Abstract

The plants of nano aerial vehicles (NAVs) are inherently unstable. Hence, a NAV needs a flight controller to accomplish a mission. Furthermore, the sensing and computational capabilities of NAV's autopilot hardware are limited. Hence, the implementation of the full state feedback controllers with gain scheduling is difficult. This paper proposes a flight controller scheme that consists of two parts: a Simultaneously Stabilizing Output Feedback Linear (SSOFL) controller and a Proximal Policy Optimization (PPO) deep reinforcement learning agent, which is connected in parallel to the SSOFL controller. In this scheme, the single SSOFL controller provides stabilization and nominal tracking performance to the NAV throughout its flight envelope by accomplishing simultaneous stabilization (SS). Additionally, the PPO agent is trained using the closed-loop (CL) nonlinear plant with this SSOFL controller to enhance the tracking performance. The effectiveness of the proposed flight controller scheme is verified using the six-degree-of-freedom nonlinear simulations of the fixed-wing nano aerial vehicle.

Abstract

This paper proposes a Reinforcement Learning (RL) based technique to develop a simple neural network controller for the task of waypoint navigation in quadrotors. In this paper, the application of Twin Delayed Deep Deterministic (TD3) Policy Gradient algorithm for high and low-level control implementation for quadrotors is discussed. The proposed methods are tested on high fidelity Gym-Pybullet-Drones simulator. The effectiveness of the methods developed is validated through numerical simulations. The simulation results indicate that both control policies are successful in navigating through the assigned waypoint, with the low-level controller being accurate in the nominal flight conditions. In the presence of disturbance inputs, the high-level controller performs better when compared to the low-level controller.

Abstract

Deep learning has shown remarkable performances in image classification, including those of plants and leaves. However, high-performing networks in terms of accuracy may not be using the salient regions for making the prediction and could be prone to biases. This work proposes a neural network architecture incorporating handcrafted features and fusing them with the learned features. Using hybrid features provides better control and understanding of the feature space while leveraging deep learning capabilities. Furthermore, a new IoU-based metric is introduced to assess the CNN-based classifier’s performance based on the regions focused on making predictions. Experiments over multiple leaf disease datasets demonstrate the performance improvement with the model using hybrid features. Classification using hybrid features performed better than the baseline models in terms of P@1 and also on the IoU …

Abstract

Note-taking methods and devices have improved tremendously over the past few decades, and people are finding new ways to write notes and take photos. Automatic extraction, recognition, and retrieval are necessary to process the huge chunk of digitized document data. However, an important step in all of these pipelines is the pre-processing step, mainly image enhancement or clean-up, which enhances the text regions and suppresses the non-text regions. In this article, we look at the problem of image enhancement or clean-up on one such important class of images (i.e., mobile captured glass board images). We present a simple yet efficient algorithm using the concepts of classical image processing techniques to solve the problem, and the obtained results are promising in comparison to the Office Lens.

Abstract

Wall-Frames, the recommended structural system for lifeline buildings is expected to have satisfactory and desirable seismic performance, especially if located in high seismic regions. In this work, fiber modeling approach is used to model inelasticity in walls and moment frames in a 6 storey-3-bay 2D wall-frame system. Displacement-controlled nonlinear static pushover analyses results from commercial software PERFORM3D suggest, the fiber type inelasticity in numerical models help predict nonlinear static behavior of study wall-frames reasonably well, compared to behavior of wall-frames with lumped inelasticity. In particular, increasing wall plan-aspect ratio enhances earthquake resistant virtues of wallframes, namely stiffness, strength, and ductility. Further, early yielding is observed in stiffer wall-frames and reasonable ductility achieved in all study wall-frames. Alongside, limit states of structural damages are also monitored to grade the damages and in turn the seismic performance of wall-frames. In addition, results obtained from nonlinear static analyses are confirmed by performing nonlinear time history analyses of study wall-frames, towards quantifying the critical earthquake resistant virtues—results and investigations from the study is a precursor towards recommending design measures to ensure postearthquake functionality of lifeline buildings.

Abstract

Wall-Frames, the recommended structural system for lifeline buildings is expected to have satisfactory and desirable seismic performance, especially if located in high seismic regions. In this work, fiber modeling approach is used to model inelasticity in walls and moment frames in a 6 storey-3-bay 2D wall-frame system. Displacement-controlled nonlinear static pushover analyses results from commercial software PERFORM3D suggest, the fiber type inelasticity in numerical models help predict nonlinear static behavior of study wall-frames reasonably well, compared to behavior of wall-frames with lumped inelasticity. In particular, increasing wall plan-aspect ratio enhances earthquake resistant virtues of wall-frames, namely stiffness, strength, and ductility. Further, early yielding is observed in stiffer wall-frames and reasonable ductility achieved in all study wall-frames. Alongside, limit states of structural damages are also monitored to grade the damages and in turn the seismic performance of wall-frames. In addition, results obtained from nonlinear static analyses are confirmed by performing nonlinear time history analyses of study wall-frames, towards quantifying the critical earthquake resistant virtues — results and investigations from the study is a precursor towards recommending design measures to ensure post-earthquake functionality of lifeline buildings. Keywords: Fiber modeling, Nonlinear static analysis, Limit states, Structural damages

Abstract

An FFT is an essential algorithm for radar signal processing in a radar system. Due to increase in computational power of FPGAs, it is possible to perform FFT operation onboard in an airborne vehicle. However, the FPGA resources have become a limitation for processing real-time signals using conventional methods. To address this issue, we have proposed a parallel pipelined FFT architecture that can achieve very high throughput with very low latency, making it capable of processing real-time continuous data. This architecture is implemented in a radar system, which works from L band to Ku band. In this radar system, the received RF signal is downconverted into an IF signal of 1 GHz frequency with a 500 MHz bandwidth and converted to digital data using a 10-bit ADC. On the converted digital data, a 512-point FFT is implemented on a Xilinx Virtex-7 XC7VX485T FPGA using 8 parallel channels with 64 data frames and is compared with the conventional IP core-based architecture. The proposed architecture takes 1.307µs to implement FFT, which is 5.15 times faster than the IP core-based architecture and requires fewer arithmetic computations. The overall total number of complex multiplications, complex additions, multipliers & adders were reduced by 10.42%, 30.64%, 10.42% & 23.90% respectively. Apart from very low latency and fewer arithmetic operations, the proposed parallel FFT architecture achieved a throughput of 1.350 Giga Samples per second (Gsps). Index Terms—Fast Fourier Transform (FFT), DSP, Radar Systems, Parallel Architecture, FPGA.

Abstract

Inverse sqrt and sqrt function have numerous applications in linear algebra and machine learning such as vector normalisation, eigenvalue computation, dimensionality reduction, clustering, etc. This paper presents a method to approximate and securely perform the inverse sqrt function using CKKS homomorphic encryption scheme. Since the CKKS homomorphic scheme allows only computation of polynomial functions, we propose a method to approximate the inverse sqrt function polynomially. In the end, we provide an implementation of our method for the inverse sqrt function.

Abstract

Air pollution is a concern to the health of all living beings. It is essential to check on the quality of air in the surroundings. This article presents an IoT-based real-time air quality index (AQI) estimation technique using images and weather sensors on Indian rods. A mixture of image features, i.e., traffic density, visibility, and sensor features, i.e., temperature and humidity, were used to predict the AQI. Object detection and localization-based Deep Learning (DL) method along with image processing techniques were used to extract image features while an Machine Learning (ML) model was trained on those features to estimate the AQI. In order to conduct this experiment, a dataset containing 5048 images along with co-located AQI values across different seasons was collected by driving on the roads of Hyderabad city in India. The experimental results report an overall accuracy of 82% for AQI prediction. Index Terms—Air Quality, CNN, Edge computing, Machine Learning.

Abstract

How does the brain represent different modes of information? Can we design a system that can automatically understand what the user is thinking? We can make progress towards answering such questions by studying brain recordings from devices such as functional magnetic resonance imaging (fMRI). The brain encoding problem aims to automatically generate fMRI brain representations given a stimulus. The brain decoding problem is the inverse problem of reconstructing the stimuli given the fMRI brain representation. Both the brain encoding and decoding problems have been studied in detail in the past two decades and the foremost attraction of studying these solutions is that they serve as additional tools for basic research in cognitive science and cognitive neuroscience. Recently, inspired by the effectiveness of deep learning models for natural language processing and computer vision, such models have been applied for neuroscience as well. In this tutorial, we plan to discuss different kinds of stimulus representations, and popular encoding and decoding architectures in detail. The tutorial will provide a working knowledge of the state of the art methods for encoding and decoding, a thorough understanding of the literature, and a better understanding of the benefits and limitations of encoding/decoding with deep learning. Encoding models that accurately predict brain activity have several practical applications in evaluation and diagnosis of neurological conditions and thus also help designing therapies for brain damage. Invertible encoding models enable principled formulation of brain decoding models which in turn are useful for designing brain-machine or brain-computer interfaces. Recent advances in the use of pretrained deep network models enable us to use them as priors for brain decoding tasks. Deep learning models are useful for improving accuracy but also offer the flexibility of decoding across a gamut of tasks and domains.