Search Results

Abstract Lane changes are stressful maneuvers for drivers, particularly during high-speed traffic flows. However, modeling driver’s lane-changing decision and implementation process is challenging due to the complexity and uncertainty of driving behaviors. To address this issue, this article presents a personalized Lane-Changing Model (LCM) for Advanced Driver Assistance System (ADAS) based on deep learning method. The LCM contains three major computational components. Firstly, with abundant inputs of Root Residual Network (Root-ResNet), LCM is able to exploit more local information from the front view video data. Secondly, the LCM has an ability of learning the global spatial-temporal information via Temporal Modeling Blocks (TMBs). Finally, a two-layer Long Short-Term Memory (LSTM) network is used to learn video contextual features combined with lane boundary based distance features in lane change events.

Abstract Real-time reconstruction of 3D environment attributed with semantic information is significant for a variety of applications, such as obstacle detection, traffic scene comprehension and autonomous navigation. The current approaches to achieve it are mainly using stereo vision, Structure from Motion (SfM) or mobile LiDAR sensors. Each of these approaches has its own limitation, stereo vision has high computational cost, SfM needs accurate calibration between a sequences of images, and the onboard LiDAR sensor can only provide sparse points without color information. This paper describes a novel method for traffic scene semantic segmentation by combining sparse LiDAR point cloud (e.g. from Velodyne scans), with monocular color image. The key novelty of the method is the semantic coupling of stereoscopic point cloud with color lattice from camera image labelled through a Convolutional Neural Network (CNN).

Abstract The main focus of this article is the online estimation of the terminal airspace sector capacity from the Air Traffic Controller 0ATC) dynamical neural model using Neural Partial Differentiation (NPD) with permissible safe separation and affordable workload. For this purpose, a primarily neural model of a multi-input-single-output (MISO) ATC dynamical system is established, and the NPD method is used to estimate the model parameters from the experimental data. These estimated parameters have a less relative standard deviation, and hence the model validation results show that the predicted neural model response is well matched with the intervention of the ATC workload. Moreover, the proposed neural network-based approach works well with the experimental data online as it does not require the initial values of model parameters, which are unknown in practice.

Abstract This paper presents a machine learning application of the force/torque sensor in a human-robot collaborative manufacturing scenario. The purpose is to simplify the programming for physical interactions between the human operators and industrial robots in a hybrid manufacturing cell which combines several robotic applications, such as parts manipulation, assembly, sealing and painting, etc. A multiclass classifier using Light Gradient Boosting Machine (LightGBM) is first introduced in a robotic application for discriminating five different contact states w.r.t. the force/torque data. A systematic approach to train machine-learning based classifiers is presented, thus opens a door for enabling LightGBM with robotic data process. The total task time is reduced largely because force transitions can be detected on-the-fly. Experiments on an ABB force sensor and an industrial robot demonstrate the feasibility of the proposed method.

Abstract In an automated machining process, monitoring the conditions of the tool is essential for deciding to replace or repair the tool without any manual intervention. Intelligent models built with sensor information and machine learning techniques are predicting the condition of the tool with good accuracy. In this study, statistical models are developed to identify the conditions of the abrasive grinding wheel using the Acoustic Emission (AE) signature acquired during the surface grinding operation. Abrasive grinding wheel conditions are identified using the abrasive wheel wear plot established by conducting experiments. The piezoelectric sensor is used to capture the AE from the grinding process, and statistical features of the abrasive wheel conditions are extracted in time and wavelet domains of the signature. Machine learning algorithms, namely, Classification and Regression Trees (CART) and Support Vector Classifiers (SVC), are used to build statistical models.

Abstract Management of expressway networks has been mainly focused on defect management without looking at the correlations with accidental risks. This causes unsustainability in expressway infrastructure maintenance since such defects may not be a contributing factor toward public safety. Thus it is necessary to incorporate accidental events for decision-making in infrastructure management. This study has developed a novel approach to machine learning (ML) that incorporates actual primary data from the last 10 years of single-vehicle accidents (SVA) by collisions with motorway facilities, or so-called single-vehicle collisions with fixed objects. The ML is firstly aimed at identifying the influential factors of SVA in relation to finding effective countermeasures for accidents by integrating the correlation analysis, multiple regression analysis, and ML techniques. The study reveals that wet pavement conditions have a significant effect on SVA.

Empirical models for engine-out oxides of Nitrogen (NOx) and smoke emissions have been developed for the purpose of minimizing transient emissions while maintaining transient response. Three major issues have been addressed: data acquisition, data processing and modeling method. Real and virtual transient parameters have been identified for acquisition. Accounting for the phase shift between transient engine events and transient emission measurements has been shown to be very important to the quality of model predictions. Several methods have been employed to account for the transient transport delays and sensor lags which constitute the phase shift. Finally several different empirical modeling methods have been used to determine the most suitable modeling method for transient emissions. These modeling methods include several kinds of neural networks, global regression and localized regression.

This recommended practice applies to vibration monitoring systems for rotorcraft and fixed-wing drive trains, airframes, propulsion systems, electric power generators, and flight control systems. It addresses all aspects of metrics, including what to measure, how to measure, and how to evaluate the results.

This study outlines an approach for speeding up the simulation of the dynamic response of vehicle models that include hysteretic nonlinear tire components. The method proposed replaces the hysteretic nonlinear tire model with a surrogate model that emulates the dynamic response of the actual tire. The approach is demonstrated via a dynamic simulation of a quarter vehicle model. In the proposed methodology, training information generated with a reduced number of harmonic excitations is used to construct the tire hysteretic force emulator using a Neural Network (NN) element. The proposed approach has two stages: a learning stage, followed by an embedding of the learned model into the quarter car model. The learning related main challenge stems from the attempt to capture with the NN element the behavior of a hysteretic element whose response depends on its loading history.

Hat sections made of steel are frequently encountered in automotive body structural components such as front rails. These components can absorb significant amount of impact energy during collisions thereby protecting occupants of vehicles from severe injury. In the initial phase of vehicle design, it will be prudent to incorporate the sectional details of such a component based on an engineering target such as peak load, mean load, energy absorption, or total crush, or a combination of these parameters. Such a goal can be accomplished if efficient and reliable data-based models are available for predicting the performance of a section of given geometry as alternatives to time-consuming and detailed engineering analysis typically based on the explicit finite element method.

The debate surrounding roof deformation and occupant injury potential has existed in the automotive community for over 30 years. In analysis of real-world rollovers, assessment of roof deformation and occupant compartment space starts with the post-accident roof position. Dynamic movement of the roof structure during a rollover sequence is generally acknowledged but quantification of the dynamic roof displacement has been limited. Previous assessment of dynamic roof deformation has been generally limited to review of the video footage from staged rollover events. Rollover testing for the evaluation of injury potential has typically been studied utilizing instrumented test dummies, on-board and off-board cameras, and measurements of residual crush. This study introduces an analysis of previously undocumented real-time data to be considered in the evaluation of the roof structure's dynamic behavior during a rollover event.

In this research, the interior noise of a passenger car was measured, and the sound quality metrics including sound pressure level, loudness, sharpness, and roughness were calculated. An artificial neural network was designed to successfully apply on automotive interior noise as well as numerous different fields of technology which aim to overcome difficulties of experimentations and save cost, time and workforce. Sound pressure level, loudness, sharpness, and roughness were estimated by using the artificial neural network designed by using the experiment values. The predicted values and experiment results are compared. The comparison results show that the realized artificial intelligence model is an appropriate model to estimate the sound quality of the automotive interior noise. The reliability value is calculated as 0.9995 by using statistical analysis.

Collision avoidance systems for rear-end collisions have been researched and developed. It is necessary to activate collision warnings and automatic braking systems with appropriate timing determined by a monitoring system of a driver's braking action. Although there are various systems to monitor driving behavior, this study aims to create a monitoring system using a driver model. This study was intended to construct a model of a driver's braking action with the Time Delay Neural Network (TDNN). An experimental scenario focuses on rear-end collisions on a highway, such as the driver of a host vehicle controlling the brake to avoid a collision into a leading vehicle in a stationary condition caused by a traffic jam. In order to examine the accuracy of the TDNN model, this study used four parameters: the number of learning, the number of neurons in the hidden layer, the sampling time with 0.01 second as a minimum value, and the number of the delay time.

Cost reduction in the automotive industry becomes a widely-adopted operational strategy not only for Original Equipment Manufacturers (OEMs) that take cost leader generic corporation strategy, but also for many OEMs that take differentiation generic corporation strategy. Since differentiation generic strategy requires an organization to provide a product or service above the industry average level, a premium is typically included in the tag price for those products or services. Cost reduction measures could increase risks for the organizations that pursue differentiation strategy. Although manufacturers in the automotive industry dramatically improved production efficiency in past ten years, they are still facing the pressure of cost control. The big challenge in cost control for automakers and suppliers is increasing prices of raw materials, energy and labor costs. These costs create constraints for the traditional economic expansion model.

Laser cladding is a method of material deposition through which a powdered or wire feedstock material is melted and consolidated by use of a laser to coat part of a substrate. Determining the parameters to fabricate the desired clad bead geometry for various configurations is problematic as it involves a significant investment of raw materials and time resources, and is challenging to develop a predictive model. The goal of this research is to develop an experimental methodology that minimizes the amount of data to be collected, and to develop a predictive model that is accurate, adaptable, and expandable. To develop the predictive model of the clad bead geometry, an integrated five-step approach is presented. From the experimental data, an artificial neural network model is developed along with multiple regression equations.

This paper proposes a new method of predicting the sound absorption performance of polymer wool using artificial neural networks (ANN) model. Some important parameters of the proposed model have been adjusted to best fit the non-linear relationship between the input data and output data. What's more, the commonly used multiple non-linear regression model is built to compare with ANN model in this study. Measurements of the sound absorption coefficient of polymer wool based on transfer function method are also performed to determine the sound absorption performance according to GB/T18696. 2-2002 and ISO10534- 2: 1998 (E) standards. It is founded that predictions of the new model are in good agreement with the experiment results.

On-board fuel identification is important to ensure engine safe operation, similar power output, fuel economy and emissions levels when different fuels are used. Real-time detection of physical and chemical properties of the fuel requires the development of identifying techniques based on a simple, non-intrusive sensor. The measured crankshaft speed signal is already available on series engine and can be utilized to estimate at least one of the essential combustion parameters such as peak pressure and its location, rate of cylinder pressure rise and start of combustion, which are an indicative of the ignition properties of the fuel. Using a dynamic model of the crankshaft numerous methods have been previously developed to identify the fuel type but all with limited applications in terms of number of cylinders and computational resources for real time control.

One of the hardest tasks involving wind tunnel characterization is to determine the air-flow condition inside the test section. The Log-Tchebycheff method and the Equal Area method allow calculation of local velocities from measured differential pressures on rectangular and circular ducts. However, these two standard methods for air flow measurement are limited by the number of accurate pressure readings by the Pitot tube. In this paper, a new approach is presented for wind tunnel calibrations. This approach is based on a limited number of dynamic pressure measurements and a predictive technique using Neural Network (NN). To optimize the NN, the extended great deluge (EGD) algorithm is used. Wind tunnel testing involves a large number of variables such as wind direction, velocity, rate flow, turbulence characteristics, temperature variation and pressure distribution on airfoils.

When the demand for either a region of airspace or an airport approaches or exceeds the available capacity, miles-in-trail (MIT) restrictions are the most frequently issued traffic management initiatives (TMIs) that are used to mitigate these imbalances. Miles-in-trail operations require aircraft in a traffic stream to meet a specific inter-aircraft separation in exchange for maintaining a safe and orderly flow within the stream. This stream of aircraft can be departing an airport, over a common fix, through a sector, on a specific route or arriving at an airport. This study begins by providing a high-level overview of the distribution and causes of arrival MIT restrictions for the top ten airports in the United States. This is followed by an in-depth analysis of the frequency, duration and cause of MIT restrictions impacting the Hartsfield-Jackson Atlanta International Airport (ATL) from 2009 through 2011.

A twist beam rear suspension system is modeled, analyzed and optimized in this paper. An ADAMS model is established based on the REC (Rigid-Elastic Coupling) Theory, which is verified by FEM (Finite Element Method) approach, the effects of the geometric parameters on the twist beam suspension performance are investigated. In order to increase the calculation efficiency and improve the simulation accuracy, a neural network model and NSGA II (Non-dominated Sorting Genetic Algorithm II) are adopted to conduct a multi-objective optimization on a twist beam rear suspension system.