1,721,340 research outputs found
LO-SC: Local-Only Split Computing for Accurate Deep Learning on Edge Devices
Full text linkSplit Computing (SC) enables deploying a Deep Neural Network (DNN) on edge devices with limited resources by splitting the workload between the edge device and a remote server. However, relying on a server can be expensive, requires a reliable network, and introduces unpredictable latency. Existing solutions for on-device DNNs deployment often sacrifice accuracy for efficiency. In this paper, we study how to use the concepts from SC to split a DNN for executing on the same device without compromising accuracy. In other words, we propose Local-Only Split Computing (LO-SC), a new approach to split a DNN for execution entirely on the edge device while maintaining high accuracy and predictable latency. We formalize LO-SC as a MixedInteger Linear Problem (MILP) problem and solve it using a multi-constrained ordered knapsack algorithm. The proposed method achieves promising results on both synthetic and realworld data, offering a viable alternative for accurately deploying DNNs on resource-constrained edge devices. The source code is available at https://github.com/intelligolabs/LO-SC
MTL-Split: Multi-Task Learning for Edge Devices using Split Computing
Full text linkSplit Computing (SC), where a Deep Neural Network (DNN) is intelligently split with a part of it deployed on an edge device and the rest on a remote server is emerging as a promising approach. It allows the power of DNNs to be leveraged for latency-sensitive applications that do not allow the entire DNN to be deployed remotely, while not having sufficient computation bandwidth available locally. In many such embedded systems scenarios, such as those in the automotive domain, computational resource constraints also necessitate Multi-Task Learning (MTL), where the same DNN is used for multiple inference tasks instead of having dedicated DNNs for each task, which would need more computing bandwidth. However, how to partition such a multi-tasking DNN to be deployed within a SC framework has not been sufficiently studied. This paper studies this problem, and MTL-Split, our novel proposed architecture, shows encouraging results on both synthetic and real-world data. The source code is available at https://github.com/intelligolabs/MTL-Split
Intelligent Chargers Will Make Mobile Devices Live Longer
No full textEditor's notes: Editor's note: Battery aging is becoming a major concern in mobile devices such as laptops or smartphones and often results in premature device replacement. This perspective article gives an overview of recent advances made in battery-health-aware charging and highlights the benefits of making chargers more intelligent to improve the cycle life of different battery-powered devices.
Artifact: WirelessEye – Seeing over WiFi Made Accessible
No full textWirelessEye is a framework for developing WiFi-based sensing systems. It serves as a link between a (patched) radio firmware and a classifier for detecting, e.g., human activity. In brief, WirelessEye supports the following functionality related to a WiFi’s channel state information (CSI) in an “out- of-the-box” manner: Visualizing CSI data in real-time, Recording CSI data using different file formats, Customizable preprocessing of CSI data, Streaming of the preprocessed data to a classifier, Visualizing machine and deep learning classification results in real-time. WirelessEye can be extended and enhanced using plugins. Ad- ditional documentation, including additional details on setting up and running a WiFi sensing system, is available in the WirelessEye repository
Dynamic Service Synthesis and Switching for Medical IoT and Ambient Assisted Living
No full textIn the Internet of Things (IoT), our surrounding will include a large variety of devices from different manufac- turers. One of its promising branches, the medical IoT, will also be accompanied by heterogeneous smart-home infrastructures. However, the efficacy of a medical IoT application will depend on how well the surrounding smart devices collaborate with it to serve the individual needs of the users. Pre-programmed solutions lack flexibility to adapt to each need and environment, and fail to make full use of the capabilities of a set of smart devices. In this paper, we propose a concept based on the flexible and user-friendly synthesis and switching of services for medical IoT applications. The crux of the concept is to provide a methodology in which non-experts can dynamically define services based on their needs. We describe a potential scenario, discuss the associated challenges, and present preliminary results on the feasibility of this approach. Particularly, we focus on design aspects for realizing the concept and propose the use of interpreters on the smart devices as alternative solution. We show that such an approach is feasible in terms of implementation and energy consumption while still maintaining the full flexibility of the service synthesis
Wear Leveling-Aware Active Battery Cell Balancing
No full textDue to manufacturing variabilities and temperature gradients within an electric vehicle's battery pack, the capacities of cells in it decrease differently over time. This reduces the usable capacity of the battery - the charge levels of one or more cells might be at the minimum threshold while most of the other cells have residual charge. Active cell balancing (i.e., transferring charge among cells) can equalize their charge levels, thereby increasing the battery pack's usable capacity. But performing balancing means additional charge transfer, which can result in energy loss and cell aging, akin to memory aging in storage technologies due to writing. This paper studies when cell balancing should be optimally triggered to minimize aging while maintaining the necessary driving capability. In particular, we propose optimization strategies for cell balancing while minimizing their impact on aging. By borrowing terminology from the storage domain, we refer to this as "wear levelingaware" active balancing
Statistical Approach to Efficient and Deterministic Schedule Synthesis for Cyber-Physical Systems
No full textCorrectness of controller implementations rely on real-time
guarantees that all control tasks finish execution by their prescribed
deadlines. However, with increased complexity and heterogeneity in hardware,
the worst-case execution time estimates are becoming very conservative.
Thus, for efficient usage of hardware resources, some control tasks
might have to miss their deadlines. Recent work has shown that a system
can still abide by its safety requirements even after missing some of
its deadlines. This paper investigates an approach to synthesize a scheduler
for control tasks that miss some deadlines without compromising
its safety requirements. But given that the number of possible schedules
increase combinatorially with the number of tasks involved, our scheduler
synthesis uses an efficient automata representation to search for the
appropriate schedule. We incorporate statistical verification techniques
to construct this automaton and accelerate the search process. Statistical
verification is advantageous compared to deterministic verification in the
synthesis process in two ways: first, it enables us to synthesize schedules
that would not be possible otherwise, and second, it drastically reduces
the time taken to synthesize such a schedule. We demonstrate both these
advantages through a case study with five controllers having different
safety specifications, but sharing the same computational resource
Process Dynamics-Aware Flexible Manufacturing for Industry 4.0
No full textThis paper studies the following basic flexible manufacturing problem: Given N machines that can perform the same job on a production item (e.g., drilling or tapping) but with different capabilities (e.g., energy requirements and speeds), what is an optimal schedule for the job on these machines? While this is a well-studied problem, the main innovation this paper introduces is the explicit modeling of the underlying process dynamics—i.e., the physical interaction of the item and the machine—using differential equations. The resulting scheduling problem is in a hybrid systems setting that involves determining the transition times between states, where the system evolution in each state is defined by differential equations. To the best of our knowledge, such a cyber-physical systems (CPS) oriented approach to machine scheduling has not been studied before, although it lies at the core of flexible manufacturing in Industry 4.0. We believe that this new formulation might lead to a renewed interest in machine scheduling problems, but now in a hybrid/CPS-oriented setting
WirelessEye - Seeing over WiFi Made Accessible
No full textWhile commonly used for communication, recently, WiFi is increasingly being used for sensing. In particular, wireless signals are altered (i.e., absorbed, reflected, and attenuated) by the human body and objects in the environment. This can be perceived by an observer to infer information on the environment and hence, to “see” over WiFi. So far, works in this area have led to a variety of custom software tools – each designed for a specific purpose. Moreover, given how scattered the literature is, it is difficult to even identify all processing steps or functionalities necessary for WiFi sensing. To the best of our knowledge, there has been no effort towards a generic solution that helps promote further research and boost new applications in the area. With this as a motivation, we propose WirelessEye, a freely available, generic software framework that allows bootstrapping WiFi sensing systems on low-cost hardware, such as a Raspberry Pi. WirelessEye consolidates all necessary processing steps in a single framework, from collecting and visualizing data to executing different machine learning models in real-time for the purpose of comparison. This way, researchers and practitioners can focus on aspects of their research/applications rather than dealing with the many implementation hurdles of WiFi sensing.While commonly used for communication, recently, WiFi is increasingly being used for sensing. In particular, wireless signals are altered (i.e., absorbed, reflected, and attenuated) by the human body and objects in the environment. This can be perceived by an observer to infer information on the environment and hence, to 'see' over WiFi. So far, works in this area have led to a variety of custom software tools - each designed for a specific purpose. Moreover, given how scattered the literature is, it is difficult to even identify all processing steps or functionalities necessary for WiFi sensing. To the best of our knowledge, there has been no effort towards a generic solution that helps promote further research and boost new applications in the area. With this as a motivation, we propose WirelessEye, a freely available, generic software framework that allows bootstrapping WiFi sensing systems on low-cost hardware, such as a Raspberry Pi. WirelessEye consolidates all necessary processi..
- …
