Modern microcontroller units (MCUs) support enhanced direct memory access (DMA) such as Linked DMA (LDMA) mechanisms that not only offload bulk I/O from the processor core but also support simple commands to minimize processor intervention between bulk transfers. However, straightforward offloading to dedicated channels results in full occupancy of the channel resources even if the actual I/O load is low. To address this problem, we propose an LDMA task scheduler that schedules groups of tasks to enable their shared access to the same set of channels. When applied to a real-life multi-sensor device, the proposed scheme reduces total channel occupancy from 100% down to 32.95% while incurring minimal processor overhead of 0.1%, thereby enabling offloading of twice as many I/O tasks as the fixed channel allocation scheme.

This article presents a family of computational storage drives (CSDs) and demonstrates their performance and power improvements due to in-storage processing (ISP) when running big data analytics applications. CSDs are an emerging class of solid state drives that are capable of running user code while minimizing data transfer time and energy. Applications that can benefit from in situ processing include distributed training, distributed inferencing, and databases. To achieve the full advantage of the proposed ISP architecture, we propose software solutions for workload balancing before and at runtime for training and inferencing applications. Other applications such as sharding-based databases can readily take advantage of our ISP structure without additional tooling. Experimental results on different capacity and form factors of CSDs show up to 3.1× speedup in processing while reducing the energy consumption and data transfer by up to 67% and 68%, respectively, compared to regular enterprise solid state drives.

Computational storage devices enable in-storage processing of data in place. These devices contain 64-bit application processors and hardware accelerators that can help improving performance and saving power by reducing or eliminating data movement between host computers and storage units. This paper proposes a framework, named Stannis, for distributed in-storage training of deep neural networks on clusters of computational storage devices. This in-storage processing style of training ensures that private data never leaves the storage while fully controlling the public sharing of data. The Stannis framework distributes the workload based on the processing power of each worker by determining the proper batch size for each node. Stannis also ensures the availability of input data for all nodes to avoid rank stall while maximizing the utilization and overall processing speed. Experimental results show up to 2.7x speedup and 69% reduction in energy consumption with no significant loss in accuracy.

Distributed training is a novel approach to accelerating training of Deep Neural Networks (DNN), but common training libraries fall short of addressing the distributed nature of heterogeneous processors or interruption by other workloads on the shared processing nodes. This paper describes distributed training of DNN on computational storage devices (CSD), which are NAND flash-based, high-capacity data storage with internal processing engines. A CSD-based distributed architecture incorporates the advantages of federated learning in terms of performance scalability, resiliency, and data privacy by eliminating the unnecessary data movement between the storage device and the host processor. The paper also describes Stannis, a DNN training framework that improves on the shortcomings of existing distributed training frameworks by dynamically tuning the training hyperparameters in heterogeneous systems to maintain the maximum overall processing speed in term of processed images per second and energy efficiency. Experimental results on image classification training benchmarks show up to 3.1x improvement in performance and 2.45x reduction in energy consumption when using Stannis plus CSD compare to the generic systems.

This paper describes EcoMicro, a miniature, self-powered, wireless inertial-sensing node in the volume of 8 x 13 x 9.5 mm3, including energy storage and solar cells. It is smaller than existing systems with similar functionality while retaining rich functionality and efficiency. It is capable of measuring motion using a inertial measurement unit (IMU) and communication over Bluetooth Low Energy (BLE) protocol. It is self-powered by miniature solar cells and can perform maximum power point tracking (MPPT). Its integrated energy-storage device combines the longevity and power density of supercapacitors with the relatively flat discharge curve of batteries. Our power-ground gating circuit minimizes leakage current during sleep mode and is used in conjunction with the real-time-clock for duty cycling. Experimental results show EcoMicro to be operational and efficient for a class of wireless sensing applications.

This paper describes a miniature, wearable wireless hand- gesture recognition (HGR) system called HANDIO. It obtains its input from not only traditional inertial sensors but also a muscle tension sensor (MTS). The addition of MTS enables recognition of a much broader range of intuitive hand gestures, particularly those involving the wrist, that would otherwise be difficult to distinguish by traditional inertial-only HGRs. Among MTSs, we choose an optical MTS over the conventional surface electromyography(sEMG) for the small size, low power consumption, wearing comfort, and good detection rate. This novel miniaturized design enables the whole system to be easily patched on the wrist area or integrated into a wearable device such as a wristband or watch without extra wiring. Experimental results show that a total of 8 hand gestures involving the wrist can be recognized with a detection rate over 93%. The average power consumption of the optical sensor is only around 258μW. This versatile system can also be used to detect other joint activities such as the elbow and knee joint.

Sternocleidomastoid (SCM) is a paired muscle that stretches along both sides of the neck area. It acts as an accessory muscle of inhalation. Abnormal SCM contraction during asthma is usually a sign of further respiratory impairment. Thus, monitoring SCM muscles has great significance in asthma assessment and control. In this work, we develop a wearable monitoring system based on an optical sensor that consists of an LED and a photo detector (PD). A voltage comparator enables the microcontroller unit (MCU) to remain in sleep mode until waken upon detecting contraction. Experimental results show that our optical sensor consumes much lower power than surface electromyography (sEMG), the most commonly used technique while offering more comfort and compactness. It is also robust to motion artifact and DC baseline wandering. These properties simplify the hardware design, while the use of the comparator further reduces the system power consumption to >450 μW on average, making it the best option for low power monitoring.

This work investigates the theoretical analysis to enable fast and accurate estimation of dominant frequencies from randomly sampled signals by compressive sensing (CS). We show that dominant frequencies can be discovered using partially computed Discrete Cosine Transform (DCT). We also propose a new system structure with an estimation unit that enables the signal reconstruction to be selectively bypassed for CS-based devices on signals with dominant frequencies, thus increasing the responsiveness and further reducing the power consumption. For verification, we design a photoplethysmagram (PPG) based heart rate monitor using the proposed algorithm. The accuracy is tested using MIMIC database. The detected heart rate is within 1 beat per minute from the reference over 99% of the data.

We consider the goal of adaptively controlling access to medium in wireless sensor networks with minimal footprint and complexity. Lightweight protocols are desirable in general and inevitable in some real-world applications such as infant monitoring. We assume high-data-rate, resource-constrained wireless sensor nodes in a star network. We take a deterministic approach to contention resolution to achieve bounded latency on data transmissions while reducing the overhead. In addition to simplicity and small footprint, our approach requires fewer message exchanges compared to the state of the art. We then present a lightweight hybrid protocol that is seamlessly integrated with the proposed contention-resolution scheme, one better suited for delay-sensitive applications with real-time constraints by providing determinism. Another feature of the proposed protocol is that it requires carrier-sensing hardware only on the base station but not on the sensor nodes. The average probing complexity of our protocol in a highly dynamic network is O ( n /log n ) per round, and experimental results show the proposed scheme to be scalable and highly adaptive to the contention level.

This work presents energy harvesting techniques from low-voltage current used to prevent galvanic corrosion between a metallic structure and a permanent copper/copper sulfate (Cu/CuSO4) reference electrode. Supercapacitors are adopted to compensate for or overcome the limitations of batteries. Then, a boost converter is used to convert the low voltage levels of galvanic corrosion to that needed by the complementary metal oxide semiconductor (CMOS) technologies used for the wireless sensor systems. Experimental results show that our proposed harvesting schemes significantly reduce the overhead of the charging circuitry, which enables nearly full charging of supercapacitors of up to 350F under the low power conditions of 3mW (i.e., 3mA at 1,V). More importantly, our system enables maintenance-free operation of remote-monitoring cathodic protection (RMCP) systems in harsh environments, where sunlight or wind power may be unavailable or unpredictable.

In this paper, we propose an estimation algorithm for spatially-variant blur due to camera motion. To estimate the most accurate latent image, we integrated depth sensor (Microsoft Kinect) and IMU sensor with the camera. The joint analysis of the blurry image, IMU data and the depth data provide better recovery of the real camera motion during the course of the exposure. The reconstructed camera trajectory along with the depth map is then used to synthesize a spatially-variant blur kernel to estimate the final latent (non-blurry) image. The results show that our algorithm effectively compensates the motion blur from the original image while taking scene geometry into account.

Power-transmission loss can be a severe problem for low-power embedded systems organized in a daisy-chain topology. The loss can be so high that it can result in failure to power the load in the first place. The first contribution of this article is a recursive algorithm for solving the transmission current on each segment of the daisy chain at a given supply voltage. It enables solving not only the transmission loss but also reports infeasible configurations if the voltage is too low. Using this core algorithm, our second contribution is to find energy-efficient configurations that use local energy buffers (LEBs) to eliminate peak load on the bus without relying on high voltage. Experimental results confirm that our proposed techniques significantly reduce the total energy consumption and enable the deployed system to operate for significantly longer.

This paper presents a light-weight hybrid protocol called Bin-MAC (Binary Medium Access Control) for highly resource-constrained wireless sensor nodes. In addition to simplicity and low footprint, a distinguishing feature of Bin-MAC is its deterministic contention resolution mechanism, which enables it to achieve bounded latency on data transmissions. As a result, Bin-MAC can be applied to delay-sensitive applications with real-time constraints, a feature not provide by most of the existing hybrid protocols. Another feature of the proposed protocol is that it requires carrier-sensing hardware only on the base station side, and sensor nodes do not have to possess such capability. Bin-MAC does not require clock synchronization, and thus clock drifts have no impact on its performance. Experimental results show Bin-MAC to be scalable and able to handle large network sizes without noticeable performance degradation.

This article presents a buffer minimization scheme with low dispatching overhead for embedded software processes. To accomplish this, we exploit behavioral transparency in the model of computation. In such a model (e.g., synchronous dataflow), the state of buffer requirements is determined completely by the firing sequence of the actors without requiring functional simulation of the actors. Fine-grained buffer allocation incurs high and code pointer overhead while coarse-grained allocation suffers from memory fragmentation. Instead, we propose a medium-grained, “access-contiguous” buffer allocation scheme that minimizes the total buffer space and pointer overhead. We formulate the buffer allocation problem as 2D tiles that represent the lifetime of the buffers to minimize their memory occupation spatially and temporally. Experimental results show that our scheme uses less data memory than existing techniques by 26% on average, or up to 57% in the best case. Our technique retains code modularity for dynamic configuration and, more importantly, enables many more applications that otherwise would not fit if implemented using previous state-of-the-art techniques.

Rapidly and accurately estimating the impact of design decisions on performance metrics is critical to both the manual and automated design of wireless sensor networks. Estimating system-level performance metrics such as lifetime, data loss rate, and network connectivity is particularly challenging because they depend on many factors, including network design and structure, hardware characteristics, communication protocols, and node reliability. This paper describes a new method for automatically building efficient and accurate predictive models for a wide range of system-level performance metrics. These models can be used to eliminate or reduce the need for simulation during design space exploration. We evaluate our method by building a model for the lifetime of networks containing up to 120 nodes, considering both fault processes and battery energy depletion. With our adaptive sampling technique, only 0.27% of the potential solutions are evaluated via simulation. Notably, one such automatically produced model outperforms the most advanced manually designed analytical model, reducing error by 13% while maintaining very low model evaluation overhead. We also propose a new, more general definition of system lifetime that accurately captures application requirements and decouples the specification of requirements from implementation decisions.

Detecting and reacting to faults is an indispensable capability for many wireless sensor network applications. Unfortunately, implementing fault detection and error correction algorithms is challenging. Programming languages and fault tolerance mechanisms for sensor networks have historically been designed in isolation. This is the first work to combine them. Our goal is to simplify the design of fault-tolerant sensor networks. We describe a system that makes it unnecessary for sensor network application developers and users to understand the intricate implementation details of fault detection and tolerance techniques, while still using their domain knowledge to support fault detection, error correction, and error estimation mechanisms. Our FACTS system translates low-level faults into their consequences for application-level data quality, i.e., consequences domain experts can appreciate and understand. FACTS is an extension of an existing sensor network programming language; its compiler and runtime libraries have been modified to support automatic generation of code for on-line fault detection and tolerance. This code determines the impacts of faults on the accuracies of the results of potentially complex data aggregation and analysis expressions. We evaluate the overhead of the proposed system on code size, memory use, and the accuracy improvements for data analysis expressions using a small experimental testbed and simulations of large-scale networks.

Energy harvesting has been studied as a candidate for powering next generation wireless sensor networks. The technologies that can harvest electric power from ambient energy sources include solar, vibration, heat and wind. However, sensor nodes powered by energy harvesting devices cannot always communicate with other nodes because the energy harvesting devices cannot provide a stable supply power. A node cannot know whether its neighboring nodes have enough energy to receive a data packet that it has transmitted. During the process of relaying the packet, each additional hop increases the overall probability of losing the packet. In this paper, we propose two data collection protocols for the energy harvesting wireless sensor networks called Probabilistic ReTransmission protocol (PRT) and PRT with Collision Consideration (PRT-CC). The idea is to derive the number of times to retransmit a packet based on the reception probability and the active intervals computed by the receivers themselves. In PRT-CC, each node computes the reception probability with packet collision consideration. The simulation results show that the proposed protocols achieve higher delivery ratio than the previous works (GR-DD and GR DD-RT).

This paper describes EcoIMU, a gyro-free, wearable inertial measurement unit (IMU) built with a pair of triaxial accelerometers that are spatially separated and wirelessly connected on the body. It can output the translation and rotation of the body for the purpose of motion tracking and dead reckoning applications. It mitigates error accumulation and drift problems using domain knowledge including pause identification and geometric constraints on the two nodes. Experimental results show EcoIMU to have less drift error while consuming significantly lower power than comparable IMUs built with MEMS gyroscopes and accelerometers.

TeleScribe is a software mechanism for efficiently reprogramming embedded systems such as wireless sensor nodes over a shared communication link. One distinguishing feature is its ability to resume update on any node after power failure, link disconnection, or many other indefinite disruptive events. The nodes are guaranteed never to be left in a bad state as a result of such incomplete reprogramming procedures. Moreover, TeleScribe efficiently disseminates the binary image to as many nodes as possible, thereby minimizing redundant communication while ensuring that, in a later round as needed, all nodes receive packets that were lost earlier. Experimental results show TeleScribe to be the fastest and smallest among similar systems, achieving an update rate of about 95 bytes per node per second in a 100-node system. The total code size of our implementation on the node is around only 2 KB, making TeleScribe easily adaptable to a wide range of platforms with little overhead

Nucleos is a new runtime system for ultra-lightweight embedded systems. Central to Nucleos is a dispatcher based on the concept of recursive threaded code, which enables layers of abstraction from the runtime system and interrupt handlers to application tasks to be composed in a structured, powerful way, all with minimal program code. When used in conjunction with models of computation with behavioral transparency such as synchronous dataflow (SDF), Nucleos can support efficient memory allocation policies for multiple communicating actors with minimal runtime overhead. This recursive structure also lends itself to in-field code update and dynamic execution. Nucleos's low runtime overhead and low RAM/ROM requirements enable it to run on compact platforms previously unsupported by the most popular sensor OSes while still providing high flexibility and composability. In some cases, applications running on Nucleos actually outperform hand-crafted code running without an OS, thanks to non-obvious memory optimizations enabled by the SDF model of computation.

DuraCap is a solar-powered energy harvesting system that stores harvested energy in supercapacitors and is voltage-compatible with lithium-ion batteries. The use of supercapacitors instead of batteries enables DuraCap to extend the operational life time from tens of months to tens of years. DuraCap addresses two additional problems with micro-solar systems: inefficient operation of supercapacitors during cold booting, and maximum power point tracking (MPPT) over a variety of solar panels. Our approach is to dedicate a smaller supercapacitor to cold booting before handing over to the array of larger-value supercapacitors. For MPPT, we designed a bound-control circuit for PFM regulator switching and an I-V tracer to enable self-configuring over the panel's aging process and replacement. Experimental results show the DuraCap system to achieve high conversion efficiency and minimal downtime.

EcoExec is a host-assisted interactive execution environment for wireless sensing systems. Users can interact with sensor nodes by viewing attributes and invoking functions via a command-line interface. Functions that are not resident in the node's firmware are automatically compiled on the host, packaged up and downloaded to the node, linked, and executed, all seamlessly and transparently to the user. By packaging these features in a dynamically object-oriented programming environment such as Python, EcoExec enables programmers to experiment with features of the wireless sensor nodes and to rapidly develop application software. Most importantly, EcoExec empowers resource-constrained wireless sensor platforms with rich functionalities that would otherwise be prohibitive, thanks to its host-assisted execution feature with code swapping over the network. Experimental results on actual wireless sensor platforms show EcoExec to perform effectively with negligible observed overhead to the user.

Enix is a lightweight dynamic operating system for tightly constrained platforms for wireless sensor networks (WSN). Enix provides a cooperative threading model, which is applicable to event-based WSN applications with little run-time overhead. Virtual memory is supported with the assistance of the compiler, so that the sensor platforms can execute code larger than the physical code memory they have. To enable firmware update for deployed sensor nodes, Enix supports remote reprogramming. The commonly used library and the main logical structure are separated; each sensor device has a copy of the dynamic loading library in the Micro-SD card, and therefore only the main function and user-defined subroutines should be updated through RF. A lightweight, efficient file system named EcoFS is also included in Enix. The code size and data size of Enix with full-function including EcoFS are 8 KB and 512 bytes, respectively, enabling Enix to run on many RF-enabled systems-on-chip that cannot run most other WSN OSs.

Firmware update over a network connection is an essential but expensive feature for many embedded systems due to not only the relatively high power consumption and limited bandwidth, but also page-granular erasure before rewriting to flash memory. This work proposes a page-level, link-time technique that minimizes not only the size of patching scripts but also perturbation to the firmware memory, over the entire sequence of updates in the system’s lifetime. We propose a tool that first clusters functions to minimize caller-callee dependency across pages, and then orders the functions within each page to minimize intrapage perturbation. Experimental results show our technique to reduce the energy consumption of firmware update by 30--42% over the state-of-the-art. Most importantly, this is the first work that has ever shown to evolve well over 41 revisions of a real-world open-source real-time operating system.

Endosymbiotic Computing entails attaching an RF-enabled micro-controller module (endomodule) to an appliance such that it appears as a networked device in the cyber world. It enables a smart phone to work as not only a universal remote control but also a surrogate GUI for inspecting the attributes of these appliances, without modifications to legacy circuits. To minimize the cost and resource requirements of the endomodules, we propose a generalized active message programming method that executes dynamically-loaded threaded code on-demand without requiring parsing.

Firmware update over a network connection is an essential but expensive feature for many embedded systems due to the relatively high power consumption and limited bandwidth. This work proposes a page-level, link-time technique that minimizes not only the size of patching scripts but also perturbation to the firmware memory, over the entire sequence of updates in the system's lifetime. Experimental results show our technique to reduce the energy consumption of firmware update by 38--42% over the state-of-the-art.

EcoPlex is an infrastructure that enables simple wireless platforms to participate seamlessly in a feature-rich, wireless ad hoc network. EcoPlex consists of gateways that are responsible for handoff support for mobility and high data rate without burdening the simple nodes to implement multihop pro

Current wireless sensor networks achieve dense deployment by low duty cycling. Unfortunately, this represents a great mismatch with many real-world applications that require non-trivial data rates. To solve these issues above, we propose EcoDAQ, a new densely distributed wireless data acquisition system with a relatively high data rate. We implement a MAC layer protocol that minimizes complexity and memory footprint on the sensor node while guaranteeing collision freedom, without using multiple frequency channels. As a result, this work is not only highly scalable to at least 50 active nodes per squared meter, but at the same time the low complexity minimizes the cost of these ultra-compact sensor devices. The full demo would show a system deployment with a minimum of 50 active sensors on a poster board of 1 m2area. The monitoring station keeps gathering up to 500 accelerometer data samples per second from all of these 50 sensor nodes and displays them simultaneously on a fifty-screen tiled display system. A portable version that emulates the tiled screens in a reduced size will be shown at the conference.

Densely distributed wireless real-time system must perform communication scheduling and protocols in addition to task scheduling to achieve schedulability and reliable communication. While many wireless sensor networks are densely distributed, they assume low duty cycling and trivial data packet rates, and they cannot handle many real-world applications. To highlight these design issues, this paper presents a case study with EcoDAQ, a wireless data acquisition system with 50 sensor nodes streaming real-time data over the same frequency channel. To achieve reliable communication, we define a pulling protocol that minimizes node complexity while guaranteeing collision freedom on a given frequency channel. Our case study illustrates the impact of communication requirements on the rest of the systems including bus and processing speeds. Experimental results show that our monitoring station can sustain gathering accelerometer data from up to 50 nodes in one square meter area at up to 15,000 samples per second with minimal latency and show very high expandability, flexibility and reliability.

This work is to augment a debris flow observation and early warning system with wireless sensor networks. Previously, the GIS at Fengchia University has constructed and deployed state-of-the-art, stationary and mobile types of observation systems at nearly 20 sites throughout Taiwan. These sites collect data from sensors ranging from rain gauges and tension cables to ultrasonic sensors and CCD cameras, and transmit them back to the GIS via a lower-orbit satellite uplink in real-time. A new wireless sensor network and middleware system are being designed and implemented to overcome several limitations with the current system. Wireless communication capabilities are being incorporated to enhance the coverage. Previously, most connections between the sensors and the server before the satellite uplink are wired or Wi-Fi with fixed topology and limited range. New wireless interfaces with a 500 m - 1 km range plus energy harvesting devices on the sensors reduces deployment effort and cost. More importantly, it is now becoming possible to construct and deploy brand new types of mobile sensor nodes that move with the debris flow along its path. Such sensor nodes are to be housed in pyramid-shaped, weather-proof capsules that contain motion sensors, GPS and other localization devices, energy harvesting and storage devices, and wireless transceivers. Normally in low-power or standby mode, these capsules would be deployed in the path of potential debris flows. They would stand steadily during normal weather conditions including wind, rain, and water flow. They would get triggered by a threshold motion detector or a rain gauge and start actively monitoring the flow. As it flows with the debris, these capsules transmit their sensor data wirelessly, via other relaying nodes if necessary. Based on the shape and mass of the capsule itself and the velocity, researchers can derive the direction and magnitude of the flow in brand new ways.

The development cost of low-power embedded systems can be significantly reduced by reusing legacy designs and applying proper modifications to meet the new power constraints. The proposed power management techniques in the literature for implementing distributed power managers in multi-processor systems are very costly in terms of hardware and software modifications. For example, extra software code for power management must be integrated into each component of the system. Furthermore, in order to turn on a component at a specific time/event, extra hardware timers and interrupt controllers must be added to each component along with proper software/device driver modifications. In this paper, we propose a new centralized power management technique that reduces the power consumption of distributed realtime systems at very low implementation cost. Our power manager does not need software and hardware modifications of each individual component. Instead, it uses the model of the system/application to compute the schedule of turning on/off commands by dynamically simulating the system for a given application scenario. The dynamic simulation can be conservative to reflect the jitter in arrival time of events and/or variation in execution delay of tasks. We applied our power management technique to a distributed software-defined radio system and achieved significant energy savings (60% to 87%) at the cost of 1% energy consumed by the power manager itself, as verified by actual hardware measurements. Furthermore, our power manager reacts to the changes in application scenario (referred to as mission) within milliseconds.

Software components are modular and can enable post-deployment update, but their high overhead in runtime and memory is prohibitive for many embedded systems. This paper proposes to minimize such overhead by exploiting behavioral transparency in models of computation. In such a model (e.g., synchronous dataflow), the state of buffer requirements is determined completely by the firing sequence of the actors without requiring functional simulation of the actors. Instead of dedicating space to each channel or actor statically, our dispatcher passes buffer pointers to an actor upon firing. Straightforward implementations are counterproductive, as fine-grained allocation incurs high pointer overhead while coarse-grained allocation suffers from fragmentation. To address this problem, we propose medium-grained, "access-contiguous" buffer allocation scheme. We formulate the problem as 2-D tiles that represent the lifetime of the buffers over time and define operators for their translation and transformations to minimize their memory occupation spatially and temporally. Experimental results on real-life applications show up to 70% data memory reduction compared to existing techniques. Our technique retains code modularity for dynamic configuration and, more importantly, enables many more applications that otherwise would not fit if implemented using previous state-of-the-art techniques.

This paper presents a new technique, called Adaptive Stochastic Gradient Voltage-and-Task Scheduling (ASG-VTS), for power optimization of multicore hard realtime systems. ASG-VTS combines stochastic and energy-gradient techniques to simultaneously solve the slack distribution and task reordering problem. It produces very efficient results with few mode transitions. Our experiments show that ASG-VTS reduces number of mode transitions by 4.8 times compared to traditional energy-gradient-based approaches. Also, our heuristic algorithm can quickly find a solution that is as good as the optimal for a real-life GSM encoder/decoder benchmark. The runtime of ASG-VTS is 150 times and 1034 times faster than energy-gradient based and optimal ILP algorithms, respectively. Since the runtime of ASG-VTS is very low, it is ideal for design space exploration in system-level design tools. We have also developed a web-based interface for ASG-VTS algorithm.

This article presents techniques for reducing idle energy by mode-sequence optimization (MSO) under timing constraints. Our component-level CoMSO algorithm computes energy-optimal mode-transition sequences for different lengths of idle intervals. Our system-level SyMSO algorithm shifts tasks within slack intervals while satisfying all timing and resource constraints in the given schedule. Experimental results on a commercial software-defined radio show that these new techniques can reduce idle energy by 50--70%, or 30--50% of total system energy over previous offline-optimal but unsequenced techniques based on localized break-even-time analysis, thanks to rich options offered by mode sequencing.

Tapper is a lightweight scripting engine for highly constrained wireless sensor nodes. Many such nodes run on 8-bit microcontrollers (MCUs) with only a few kilobytes of on-chip memory, often with segmented memory and unorthogonal registers that make it difficult to perform even the most basic tasks. To facilitate development of software routines and enable dynamic reconfigurability, Tapper provides a software layer that can interpret human readable commands either interactively or in batch. It can invoke compiled routines ranging from timer and interrupt configurations and task scheduling to accessing I/O devices. Experiments on platforms including 8051, ATMega 169V, and Freescale MC9S12NE64 with built-in Ethernet show that the Tapper setup has the same modular, dynamic benefits as middleware systems but without the high overhead in code size and data memory.

Harvesting energy from the environment is a desirable and increasingly important capability in several emerging applications of embedded systems such as sensor networks, biomedical implants, etc. While energy harvesting has the potential to enable near-perpetual system operation, designing an efficient energy harvesting system that actually realizes this potential requires an in-depth understanding of several complex tradeoffs. These tradeoffs arise due to the interaction of numerous factors such as the characteristics of the harvesting transducers, chemistry and capacity of the batteries used (if any), power supply requirements and power management features of the embedded system, application behavior, etc. This paper surveys the various issues and tradeoffs involved in designing and operating energy harvesting embedded systems. System design techniques are described that target high conversion and storage efficiency by extracting the most energy from the environment and making it maximally available for consumption. Harvesting aware power management techniques are also described, which reconcile the very different spatio-temporal characteristics of energy availability and energy usage within a system and across a network.

This paper describes a supercapacitor-operated, solar-powered wireless sensor node called Everlast. Unlike traditional wireless sensors that store energy in batteries, Everlast's use of supercapacitors enables the system to operate for an estimated lifetime of 20 years without any maintenance. The novelty of this system lies in the feedforward, PFM (pulse frequency modulated) converter and open-circuit solar voltage method for maximum power point tracking, enabling the solar cell to efficiently charge the supercapacitor and power the node. Experimental results show that Everlast can achieve low power consumption, long operational lifetime, and high transmission rates, something that traditional sensor nodes cannot achieve simultaneously and must trade-off.

This paper reports on recent development of a wearable wireless sensor platform for interactive dance performances. At a fraction of a cubic-centimeter in volume, this platform is truly wearable and scalable informing wireless networks. Integrated with a wide variety of different sensing devices, it is a real-time monitoring system for activities and physical conditions of the human body. The effectiveness of this platform is demonstrated with an interactive dance performance.

Quantitative evaluation of wireless sensor platforms is difficult. Unlike general purpose computers that can run SPEC benchmarks from a file, it is difficult to reproduce the environmental input needed to stimulate the sensor nodes. Even if possible, open-loop playback would be unable to correctly account for adaptivity built into the behavior of these nodes. As a result, researchers resort to simulations, which do not consider all relevant factors without significant speed penalty. To address this problem, we propose EmPro, an environment/energy emulation and profiling system for WSNs. It accurately outputs electrical signals to emulate not only digital and analog inputs to the sensors but also the power sources as well as RF attenuation according to pre-programmed sequences. This emulation approach enables researchers to run the networked sensors in real-time in a realistic manner with full controllability and reproducibility. EmPro in profiling mode can also capture the observable behavior of WSNs for detailed analysis. Experimental results on the Eco and MICA2 WSN platforms show that EmPro can drive these hardware systems in real-time with high accuracy. We expect EmPro will expedite testing and serve as a sorely needed standard benchmarking tool for WSN platforms

AmbiMax is an energy harvesting circuit and a supercapacitor based energy storage system for wireless sensor nodes (WSN). Previous WSNs attempt to harvest energy from various sources, and some also use supercapacitors instead of batteries to address the battery aging problem. However, they either waste much available energy due to impedance mismatch, or they require active digital control that incurs overhead, or they work with only one specific type of source. AmbiMax addresses these problems by first performing maximum power point tracking (MPPT) autonomously, and then charges supercapacitors at maximum efficiency. Furthermore, AmbiMax is modular and enables composition of multiple energy harvesting sources including solar, wind, thermal, and vibration, each with a different optimal size. Experimental results on a real WSN platform, Eco, show that AmbiMax successfully manages multiple power sources simultaneously and autonomously at several times the efficiency of the current state-of-the-art for WSNs

eCAM is an ultra-compact, high data-rate wireless sensor node(WSN) with a miniature camera. It is constructed by interfacing a VGA quality digital video camera with the Eco node. The purpose of this demo is to show that Eco is not only one of the world's smallest self-contained WSNs, but its expansion capability facilitates the construction of new sensing systems over its expansion port. The results show that Eco is not only ultra-compact but also much higher performance than sensor nodes in its class.

Embedded systems are being built with renewable power sources such as solar cells to replenish the energy of batteries. The renewable power sources have a wide range of efficiency levels that depend on environment parameters and the current drawn from the circuit. Unlike low-power designs whose goal is to minimize energy consumption, systems with renewable power sources should maximize the efficiency of the sources by load matching. To match the wide dynamic range of solar output, it is necessary to exploit multiple power "knobs" simultaneously. This paper combines computation vs. communication trade-offs, algorithm selection, scheduling and dynamic voltage scaling to maximize the dynamic range of the load over time. Experimental results show one to two orders of magnitude performance improvement for a wireless handheld system running image compression applications.

Real-world wireless sensing applications demand system platforms with a wide range of size, cost, power consumption, connectivity, performance, and flexibility requirements. These goals cannot be achieved without understanding the nature of the sensing functions in the first place, which can be classified into passive vs. active sensing, event detection vs. data acquisition, and real-time monitoring vs. data logging. This paper discusses platform design techniques for supporting these design goals through the trade-offs of sensing devices, wireless interfaces, and computation and control units. We also cover power subsystem design for supply-aware optimizations, including load/supply matching, power defragmentation in multi-supply systems, and use of supercapacitors. To evaluate these system platforms, we describe an emulation-based benchmarking methodology to quantify fitness metrics.

Eco is an ultra-compact wireless sensor node. Only 648 mm/sup 3/ in volume and weighing under 1.6 grams, Eco was initially designed to monitor the spontaneous motion of preterm infants over 2.4 GHz radio links at the maximum data rate of 1 Mbps. The compact form factor and low power consumption also make Eco nodes highly suitable for many other applications, including medicine, environmental monitoring, new computer-human interface, and ambient intelligence. This paper describes the hardware and software designs of the second generation Eco nodes and the host interface. We also present an evaluation and comparison against other popular sensor nodes in the similar class.

For accurate estimation of battery lifetime, researchers have developed analytical and empirical models and applied them to representative load profiles. However, accurate battery models are not available for most batteries on the market. Although high-accuracy simulation models exist for certain battery chemistries, they are computationally intensive and still require calibration through trial and error. To address this problem, this paper presents a low-cost load emulation platform for automated, accurate battery estimation. By draining a battery with high-frequency emulation of a system power profile, all of the battery characteristics are accounted for, including the discharge rate and recovery effects. A designer can then accurately observe how the system effects battery life, quantify lifetime performance for multiple batteries, and ultimately optimize the system's power scheduling around a particular battery

This paper describes a supercapacitor-operated, solar-powered wireless sensor node called Everlast. Unlike traditional wireless sensors that store energy in batteries, Everlast's use of supercapacitors enables the system to operate for an estimated lifetime of 20 years without any maintenance. The novelty of this system lies in the feed-forward, pulse-frequency modulated (PFM) converter and open-circuit solar voltage method for maximum power point tracking (MPPT), enabling the solar cell to efficiently charge the supercapacitor and power the node. Experimental results show that Everlast can achieve low power consumption, long operational lifetime, and high transmission rates, something that traditional sensor nodes cannot achieve simultaneously and must trade-off.

New embedded systems offer rich power management features in the form of multiple operational and nonoperational power modes. While they offer mechanisms for better energy efficiency, they also complicate power management decisions in the presence of realtime constraints. A traditional dynamic power management techniques based on localized break-even-time analysis with simple on/off power controls often yield suboptimal if not incorrect results globally. To address these problems, This work presents two core algorithms for reducing idle energy consumption at the component level and system level. The first algorithm discovers the optimal sequence for mode transition over multiple power modes under timing constraints. It assists the second algorithm that performs a sophisticated global search strategy to aggressively explore system-wide energy savings by correctly interpreting the constraints across all subsystems. Experimental results show that in an embedded radio system where idle energy cost matches or exceeds the active energy consumption, our technique can further reduce the idle energy by 50-70%, which translates into 30-50% of overall system energy compared to existing techniques.

Summary form only given. A multiple-processor system can potentially achieve higher energy savings than a single processor, because the reduced workload on each processor creates new opportunities for dynamic voltage scaling (DVS). However, as the cost of communication starts to match or surpass that of computation, many new challenges arise in making DVS effective in a distributed system under communication-intensive workload. We discuss implementation issues for supporting DVS on distributed embedded processors. We implemented and evaluated four distributed schemes: (1) DVS during I/O, (2) partitioning, (3) power-failure recovery, and (4) node rotation. We validated the results on a distributed embedded system with the Itsy pocket computers connected by serial links. Our experiments confirmed that a distributed system can create new DVS opportunities and achieve further energy savings. However, a surprising result was that aggregate energy savings do not translate directly into a longer battery life. In fact, the best partitioning scheme, which distributed the workload onto two nodes and enabled the most power-efficient CPU speeds at 30-50%, resulted in only 15% improvement in battery lifetime. Of the four techniques evaluated, node rotation showed the most measurable improvement to battery lifetime at 45% by balancing the discharge rates among the nodes.

Increasing demand for larger high-performance applications requires developing more complex systems with hundreds of processing cores on a single chip. To allow dynamic voltage scaling in each on-chip cores individually, many on-chip voltage regulators must be used. However, the limitations in implementation of on-chip inductors can reduce the efficiency, accuracy and the number of voltage modes generated by regulators. Therefore the future voltage scheduling algorithms must be efficient, even in the presence of few voltage modes; and fast, in order to handle complex applications. Techniques proposed to date, need many fine-grained voltage modes to produce energy efficient results and their quality degrades significantly as the number of modes decreases. This paper presents a new technique called Adaptive Stochastic Gradient Voltage and Task Scheduling (ASG-VTS) that quickly generates very energy efficient results irrespective of the number of available voltage modes. The results of comparing our algorithm to the most efficient approaches (RVS and EE-GLSA) show that in the presence of only four valid modes, the ASG-VTS saves up to 26% and 33% more energy. On the other hand, other approaches require at least ten modes to reach the same level of energy saving that ASG-VTS achieves with only four modes. Therefore our algorithm can also be used to explore and minimize the number of required voltage levels in the system.

Solar power is an important source of renewable energy for many low-power systems. Matching the power consumption level with the supply level can make a great difference in the efficiency of power utilization. This paper proposes a source-tracking power management strategy that maximizes the panel's total energy output under a given solar profile by load matching. The power efficiency was validated by extensive measurement. Compared to a conventional solar powered system, our load matching strategy improves the power utilization by 132% for a portable system performing image processing and wireless communication tasks.

For embedded systems that rely on multiple power sources (MPS), power management must distribute the power by matching the supply and demand in conjunction with the traditional power management tasks. Proper load matching is especially critical for renewable power sources such as solar panels and wind generators, because it directly affects the utility of the available power. This paper proposes a power distribution switch and a source-consumption matching algorithm that maximizes the total utility of the available power from these ambient power sources. Our method yields over 30% more usable power than conventional MPS designs.

The trend towards distributed, networked embedded systems is changing the way power should be managed. Power consumed by bus and network interfaces now matches if not surpasses that of the CPU and is thus becoming a prime candidate for reduction. This paper explores energy-efficient bus topologies as a new technique for global power optimization of embedded systems that are interconnected by high-speed serial network-like busses such as FireWire and a new generation of SoC buses. Our grammar-based representation for these networks enables selection of energy-efficient bus topologies. Experimental results show 15-20% energy saving on the network interfaces without sacrificing system performance.

To save energy, many power management policies rely on issuing mode-change commands to the components of the system. Efforts to date have focused on how these policies interact with the external workload. However, the energy savings are ultimately limited by the set of power saving modes available to the power manager. This paper exposes new power-saving opportunities to existing system-level power managers by handling each desired mode change in terms of an optimal sequence of mode transitions involving multiple components. We employ algorithms to optimize these transition sequences in polynomial time, making them applicable to static and dynamic policies. The decoupling between policies and mechanisms also makes this approach modular and scalable to devices with complex modes and intricate dependencies on other devices in the system. Experimental results show significant energy savings due to these sequentialized mode-change opportunities that would otherwise be difficult to discover manually even by experienced designers.

Transmitting compressed data can reduce inter-processor communication traffic and create new opportunities for DVS (dynamic voltage scaling) in distributed embedded systems. However, data compression alone may not be effective unless coordinated with functional partitioning. This paper presents a dynamic programming technique that combines compression and functional partitioning to minimize energy on multiple voltage-scalable processors running pipelined data-regular applications under performance constraints. Our algorithm computes the optimal functional partitioning, CPU speed for each node, and their respective compression ratios. We validate the algorithm's effectiveness on a real distributed embedded system running an image processing algorithm.

This paper describes B# (B-sharp), a programmable power supply that emulates the behavior of a battery. It measures the current load, calls a battery simulation program to compute the voltage in real time, and controls a linear regulator to mimic the voltage output of a battery. This instrument enables validation of battery-aware power-optimization techniques with accurate, controllable, reproducible results. This instrument also supports training mode with actual batteries, and it can even be used for recording and playback of a solar power source. This design has been prototyped and tested on hand-held devices with high accuracy and fast response time.

Power-aware systems are those that must make the best use of available power. They subsume traditional low-power systems in that they must not only minimize power when the budget is low, but also deliver high performance when required. This paper presents a new scheduling technique for supporting the design and evaluation of a class of power-aware systems in mission-critical applications. It satisfies stringent min/max timing and max power constraints. It also makes the best effort to satisfy the min power constraint in an attempt to fully utilize free power or to control power jitter. Experimental results show that our scheduler can improve performance and reduce energy cost simultaneously compared to hand-crafted designs for previous missions. This tool forms the basis of the IM-PACCT system-level framework that will enable designers to explore many power-performance trade-offs with confidence. 1.

A main advantage of control composition with modal processes [4] is the enhanced retargetability of the composed behavior over a wide variety of target architectures. Unlike previous component models that hardwire the coordination behavior either explicitly in the components or implicitly in the underlying model of computation, modal processes decouple component functionality and coordination protocols. Retargetability is achieved through the synthesis of distributed mode managers, which abstract away low-level synchronization and control communication details that would otherwise be exposed to the component designer. This paper presents an algorithm for the synthesis and optimization of distributed coordination controllers by computing an optimal projection of the global state space onto each processor. It not only minimizes interprocessor communication traffic for coordination but also reduces controller complexity by minimizing replication.

Article ipChinook: an integrated IP-based design framework for distributed embedded systems Share on Authors: Pai Chou Consystant Design Technologies, Inc., Seattle, WA Consystant Design Technologies, Inc., Seattle, WAView Profile , Ross Ortega Department of Computer Science and Engineering, Box 352350, University of Washington, Seattle, WA Department of Computer Science and Engineering, Box 352350, University of Washington, Seattle, WAView Profile , Ken Hines Department of Computer Science and Engineering, Box 352350, University of Washington, Seattle, WA Department of Computer Science and Engineering, Box 352350, University of Washington, Seattle, WAView Profile , Kurt Patridge Department of Computer Science and Engineering, Box 352350, University of Washington, Seattle, WA Department of Computer Science and Engineering, Box 352350, University of Washington, Seattle, WAView Profile , Gaetano Borriello Department of Computer Science and Engineering, Box 352350, University of Washington, Seattle, WA Department of Computer Science and Engineering, Box 352350, University of Washington, Seattle, WAView Profile Authors Info & Claims DAC '99: Proceedings of the 36th annual ACM/IEEE Design Automation ConferenceJune 1999 Pages 44–49https://doi.org/10.1145/309847.309865Online:01 June 1999Publication History 25citation519DownloadsMetricsTotal Citations25Total Downloads519Last 12 Months0Last 6 weeks0 Get Citation AlertsNew Citation Alert added!This alert has been successfully added and will be sent to:You will be notified whenever a record that you have chosen has been cited.To manage your alert preferences, click on the button below.Manage my AlertsNew Citation Alert!Please log in to your account Save to BinderSave to BinderCreate a New BinderNameCancelCreateExport CitationPublisher SiteGet Access

To explore different points in the design space of an embeddedsystem, it is important to be able to compose a designfrom reusable design components, and then map the resultingsystem description onto several possible target architectureswith different partitionings of functionality. Today's specificationmodels support composition styles that work well fordata communication but not for control communication betweenconcurrent processes to be mapped onto a distributedarchitecture. We propose a new retargetable system specificationmodel that combines the best properties of process-basedand hierarchical-FSM-based methods for modular compositionof data and control. The model lends itself to automatedsynthesis of the run-time system for coordinating tasks ondifferent processors in the system. The model and synthesismethod are illustrated with several examples of embeddedsystems.

In traditional distributed embedded system designs, control information is often replicated across severrd processes and kept coherent by application-specific mechanisms.Consequently, processes cannot be reused in a new system without tailoring the code to ded with the new system's control information.The modal process framework [5] provides a high-level way to specify the coherence of replicated control information independently of the behavior of the processes.Thus multiple processes can be composed without internal tailoring and without suffering from errors common in lower-level specification styles.This paper first describes a kernel-language representation for the high-level composition operators; it also presents a synthesis algorithm for the mode manager, the runtime code that maintains control information coherence within and between distributed processors.

A central problem in embedded system co-synthesis is the generation of software for lowlevel I/O.Scheduling still remains a manual task because existing coarse-grained real-time scheduling algorithms are not applicable: they assume xed delays even though the run times are often variable, and they incur too much o v erhead.To solve this problem, we present a new static ordering technique, called interval scheduling, for meeting general timing constraints on ne-grained, variable-delay operations without using a run-time executive.

A key aspect of the synthesis of embedded systems is the automatic integration of system components. This entails the derivation of both the hardware and software interfaces that will bind these elements together and permit them to communicate correctly and efficiently. Without the automatic synthesis of these interfaces, designers are not able to fully simulate and evaluate their systems. Frequently, they are discouraged from exploring the design space of different hardware/software partitions because practical concerns mandate minimizing changes late in the design cycle, thus leading to more costly implementations than necessary. This paper presents a set of techniques that form the basis of a comprehensive solution to the synthesis of hardware/software interfaces. Software drivers and glue logic are generated to connect processors to peripheral devices, hardware co-processors, or communication interfaces while meeting bandwidth and performance requirements. We use as examples a set of devices that communicate over an infrared local communications network (highlighting a video wrist-watch display) to explain our techniques and the need for design space exploration tools for embedded systems.

Abstract { Existing software scheduling techniques limit the functions that can be implemented in software to those with a restricted class of timing constraints, in particular those with a coarse-grained, uniform, periodic behavior. In practice, however, many systems change their I/O behavior in response to the inputs from the environment. This paper considers one such class of systems, called reactive real-time systems, where timing requirements can include sequencing, rate, and response time constraints. We present a static, non-preemptive, ne-grained software scheduling algorithm to meet these constraints. This algorithm is suitable for control-dominated embedded systems with hard real-time constraints, and is part of the core of a hardware/software co-synthesis system. I

A tool that automates the synthesis of the hardware/software interface between a microcontroller and the devices it controls is described. Optimizations of the synthesis process focus primarily on the minimization of interface hardware and also the code size of the device driver routines. This interface synthesis problem is specified and the data structures used to represent the microcontroller, devices, and hardware and software primitives are described. The port allocation algorithm is presented and the features of this interface synthesis tool are highlighted with two practical examples.>