required training data. Our research leverages an ensemble of environment-conditioned neural networks and employs Model Predictive Path Integral in HVAC control. While MB2C presents notable benefits, it also has limitations. Further research and development are required to optimize its performance across different building environments and specific use cases. Future directions should focus on addressing the safety challenges associated with real-world deployment. Beyond HVAC control, the principles and methods explored in this research have potential applications in various automation domains, such as robotics, industrial automation, and manufacturing processes.

In recent years, the focus has been on enhancing user comfort in commercial buildings while cutting energy costs. Efforts have mainly centered on improving HVAC systems, the central control system. However, it’s evident that HVAC alone can’t ensure occupant comfort. Lighting, blinds, and windows, often overlooked, also impact energy use and comfort. This paper introduces a holistic approach to managing the delicate balance between energy efficiency and occupant comfort in commercial buildings. We present OCTOPUS , a system employing a deep reinforcement learning (DRL) framework using data-driven techniques to optimize control sequences for all building subsystems, including HVAC, lighting, blinds, and windows. OCTOPUS ’s DRL architecture features a unique reward function facilitating the exploration of tradeoffs between energy usage and user comfort, effectively addressing the high-dimensional control problem resulting from interactions among these four building subsystems. To meet data training requirements, we emphasize the importance of calibrated simulations that closely replicate target-building operational conditions. We train OCTOPUS using 10-year weather data and a calibrated building model in the EnergyPlus simulator. Extensive simulations demonstrate that OCTOPUS achieves substantial energy savings, outperforming state-of-the-art rule-based and DRL-based methods by 14.26% and 8.1%, respectively, in a LEED Gold Certified building while maintaining desired human comfort levels.

Buildings are responsible for a significant portion of energy consumption in the US, accounting for more than 40% of US primary energy consumption. Heating, ventilation and air-conditioning (HVAC) accounts for nearly 50% of that use. Conditioning buildings is important since people spend 87% of their time in the place they live (residential) and the place they work (commercial). Despite this massive expense, many users are dissatisfied with the thermal conditions in buildings. Savings made in HVAC systems, therefore, have a major impact on energy consumption and cost, together with the reduction of greenhouse emissions for the nation. Equally critical is to provide thermal quality of service to their users, so people are comfortable in the place they reside and work.In this paper, we explore the tradeoff between commercial building HVAC energy consumption and the quality of thermal conditioning provided to users. We argue that optimal HVAC control cannot be achieved due to lack of critical information, namely where the users are inside the building, what do they want with respect to thermal comfort and how each zone responds to thermal changes. In this work, we present OFFICE, a model predictive control (MPC) framework for smart building HVAC control. The framework has several components that help to address the current HVAC control systems shortcomings, including (a) occupancy sensing in real-time, (b) occupancy prediction models based on historical occupancy data, (c) human-in-the-loop comfort feedback, (d) data-driven thermodynamic building models, and (e) weather forecasting data. All these components provide the necessary input to our model predictive control optimization framework that minimizes monetary costs in energy use while maintaining quality comfort bounds for the building’s users based on real-time user’s feedback. We developed a large system that involves all the above components, replacing the Building Management System control algorithms, taking over full control of the HVAC system. We tested OFFICE in a real LEED Gold certified university building with over 20 workers performing their daily tasks for 4 weeks, and we showed that we could obtain monetary costs savings of more than 10% while at the same time reducing the users’ dissatisfaction levels with thermal comfort from 25% to 0% dissatisfaction, significantly improving the quality of thermal service provided to the building’s users.

Lawns, also known as turf, cover an estimated 128,000 km 2 in North America alone, with landscape requirements representing 30% of freshwater consumed in the residential domain. With this consumption comes a large amount of environmental, economic, and social incentive to make turf irrigation systems as efficient as possible. Recent work introduced the concept of distributed control in irrigation systems, but existing control strategies either do not take advantage of the distributed control, or do not revise the strategy over time in response to collected data. In this work, we introduce OPTICS, a data-driven control strategy that self-improves over time, adapts to the local specific conditions and weather changes, and requires virtually no human input in both setup and maintenance providing a plug-and-play system that requires minimal pre-deployment efforts. In addition to substantial improvements in ease-of-use, we find across 4 weeks of large-scale irrigation system deployment that OPTICS improves system efficiency by 12.0% in comparison to industry best and 3.3% in comparison to academic state of the art. Despite using less water, OPTICS also was found to improve quality of service by a factor of 4.0× compared to industry best and 2.5× compared to academic state of the art.

As lawn irrigation is estimated to consume 7 billion gallons of scarce fresh water each day in North America alone, lawn irrigation systems are a high priority for improvements in efficiency. To this end, recent work has introduced several key advancements in irrigation control. Distributed actuation systems allow the irrigation system to apply water completely independently across the field allowing flexibility of control, and the use of fluid flow modeling and optimization allows more efficient schedules to be computed automatically, significantly improving the irrigation quality of service as well. However, the proposed systems are designed with centralized architectures that introduce single points of failure, computational bottlenecks in data processing, and significant network energy for data forwarding used by the centralized data-driven modeling strategies. In response to these challenges, we propose and demonstrate WISDOM, whose novel and flexible hardware and processing pipelines enable the use of a distributed system for the management of irrigation systems of any scale, with energy independence by way of energy harvesting. Across 4 weeks of live system deployment, we find that the WISDOM system can save up to 32.9% of water in comparison to industry-best, while maintaining a perfect quality of service to the plant. Furthermore, with substantial analysis in simulation we find that in addition to practical system improvements, the use of the proposed distributed system within typical operating conditions will provide all of the efficiency and quality-of-service benefits of the globally-modeled, centrally controlled systems, while allowing the robust control of irrigation systems of any size.

Lawns make up the largest irrigated crop by surface area in North America and carry with it a demand for over 7B gallons of freshwater each day. Despite recent developments in irrigation control and sprinkler technology, state-of-the-art irrigation systems do nothing to compensate for areas of turf with heterogeneous water needs. In this work, we overcome the physical limitations of the traditional irrigation system with the development of a sprinkler node that can sense the local soil moisture, communicate wirelessly, and actuate its own sprinkler based on a centrally computed schedule. A model is then developed to compute moisture movement from runoff, absorption, and diffusion. Integrated with an optimization framework, optimal valve scheduling can be found for each sprinkler node in the space. In a turf area covering over 10,000ft 2 , two separate deployments with four weeks of fine-grained data collection show that DICTUM can reduce water consumption by 23.4% over traditional campus scheduling, and by 12.3% over state-of-the-art evapotranspiration systems while substantially improving conditions for plant health. In addition to environmental, social, and health benefits, DICTUM is shown to return its investment in 16 to 18 months based on water consumption alone.

Lawns, also known as turf, cover an estimated 128,000km2[9] in North America alone, with landscape requirements representing 30% of freshwater consumed in the residential domain [27]. With this consumption comes a large amount of environmental, economic, and social incentive to make turf irrigation systems as efficient as possible. Recent work introduced the concept of distributed control in irrigation systems, but existing control strategies either do not take advantage of the distributed control, or don't revise the strategy over time in response to collected data. In this work, we introduce PICS, a data-driven control strategy that self-improves over time, adapts to the local specific conditions and weather changes, and requires virtually no human input in both setup and maintenance providing a plug-and-play system that requires minimal pre-deployment efforts. In addition to substantial improvements in ease-of-use, we find across 4 weeks of large-scale irrigation system deployment that PICS improves system efficiency by 12.0% in comparison to industry best and 3.3% in comparison to academic state-of-the-art. Despite using less water, PICS also was found to improve quality of service by a factor of 4.0x compared to industry best and 2.5x compared to academic state of the art.

Humans spend 90% of their lives inside buildings, but often the Heating, Ventilation, and Air Conditioning (HVAC) systems of commercial buildings do not properly maintain occupant comfort. Use of feedback through comfort voting applications has been shown to improve the quality of service, but the effects of application feedback and user interface design has not been investigated. In this work, we present several methods of feedback that use data presentation and environmental interaction in comfort voting applications. Through a 40 week user study of 61 University employees across 3 buildings, we show that feedback systems can be used to increase user satisfaction with thermal conditions from 33.9% to 93.3% and reduce energy consumption up to 18.99% compared to a system without voting. In addition, we find that by including a drifting control strategy, we find energy savings up to 37% can be realized without a significant reduction in satisfaction.

Lawns make up the largest irrigated crop by surface area in North America, and carries with it a demand for over 9 billion gallons of freshwater each day. Despite recent developments in irrigation control and sprinkler technology, state-of-the-art irrigation systems do nothing to compensate for areas of turf with heterogeneous water needs. In this work, we overcome the physical limitations of the traditional irrigation system with the development of a sprinkler node that can sense the local soil moisture, communicate wirelessly, and actuate its own sprinkler based on a centrally- computed schedule. A model is then developed to compute moisture movement from runoff, absorption, and diffusion. Integrated with an optimization framework, optimal valve scheduling can be found for each node in the space. In a turf area covering over 10,000ft2, two separate deployments spanning a total of 7 weeks show that MAGIC can reduce water consumption by 23.4% over traditional campus scheduling, and by 12.3% over state-of-the- art evapotranspiration systems, while substantially improving conditions for plant health. In addition to environmental, social, and health benefits, MAGIC is shown to return its investment in 16-18 months based on water consumption alone.

Buildings are responsible for more than 40% of US primary energy consumption. Of that, nearly 50% goes to heating, ventilation and air-conditioning (HVAC) systems. Savings made in HVAC systems therefore have a major impact in energy consumption and are of critical importance. We present a model predictive control framework (MPC) for smart building control. The framework has several components, including (a) occupancy sensing in real time, (b) occupancy prediction models based on historical occupancy data, (c) thermodynamic building models, (d) weather forecasting data, and (e) a model predictive control optimization that tries to minimize monetary costs in energy use while maintaining quality comfort bounds for the building's users. We tested the system in a real building with over 40 workers, and we show that we could obtain monetary savings of more than 50% without sacrificing user comfort.

We present a holistic architecture for energy management in sensor networks. Our architecture is based on a model-driven approach which attempts to (a) establish functional relationships across different components of the software stack and the interrelated parameters based on empirical data, (b) use the maximum sensor value and time-synchronization errors acceptable by the users of the sensor network application as input to establish minimum quality of service requirements, and (c) optimize the parameter values of all the software modules within the node's application stack to minimize total energy consumption for each sensor node. We explore the trade-offs of the design space by using a non-trivial application that includes sensing, time synchronization and routing modules and show that when using our architecture, we can provide energy savings in the average of 37% to 76% while still maintaining quality of service both in terms of the expected sensing and time-synchronization errors. We further show that even when using modules that perform significantly better than others with default values (e.g. ORW >> CTP), we can still reduce overall energy consumption by properly adjusting the parameters of lowest performance modules and provide energy savings in the average of 30% to 43%.

Lawns, both public and private, make up the largest irrigated crop in North America by surface area. Although there have been improvements in sprinkler head technology and weather assimilation, state-of-the-art irrigation systems do nothing to adjust for heterogeneous terrain or varying lawn environments. In this work, a computationally lightweight soil moisture movement model is developed, which allows the computation of optimal irrigation valve scheduling using standard optimization techniques. A prototype sprinkler head is produced with the ability to sense local soil moisture conditions, wirelessly communicate, and independently actuate based on the optimal schedule centrally computed. This prototype is then deployed to control two parallel irrigation systems covering a total of more than 10,000 ft$^2$ for a duration of 5 weeks. It is shown that lawn health can be maintained by using the topography of the space to take advantage of runoff to provide improved coverage while using an average of 23.4\% less water. We also show that the initial capital and operating costs of our system could be amortized by our water savings in $~$13 months while maintaining and/or improving quality of irrigation and lawn health.

Heating, cooling and ventilation accounts for 35% energy usage in the United States. Currently, most modern buildings still condition rooms assuming maximum occupancy rather than actual usage. As a result, rooms are often over-conditioned needlessly. Thus, in order to achieve efficient conditioning, we require knowledge of occupancy. This article shows how real time occupancy data from a wireless sensor network can be used to create occupancy models, which in turn can be integrated into building conditioning system for usage-based demand control conditioning strategies. Using strategies based on sensor network occupancy model predictions, we show that it is possible to achieve 42% annual energy savings while still maintaining American Society of Heating, Refrigerating and Air-Conditioning Engineers (ASHRAE) comfort standards.

As an integral part of reliable communication in wireless networks, effective link estimation is essential for routing protocols. However, due to the dynamic nature of wireless channels, accurate link quality estimation remains a challenging task. In this article, we propose 4C, a novel link estimator that applies link quality prediction along with link estimation. Our approach is data driven and consists of three steps: data collection, offline modeling, and online prediction. The data collection step involves gathering link quality data, and based on our analysis of the data, we propose a set of guidelines for the amount of data to be collected in our experimental scenarios. The modeling step includes offline prediction model training and selection. We present three prediction models that utilize different machine learning methods, namely, naive Bayes classifier, logistic regression, and artificial neural networks. Our models take a combination of PRR and the physical-layer information, that is, Received Signal Strength Indicator (RSSI), Signal-to-Noise Ratio (SNR), and Link Quality Indicator (LQI) as input, and output the success probability of delivering the next packet. From our analysis and experiments, we find that logistic regression works well among the three models with small computational cost. Finally, the third step involves the implementation of 4C, a receiver-initiated online link quality prediction module that computes the short temporal link quality. We conducted extensive experiments in the Motelab and our local indoor testbeds, as well as an outdoor deployment. Our results with single- and multiple-senders experiments show that with 4C, CTP improves the average cost of delivering a packet by 20% to 30%. In some cases, the improvement is larger than 45%.

Link quality estimation is a fundamental component of the low-power wireless network protocols and is essential for routing protocols in Wireless Sensor Networks (WSNs). However, accurate link quality estimation remains a challenging task due to the notoriously dynamic and unpredictable wireless environment. In this article we argue that, in addition to the estimation of current link quality, prediction of the future link quality is more important for the routing protocol to establish low-cost delivery paths. We propose to apply machine learning methods to predict the link quality in the near future to facilitate the utilization of intermediate links with frequent quality changes. Moreover, we show that, by using online learning methods, our adaptive link estimator (TALENT) adapts to network dynamics better than statically trained models without the need of a priori data collection for training the model before deployment. We implemented TALENT in TinyOS with Low-Power Listening (LPL) and conducted extensive experiments in three testbeds. Our experimental results show that the addition of TALENT increases the delivery efficiency 1.95 times on average compared with a 4B, state-of-the-art link quality estimator, as well as improves the end-to-end delivery rate when tested on three different wireless testbeds.

High-quality wireless link models can enable better simulations and reduce the development time for new algorithms and protocols. However, the models underlying current simulators are either based on too simple assumptions, so they are unrealistic, or are based on sophisticated machine learning techniques that require extensive training data from the target link, so they are more realistic but impractical. We consider the practical scenario where data collection time is limited (e.g. a few minutes) and cannot afford to deploy a testbed infrastructure with cabling, power and storage. We propose techniques that can construct an accurate machine learning model of the short-term behavior of a target wireless link given only limited training data for the latter, by adapting a reference model that was trained with abundant data. The parameters of the target model are a constrained transformation of the parameters of the reference model, thus the actual number of free parameters is much smaller, and can be reliably estimated with much less data. While estimating the target model from scratch requires 1 to 5 hours of target link data, we show our adaptation technique only requires under 3 minutes of data, for all packet reception rate regimes. We also show that we can construct adapted models for target links in different environments, packet sizes, interference conditions and radio technology (802.15.4 or 802.11b).

Buildings account for 40% of US primary energy consumption and 72% of electricity. Of this total, 50% of the energy consumed in buildings is used for Heating Ventilation and Air-Conditioning (HVAC) systems. Current HVAC systems only condition based on static schedules; rooms are conditioned regardless of occupancy. By conditioning rooms only when necessary, greater efficiency can be achieved. This paper describes POEM, a complete closed-loop system for optimally controlling HVAC systems in buildings based on actual occupancy levels. POEM is comprised of multiple parts. A wireless network of cameras called OPTNet is developed that functions as an optical turnstile to measure area/zone occupancies. Another wireless sensor network of passive infrared (PIR) sensors called BONet functions alongside OPTNet. This sensed occupancy data from both systems are then fused with an occupancy prediction model using a particle filter in order to determine the most accurate current occupancy in each zone in the building. Finally, the information from occupancy prediction models and current occupancy is combined in order to find the optimal conditioning strategy required to reach target temperatures and minimize ventilation requirements. Based on live tests of the system, we estimate ~30.0% energy saving can be achieved while still maintaining thermal comfort.

Buildings are often inefficiently conditioned. Rooms that are empty are needlessly conditioned and partially filled rooms are conditioned assuming maximum occupancy. In this demonstration, we describe a system that reduces energy consumption by opportunistically reducing energy consumption based on room usage; we only condition rooms currently occupied and condition the space based on real-time occupancy measurements. We will show how a thermal sensor array can be used measure occupancy in real-time and how this occupancy information can be integrated with a real building management system in order to control the heating, cooling, ventilation and lighting of a building to optimize energy usage.

Modeling the behavior of 802.15.4 links is a nontrivial problem, because 802.15.4 links experience different level of dynamics at short and long time scales. This makes the design of a suitable model that combines the different dynamics at different time scales a nontrivial problem. We propose a novel multilevel approach, the M&M model, involving hidden Markov models (HMMs) and mixtures of multivariate Bernoullis (MMBs) for modeling the long and short time-scale behavior of wireless links from 802.15.4 test beds. We characterize the synthetic traces generated from our model of the wireless link in terms of the mean and variance of the packet reception rates from the data traces, comparison of distributions of run lengths, and conditional packet delivery functions of successive packet receptions (1's) and losses (0's). Our results show that when compared to the closest-fit pattern matching model in TOSSIM, the proposed modeling approach is able to mimic the behavior of the data traces quite closely, with differences in packet reception rates of the empirical and simulated traces of less than 1.9%; on average and 6.6% in the worst case. Moreover, the simulated links from our proposed approach were able to account for long runs of 1's and 0's as observed in empirical data traces.

Link quality estimation is a fundamental component of the low power wireless network protocols and is essential for routing protocols in Wireless Sensor Networks (WSNs). However, accurate link quality estimation remains a challenging task due to the notoriously dynamic and unpredictable wireless environment. In this paper, we argue that in addition to the estimation of current link quality, prediction of the future link quality is more important for the routing protocol to establish low cost delivery paths. We propose to apply machine learning methods to predict the link quality in the near future to facilitate the utilization of intermediate links with frequent quality changes. Moreover, we show that by using online learning methods, our adaptive link estimator (TALENT) adapts to network dynamics better than statically trained models without the need of a priori data collection for training the model before deployment. We implemented TALENT in TinyOS with Low-Power Listening (LPL) and conducted extensive experiments in three testbeds. Our experimental results show that the addition of TALENT increases the delivery efficiency 1.95 times on average compared with 4B, the state of the art link quality estimator, as well as improve the end-to-end delivery rate when tested on three different wireless testbeds.

The lossy nature of wireless communication leads to many challenges while designing multihop networks. As an integral part of reliable communication in wireless networks, effective link estimation is essential for routing protocols. Recent studies have shown that link reliability-based metrics like ETX have better performance than traditional metrics such as hop count or latency. Usually, such metrics employ techniques like blacklisting, involving thresholds during the link estimation process. In this paper, we conduct a detailed performance analysis of three commonly used link-quality metrics in wireless sensor networks: ETX, 4Bit, and RNP. We study the interplay between these metrics and CTP, a tree-based routing protocol provided by TinyOS. The objectives of our experiment are two fold. First, by applying different link-quality metrics to the same routing protocol, we provide extensive evaluation on ETX, 4Bit and RNP with insights on their performance under different criteria. Second, we study the impact of the presence or absence of a blacklisting policy when using these link quality estimation metrics. As to our knowledge, this paper is the first to compare the performance between these link quality based metrics with networks of different qualities under realistic conditions.

Running efficiency is an important factor to consider in order to avoid injury. In particular, foot pronation, the angle of the foot as it hits the ground, is a common cause for many types of injuries among runners. Though pronation is common, diagnosing pronation is difficult and imprecise. Currently there is no method of diagonsis which can quantify the severity of pronation. In this paper we propose using WISP (Wireless Identification and Sensing Platform) sensors to help identify and quantify foot pronation.

The ease of deploying wireless camera sensor nodes has grown with the reduction of manufacturing costs of low power, high resolution cameras. Although current wireless sensor network platforms have limited on-board resources for solving highly complex computer vision problems, we show that by splitting the processing costs between the sensor node and a powerful backend, we can achieve better classification results. Using such a distributed processing approach, we balance the computational and communication costs for achieving better detection performance while improving the system lifetime.

802.15.4 links experience different level of dynamics at short and long time scales. This makes the design of a suitable model that combines the different dynamics at different timescales a non-trivial problem. In this paper, we propose a novel multilevel approach involving Hidden Markov Models (HMMs) and Mixtures of Multivariate Bernoullis (MMBs) for modeling the long and short time scale behavior of wireless links using experimental data traces collected from multiple 802.15.4 testbeds. We characterize the synthetic traces generated from the model of the wireless link in terms of statistical characteristics as compared to an empirical trace with similar PRR characteristics, such as the mean and variance of the packet reception rates from the data traces, comparison of distributions of run lengths and conditional packet delivery functions of successive packet receptions (1's) and losses (0's). We modified TOSSIM to utilize data traces created using our modeling approach and compare them against the existing radio model in TOSSIM, which uses the Closest-fit Pattern Matching model for modeling variations in noise which affect the link quality. The results show that our proposed modeling approach is able to mimic the behavior of the data traces quite closely, with difference in packet reception rates of the empirical and simulated traces of less than 2.5% on average and 9% in the worst case. Moreover, the simulated links from our proposed approach were able to account for long runs of 1's and 0's as observed in empirical data traces.

Modeling the behavior of 802.15.4 links is a non-trivial problem because of the widespread heterogeneity in the quality of any given link over short and long time scales. We propose a novel multilevel approach involving Hidden Markov Models (HMMs) and Mixtures of Multivariate Bernoullis (MMBs) for modeling the long and short time scale behavior of wireless links using experimental data traces collected from multiple 802.15.4 testbeds. We characterize the synthetic traces generated from the proposed model in terms of statistical characteristics as compared to an empirical trace with similar PRR characteristics.

We propose modeling environmental noise in order to efficiently and accurately simulate wireless packet delivery. We measure noise traces in many different environments and propose three algorithms to simulate noise from these traces. We evaluate applying these algorithms to signal-to-noise curves in comparison to existing simulation approaches used in EmStar, TOSSIM, and ns2. We measure simulation accuracy using the Kantorovich-Wasserstein distance on conditional packet delivery functions. We demonstrate that using a closest-fit pattern matching (CPM) noise model can capture complex temporal dynamics which existing approaches do not, increasing packet simulation fidelity by a factor of 2 for good links, a factor of 1.5 for bad links, and a factor of 5 for intermediate links. As our models are derived from real-world traces, they can be generated for many different environments.

Recently, several wireless sensor network studies demonstrated large discrepancies between experimentally observed communication properties and properties produced by widely used simulation models. Our first goal is to provide sound foundations for conclusions drawn from these studies by extracting relationships between location (e.g. distance) and communication properties (e.g. reception rate) using non-parametric statistical techniques. The objective is to provide a probability density function that completely characterizes the relationship. Furthermore, we study individual link properties and their correlation with respect to common transmitters, receivers and geometrical location. The second objective is to develop a series of wireless network models that produce networks of arbitrary sizes with realistic properties. We use an iterative improvement-based optimization procedure to generate network instances that are statistically similar to empirically observed networks. We evaluate the accuracy of our conclusions using our models on a set of standard communication tasks, like connectivity maintenance and routing.

Recently, several studies have analyzed the statistical properties of low power wireless links in real environments, clearly demonstrating the differences between experimentally observed communication properties and widely used simulation models. However, most of these studies have not performed in depth analysis of the temporal properties of wireless links. These properties have high impact on the performance of routing algorithms.Our first goal is to study the statistical temporal properties of links in low power wireless communications. We study short term temporal issues, like lagged autocorrelation of individual links, lagged correlation of reverse links, and consecutive same path links. We also study long term temporal aspects, gaining insight on the length of time the channel needs to be measured and how often we should update our models.Our second objective is to explore how statistical temporal properties impact routing protocols. We studied one-to-one routing schemes and developed new routing algorithms that consider autocorrelation, and reverse link and consecutive same path link lagged correlations. We have developed two new routing algorithms for the cost link model: (i) a generalized Dijkstra algorithm with centralized execution, and (ii)a localized distributed probabilistic algorithm.

Advances in microsensor and radio technology enable small but smart sensors to be deployed for a wide range of environmental monitoring applications. The low-per node cost allows these wireless networks of sensors and actuators to be densely distributed. The nodes in these dense networks coordinate to perform the distributed sensing and actuation tasks. Moreover, as described in this paper, the nodes can also coordinate to exploit the redundancy provided by high density so as to extend overall system lifetime. The large number of nodes deployed in this systems preclude manual configuration, and the environmental dynamics precludes design-time preconfiguration. Therefore, nodes have to self-configure to establish a topology that provides communication under stringent energy constraints. ASCENT builds on the notion that, as density increases, only a subset of the nodes is necessary to establish a routing forwarding backbone. In ASCENT, each node assesses its connectivity and adapts its participation in the multihop network topology based on the measured operating region. This paper motivates and describes the ASCENT algorithm and presents analysis, simulation, and experimental measurements. We show that the system achieves linear increase in energy savings as a function of the density and the convergence time required in case of node failures while still providing adequate connectivity.

Advances in micro-sensor and radio technology will enable small but smart sensors to be deployed for a wide range of environmental monitoring applications. The low per-node cost will allow these wireless networks of sensors and actuators to be densely distributed. The nodes in these dense networks will coordinate to perform the distributed sensing tasks. Moreover, as described in this paper, the nodes can also coordinate to exploit the redundancy provided by high density, so as to extend overall system lifetime. The large number of nodes deployed in these systems will preclude manual configuration, and the environmental dynamics will preclude design-time pre-configuration. Therefore, nodes will have to self-configure to establish a topology that provides communication and sensing coverage under stringent energy constraints. In ASCENT, each node assesses its connectivity and adapts its participation in the multi-hop network topology based on the measured operating region. This paper motivates and describes the ASCENT algorithm and presents simulation and experimental measurements.