Search

Multi-FPGA platforms, like Amazon AWS F1, can run in the cloud multikernel pipelined applications, like convolutional neural networks (CNNs), with excellent performance and lower energy consumption than CPUs or GPUs. We propose a method to efficiently map these applications on multi-FPGA platforms to maximize the application throughput. Our methodology finds, for the given resources, the optimal number of parallel instances of each kernel in the pipeline and their allocation to one or more among the available FPGAs. We obtain this by formulating and solving a mixed-integer, nonlinear optimization problem, in which we model the performance of each component and the duration of the phases in which the accelerated computation can be split into, namely: 1) data transfer from a host CPU to the DDR memory of each FPGA; 2) data transfer from FPGA DDR to FPGA on-chip memory; 3) kernel computation on the FPGA; 4) data transfer from FPGA on-chip memory to FPGA DDR; and 5) data transfer from FPGA DDR to host. Finding the optimal solution using a mixed-integer nonlinear programming (MINLP) solver is often highly inefficient. Hence, we provide a fast heuristic method that according to our experiments can be much more efficient than the MINLP solver and finds comparable results. For larger problems (more CNN layers), our heuristic method can quickly find (several thousand times faster) much better solutions than the MINLP solver, even if we run the latter for a very long time. [ABSTRACT FROM AUTHOR]

Multi-FPGA platforms like Amazon Web Services F1 are perfect to accelerate multi-kernel pipelined applications, like Convolutional Neural Networks (CNNs). To reduce energy consumption, we propose to upload at runtime the best power-optimized CNN implementation for a given throughput constraint. Our design method gives the best number of parallel instances of each kernel, their allocation to the FPGAs, the number of powered-on FPGAs and their clock frequency. This is obtained by solving a mixed-integer, non-linear optimization problem that models power and performance of each component, as well as the duration of the computation phases—data transfer between a host CPU and the FPGA memory (typically DDR), data transfer between DDR and FPGA, and FPGA computation. The results show that the power saved compared to simply clock gating the fastest implementation is obviously very high, but it is also much more significant than simply scaling the frequency of the fastest implementation or replicating the slowest implementation on multiple FPGAs. [ABSTRACT FROM AUTHOR]

Low cost, ubiquitous, tagless, and privacy aware indoor monitoring is essential to many existing or future applications, such as assisted living of elderly persons. We explore how well different types of neural networks in basic configurations can extract location and movement information from noisy experimental data (with both high-pitch and slow drift noise) obtained from capacitive sensors operating in loading mode at ranges much longer than the diagonal of their plates. Through design space exploration, we optimize and analyze the location and trajectory tracking inference performance of multilayer perceptron (MLP), autoregressive feedforward, 1D Convolutional (1D-CNN), and Long-Short Term Memory (LSTM) neural networks on experimental data collected using four capacitive sensors with 16 cm $\times$ 16 cm plates deployed on the boundaries of a 3 m $\times$ 3 m open space in our laboratory. We obtain the minimum error using a 1D-CNN [0.251 m distance Root Mean Square Error (RMSE) and 0.307 m Average Distance Error (ADE)] and the smoothest trajectory inference using an LSTM, albeit with higher localization errors (0.281 m RMSE and 0.326 m ADE). 1D Convolutional and window-based neural networks have best inference accuracy and smoother trajectory reconstruction. LSTMs seem to infer best the person movement dynamics. [ABSTRACT FROM AUTHOR]

Although useful for many applications, the practical use of tagless remote human identification is often hampered by privacy, usability, reliability, or cost concerns. In this paper, we explore the use of capacitive sensors, which appear to address most of these concerns, to identify different persons based on the unique electric and dielectric properties of their bodies given by their specific tissue composition. We present experimental results obtained by measuring the capacitance of a 16~\text cm\times 16~\text cm transducer plate 70 cm in front of different human bodies at different frequencies in the 5–160-kHz range. The measurements show clearly distinct signatures of capacitance variation with frequency for each person in the experiment, even after accounting for capacitance variations due to different body masses or physical dimensions. This paper focuses on the contactless identification of human body using capacitive sensors in smart home environments. [ABSTRACT FROM PUBLISHER]

Wireless sensor networks (WSNs) are a key part of the Internet of Things vision which aims at bridging together the physical and the digital worlds in several application domains. In the building automation field, WSNs are widely adopted for energy optimization, safety and security purposes, and could greatly benefit from existing information already available in preexisting building information models (BIMs). Such BIMs are normally developed during the building design phase and reused continuously during the construction and operation phases of the building life-cycle. In current deployments, however, due to lack of interoperability such information (e.g., walls geometry, materials, and so on) must often be recollected and reinputed by WSN commissioning specialists. Open development and planning tools and methodologies can play a key role in fostering interoperability and convergence of BIM and WSN systems. This paper aims at demonstrating how increased interoperability between WSN development-support tools and BIM systems could provide advantages to developers, integrators, domain specialist, and BIM users. The methodology is validated by applying a newly proposed tool exploiting BIM interoperability to support the planning of the topology of a WSN based on ultrawideband technologies. The proposed approach is evaluated through a small-scale experimentation held in a historical building in Torino (Italy). [ABSTRACT FROM PUBLISHER]

In this paper, we present an interactive design tool that can assist rapid prototyping and deployment of wireless sensor networks for building automation systems. We argue that it is possible to design networks that are more resilient to failures and have longer lifetime if the behavior of routing algorithms (RAs) is taken into account at design time. Resiliency can be increased by algorithmically adding redundancy to the network at locations where it can be maximally leveraged by RAs during operation. Lifetime can be increased by placing routers where they are most needed according to the expected data traffic patterns to improve the quality of the transmission. The network synthesis problem is formulated as an optimization problem. We propose a mixed-integer linear program to solve it exactly and a polynomial-time heuristic that returns close-to-optimal results in a shorter time. We analyze the performance of the designed networks by using OPNET simulation. Results show that our tool can assist in designing sensor networks that have high throughput and consume power efficiently. [ABSTRACT FROM PUBLISHER]