Deep Learning (DL) algorithms are emerging as a key alternative to computationally expensive CFD simulations. However, state-of-the-art DL approaches require large and high-resolution training data to learn accurate models. The size and availability of such datasets are a major limitation for the development of next-generation data-driven surrogate models for turbulent flows. This paper introduces SURFNet, a transfer learning-based super-resolution flow network. SURFNet primarily trains the DL model on low-resolution datasets and transfer learns the model on a handful of high-resolution flow problems - accelerating the traditional numerical solver independent of the input size. We propose two approaches to transfer learning for the task of super-resolution, namely one-shot and incremental learning. Both approaches entail transfer learning on only one geometry to account for fine-grid flow fields requiring 15x less training data on high-resolution inputs compared to the tiny resolution (64x256) of the coarse model, significantly reducing the time for both data collection and training. We empirically evaluate SURFNet's performance by solving the Navier-Stokes equations in the turbulent regime on input resolutions up to 256x larger than the coarse model. On four test geometries and eight flow configurations unseen during training, we observe a consistent 2-2.1x speedup over the OpenFOAM physics solver independent of the test geometry and the resolution size (up to 2048x2048), demonstrating both resolution-invariance and generalization capabilities. Our approach addresses the challenge of reconstructing high-resolution solutions from coarse grid models trained using low-resolution inputs (super-resolution) without loss of accuracy and requiring limited computational resources.

Proximity perception is a technology that has the potential to play an essential role in the future of robotics. It can fulfill the promise of safe, robust, and autonomous systems in industry and everyday life, alongside humans, as well as in remote locations in space and underwater. In this survey paper, we cover the developments of this field from the early days up to the present, with a focus on human-centered robotics. Here, proximity sensors are typically deployed in two scenarios: first, on the exterior of manipulator arms to support safety and interaction functionality, and second, on the inside of grippers or hands to support grasping and exploration. Starting from this observation, we propose a categorization for the approaches found in the literature. To provide a basis for understanding these approaches, we devote effort to present the technologies and different measuring principles that were developed over the years, also providing a summary in form of a table. Then, we show the diversity of applications that have been presented in the literature. Finally, we give an overview of the most important trends that will shape the future of this domain.

We present a novel implementation of classification using the machine learning / artificial intelligence method called boosted decision trees (BDT) on field programmable gate arrays (FPGA). The firmware implementation of binary classification requiring 100 training trees with a maximum depth of 4 using four input variables gives a latency value of about 10 ns, independent of the clock speed from 100 to 320 MHz in our setup. The low timing values are achieved by restructuring the BDT layout and reconfiguring its parameters. The FPGA resource utilization is also kept low at a range from 0.01% to 0.2% in our setup. A software package called fwXmachina achieves this implementation. Our intended user is an expert of custom electronics-based trigger systems in high energy physics experiments or anyone that needs decisions at the lowest latency values for real-time event classification. Two problems from high energy physics are considered, in the separation of electrons vs. photons and in the selection of vector boson fusion-produced Higgs bosons vs. the rejection of the multijet processes.

Surface defect detection plays an increasingly important role in manufacturing industry to guarantee the product quality. Many deep learning methods have been widely used in surface defect detection tasks, and have been proven to perform well in defects classification and location. However, deep learning-based detection methods often require plenty of data for training, which fail to apply to the real industrial scenarios since the distribution of defect categories is often imbalanced. In other words, common defect classes have many samples but rare defect classes have extremely few samples, and it is difficult for these methods to well detect rare defect classes. To solve the imbalanced distribution problem, in this paper we propose TL-SDD: a novel Transfer Learning-based method for Surface Defect Detection. First, we adopt a two-phase training scheme to transfer the knowledge from common defect classes to rare defect classes. Second, we propose a novel Metric-based Surface Defect Detection (M-SDD) model. We design three modules for this model: (1) feature extraction module: containing feature fusion which combines high-level semantic information with low-level structural information. (2) feature reweighting module: transforming examples to a reweighting vector that indicates the importance of features. (3) distance metric module: learning a metric space in which defects are classified by computing distances to representations of each category. Finally, we validate the performance of our proposed method on a real dataset including surface defects of aluminum profiles. Compared to the baseline methods, the performance of our proposed method has improved by up to 11.98% for rare defect classes.

Convolutional Neural Networks achieve state-of-the-art accuracy in object detection tasks. However, they have large computational and energy requirements that challenge their deployment on resource-constrained edge devices. Object detection takes an image as an input, and identifies the existing object classes as well as their locations in the image. In this paper, we leverage the prior knowledge about the probabilities that different object categories can occur jointly to increase the efficiency of object detection models. In particular, our technique clusters the object categories based on their spatial co-occurrence probability. We use those clusters to design an adaptive network. During runtime, a branch controller decides which part(s) of the network to execute based on the spatial context of the input frame. Our experiments using COCO dataset show that our adaptive object detection model achieves up to 45% reduction in the energy consumption, and up to 27% reduction in the latency, with a small loss in the average precision (AP) of object detection.

Within electrical distribution networks, substation constraints management requires that aggregated power demand from residential users is kept within suitable bounds. Efficiency of substation constraints management can be measured as the reduction of constraints violations w.r.t. unmanaged demand. Home batteries hold the promise of enabling efficient and user-oblivious substation constraints management. Centralized control of home batteries would achieve optimal efficiency. However, it is hardly acceptable by users, since service providers (e.g., utilities or aggregators) would directly control batteries at user premises. Unfortunately, devising efficient hierarchical control strategies, thus overcoming the above problem, is far from easy. We present a novel two-layer control strategy for home batteries that avoids direct control of home devices by the service provider and at the same time yields near-optimal substation constraints management efficiency. Our simulation results on field data from 62 households in Denmark show that the substation constraints management efficiency achieved with our approach is at least 82% of the one obtained with a theoretical optimal centralized strategy.

Artificial intelligence is already capable of doing some incredibly useful things, like predicting flooding, diagnosing disease, and instantly translating languages. Advances in neural networks coupled with enormous computational power have allowed tech companies to create incrementally smarter AI models over the last decade. Jeff Dean, Google's AI chief, thinks we're just scratching the surface. Speaking at the TED conference in Monterey, California, this week, he revealed that Google is developing a nimble, multi-purpose AI that can perform millions of tasks. Called Pathways, Google's solution seeks to centralize disparate AI into one powerful, all-knowing algorithm.

Electrical workers check solar panels at a photovoltaic power station built in a fishpond in Haian in China's eastern Jiangsu province. Electrical workers check solar panels at a photovoltaic power station built in a fishpond in Haian in China's eastern Jiangsu province. As the world's top climate scientists released a report full of warnings this week, they kept insisting that the world still has a chance to avoid the worst effects of climate change. "It is still possible to forestall most of the dire impacts, but it really requires unprecedented, transformational change," said Ko Barrett, vice chair of the Intergovernmental Panel on Climate Change. "The idea that there still is a pathway forward, I think, is a point that should give us some hope."

Artificial intelligence, or AI, is a simulation of intelligent human behavior. It's a computer or system designed to perceive its environment, understand its behaviors, and take action. Consider self-driving cars: AI-driven systems like these integrate AI algorithms, such as machine learning and deep learning, into complex environments that enable automation. AI is estimated to create $13 trillion in economic value worldwide by 2030, according to a McKinsey forecast. That's because AI is transforming engineering in nearly every industry and application area.

Our team is proposing to run a full-scale energy demand response experiment in an office building. Although this is an exciting endeavor which will provide value to the community, collecting training data for the reinforcement learning agent is costly and will be limited. In this work, we examine how offline training can be leveraged to minimize data costs (accelerate convergence) and program implementation costs. We present two approaches to doing so: pretraining our model to warm start the experiment with simulated tasks, and using a planning model trained to simulate the real world's rewards to the agent. We present results that demonstrate the utility of offline reinforcement learning to efficient price-setting in the energy demand response problem.