Browse Topic: Neural networks
ABSTRACT New generations of ground vehicles are required to perform tasks with an increased level of autonomy. Autonomous navigation and Artificial Intelligence on the edge are growing fields that require more sensors and more computational power to perform these missions. Furthermore, new sensors in the market produce better quality data at higher rates while new processors can increase substantially the computational power. Therefore, near-future ground vehicles will be equipped with large number of sensors that will produce data at rates that has not been seen before, while at the same time, data processing power will be significantly increased. This new scenario of advanced ground vehicles applications and increase in data amount and processing power, has brought new challenges with it: low determinism, excessive power needs, data losses and large response latency. In this article, a novel approach to on-board artificial intelligence (AI) is presented that is based on state-of-the
ABSTRACT As the Army leverages Prognostic and Predictive Maintenance (PPMx) models to migrate ground vehicle platforms toward health monitoring and prescriptive maintenance, the need is imminent for a pipeline to quickly and constantly move operational and maintenance data off the platform, through analytic models, and push the insights gained back out to the edge. This process will reduce data-to-decision time and operation and sustainment costs while increasing reliability for the platform and situational awareness for analysts, subject matter experts, maintainers, and operators. The US Army Ground Vehicle Systems Center (GVSC) is collaborating with The US Army Engineer Research and Development Center (ERDC) to develop a system of systems approach to stream operational and maintenance data to appropriate computing resources, collocating the data with DoD High-Performance Computing (HPC) processing capabilities where appropriate, then channeling the generated insights to maintainers
ABSTRACT This paper describes the use of neural networks to enhance simulations for subsequent training of anomaly-detection systems. Simulations can provide edge conditions for anomaly detection which may be sparse or non-existent in real-world data. Simulations suffer, however, by producing data that is “too clean” resulting in anomaly detection systems that cannot transition from simulated data to actual conditions. Our approach enhances simulations using neural networks trained on real-world data to create outputs that are more realistic and variable than traditional simulations. Citation: P.Feldman, “Training robust anomaly detection using ML-Enhanced simulations”, In Proceedings of the Ground Vehicle Systems Engineering and Technology Symposium (GVSETS), NDIA, Novi, MI, Aug. 11-13, 2020
ABSTRACT To advance development of the off-road autonomous vehicle technology, software simulations are often used as virtual testbeds for vehicle operation. However, this approach requires realistic simulations of natural conditions, which is quite challenging. Specifically, adverse driving conditions, such as snow and ice, are notoriously difficult to simulate realistically. The snow simulations are important for two reasons. One is mechanical properties of snow, which are important for vehicle-snow interactions and estimation of route drivability. The second one is simulation of sensor responses from a snow surface, which plays a major role in terrain classification and depends on snow texture. The presented work describes an overview of several approaches for realistic simulation of snow surface texture. The results indicate that the overall best approach is the one based on the Wiener–Khinchin theorem, while an alternative approach based on the Cholesky decomposition is the second
ABSTRACT Due to the high complexity of modern internal combustion engines and powertrain systems, the proper calibration of the electronic control unit’s (ECU) parameters has a strong impact on project targets like fuel consumption, emissions and drivability, as well as development costs and project duration. Simulation methods representing the system behavior with a model can support the calibration process considerably. However, standard physics-based models are often not able to describe all effects with sufficient accuracy, or the effort to set up a detailed model is too high. Physics-based models can also have a high computational demand, so that their simulation is not real-time capable. More suited for ECU calibration are data-driven models, combined with Design of Experiment (DoE). The system to be calibrated is identified with few specific test bench or vehicle measurements. From these measurements, a mathematical regression model is built. This paper describes recently
Butterflies can see more of the world than humans, including more colors and the field oscillation direction, or polarization, of light. This special ability enables them to navigate with precision, forage for food, and communicate with one another. Other species, like the mantis shrimp, can sense an even wider spectrum of light, as well as the circular polarization, or spinning states, of light waves. They use this capability to signal a “love code,” which helps them find and be discovered by mates
The integration of Artificial Intelligence (AI) and Machine Learning (ML) technologies has significantly changed various industries. This study demonstrates the application of a Convolutional Neural Network (CNN) model in Computational Fluid Dynamics (CFD) to predict the drag coefficient of a complete vehicle profile. We have developed a design advisor that uses a custom 3D CNN with a U-net architecture in the DEP MeshWorks environment to predict drag coefficients (Cd) based on car shapes. This model understands the relationship between car shapes and air drag coefficients calculated using computational fluid dynamics (CFD). The AI/ML-based design advisor feature has the potential to significantly decrease the time required for predicting drag coefficients by conducting CFD calculations. During the initial development phase, it will serve as an efficient tool for analyzing the correlation between multiple design proposals and aerodynamic drag forces within a short time frame
ERRATUM
The industrial internet of things (IIoT) is the nervous system in manufacturing facilities worldwide, with programmable logic controllers (PLCs) serving as its vital synapses. This digital neural network is transforming isolated machines into interconnected ecosystems of unprecedented intelligence and efficiency. PLCs have evolved from simple control devices into sophisticated nodes in a vast, responsive network
Tracking of energy consumption has become more difficult as demand and value for energy have increased. In such a case, energy consumption should be monitored regularly, and the power consumption want to be reduced to ensure that the needy receive power promptly. Our objective is to identify the energy consumption of an electric vehicle from battery and track the daily usage of it. We have to send the data to both the user and provider. We have to optimize the power usage by using anomaly detection technique by implementing deep learning algorithms. Here we are going to employ a LSTM auto-encoder algorithm to detect anomalies in this case. Estimating the power requirements of diverse locations and detecting harmful actions are critical in a smart grid. The work of identifying aberrant power consumption data is vital and it is hard to assure the smart meter’s efficiency. The LSTM auto-encoder neural network technique is used here for predicting power consumption and to detect anomalies
An approach to performance assessment and evaluation of robustness' of delivered AI (artificial intelligence) solutions (e.g., Aided Target Detection and Recognition (AiTDR) or ground vehicle autonomy behaviors) is presented. The initial development assumes the AI solution is delivered as a blackbox solution taking specified inputs and delivering output or outputs per stated requirements. The methods developed seek to not only confirm that requirements are met but to confirm performance in a manner that provides supporting evidence of which input information contributed to the AI output. Methods are developed for both AI solutions that take a single sample input as well as AIs that take a sequence of inputs to produce an output. Examples are included for convolutional neural networks (CNN). Planned extensions to the blackbox methods are discussed. Extensions include adaptations to support Residual Network (ResNet) solutions, evaluation of solution robustness when hidden layer
Items per page:
50
1 – 50 of 1210