In general, when a problem occurs in a component, various phenomena appear, and abnormal noise is one of them. The service technicians diagnose the noise through the analysis using hearing and equipment. Depending on their experiences, the analysis time and diagnosis accuracy vary widely. The newly developed AI-based diagnostic technology diagnoses parts that cause abnormal noises within seconds when a noise is input to the equipment. To create a learning model for diagnosis, we collected as many abnormal noises as possible from various parts, and selected good and bad data. This process is very important in the development of diagnostic techniques. Artificial intelligence was learned by deep learning with selected good data. This paper is about the technology that can diagnose the abnormal noises generated from the engine, transmission, drivetrain and PE (Power Electric) parts of the eco-friendly vehicle through the diagnosis model composed of various methods of deep learning.

Formula Society of Automotive Engineers (FSAE) International is a student design competition organized by SAE. The student design involves engineering and manufacturing a formula style racecar and evaluating its performance. Testing and validation of the vehicle is an integral part of the design and performance during the competition. At the collegiate level the drivers are at the amateur level. As a result, the human factor plays a significant role in the outcome of the dynamic events. In order to reduce the uncertainty factor and improve the general performance, driver training is necessary. Instead of overall performance of the driver based on individual lap, our current research focuses on the more detailed components of the driver’s actions throughout different sections of the lap. A complete lap consists of several components, such as, straight line acceleration and braking, max and min radius cornering, slalom or “S” movements, and bus stops or quick braking and turning. In order to evaluate the performance of each driver in each of these components, an AiM data acquisition system is mounted in…

Current artificial intelligence techniques for end to end driving of autonomous vehicles typically rely on a single form of learning or training processes along with a corresponding dataset or simulation environment. Relatively speaking, success has been shown for a variety of learning modalities in which it can be shown that the machine can successfully “drive” a vehicle. However, the realm of real-world driving extends significantly beyond the realm of limited test environments for machine training. This creates an enormous gap in capability between these two realms. With their superior neural network structures and learning capabilities, humans can be easily trained within a short period of time to proceed from limited test environments to real world driving. For machines though, this gap is guarded by at least two challenges: 1) machine learning techniques remain brittle and unable to generalize to a wide range of scenarios, and 2) effective training data that enhances generalization and generates the desired driving behavior. Further, each challenge can be computationally intensive on its own thereby exasperating the gap. Moreover, is has…

While machine learning in autonomous vehicles development has increased significantly in the past few years, the use of reinforcement learning (RL) methods has only recently been applied. Convolutional neural networks (CNNs) became common for their powerful object detection and identification and even provided end-to-end control of an autonomous vehicle. However, one of the requirements of a CNN is a large amount of labeled data to inform the neural network. While data is becoming more accessible, these networks are still sensitive to the format and collection environment which makes the use of others’ data more difficult. In contrast, RL develops solutions in a simulation environment by trial and error without labeled data. Our research expands upon previous research in RL and proximal policy optimization (PPO) and the application of these algorithms to 1/18th scale cars by expanding the application of this control strategy to a full-sized passenger vehicle. By using this method of unsupervised learning, our research demonstrates the ability to learn new control strategies while in a simulation environment without the need for large amounts…

This paper argues the importance of engineering heuristics and introduces an educational data-driven tool to help novice engineers develop their engineering heuristics more effectively. The main objective in engineering practice is to identify opportunities for improvement and apply methods to effect change. Engineers do so by applying ‘how to’ knowledge to make decisions and take actions. This ‘how to’ knowledge is encoded in engineering heuristics.In this paper, we describe a tool that aims to provide heuristic knowledge to users by giving them insight into heuristics applied by experts in similar situations. A repository of automotive data is transformed into a tool with powerful search and data visualization functionalities. The tool can be used to educate novice automotive engineers alongside the current resource intensive practices of teaching engineering heuristics through social methods such as an apprenticeship. The tool can do so by providing novices with powerful search and data visualization capabilities which will allow them to understand tradeoffs between vehicle attributes, to make assumptions from initial information, and to benchmark the vehicle design.

More than four decades ago the concept of zero defects was coined by Phillip Crosby. At that time it was only a vision, but today with the introduction of Artificial Intelligence in manufacturing it has become attainable. Since most mature manufacturing organizations have merged traditional quality philosophies and techniques, their processes generate only a few defects per million of opportunities. Therefore, detecting these rare quality events is one of the modern intellectual challenges posed by this industry. Process Monitoring for Quality is a big data-driven quality philosophy aimed at defect detection and empirical knowledge discovery. Detection is formulated as a binary classification problem, where the right machine learning, optimization and statistics techniques are applied to develop an effective predictive system. Manufacturing-derived data sets for binary classification of quality tend to be highly/ultra unbalanced, making it very difficult for the learning algorithms to learn the minority (defective) class. In this paper, the learning and deployment paradigm of Process Monitoring for Quality is presented, a discussion of how it interacts with traditional quality philosophies to enable the…

Rapid changes in the automotive industry, including the growth of advanced vehicle controls and autonomy, are driving the need for more dedicated proving ground spaces where these systems can be developed safely. To address this need, Kettering University has created the GM Mobility Research Center, a 21-acre proving ground located in Flint, Michigan at the former “Chevy in the Hole” factory location. Construction of a proving ground on this site represents a beneficial redevelopment of an industrial brownfield, as well as a significant expansion of the test facilities available at the campus of Kettering University. Test facilities on the site include a road course and a test pad, along with a building that has garage space, a conference room, and an indoor observation platform. All of these facilities are available to the students and faculty of Kettering University, along with their industrial partners, for the purpose of engaging in advanced transportation research and education. This work describes the history of the proving ground development and outlines its design. Special emphasis is placed on a detailed…

Forecasting the motion of the leading vehicle is a critical task for connected autonomous vehicles as it provides an efficient way to model the leading-following vehicle behavior and analyze the interactions. In this study, a personalized time-series modeling approach for leading vehicle trajectory prediction considering different driving styles is proposed. The method enables a precise, personalized trajectory prediction for leading vehicles with limited inter-vehicle communication signals, such as vehicle speed, acceleration, space headway, and time headway of the front vehicles. Based on the learning nature of human beings that a human always tries to solve problems based on grouping and similar experience, three different driving styles are first recognized based on an unsupervised clustering with a Gaussian Mixture Model (GMM). The GMM generates a specific driving style for each vehicle based on the speed, acceleration, jerk, time, and space headway features of the leading vehicle. Then, a personalized joint time-series modeling (JTSM) method based on the Long Short-Term Memory (LSTM) Recurrent Neural Network model (RNN) is proposed to predict the trajectory of the front vehicle.…

People through many years of experience, have developed a great intuitive sense for navigation and spatial awareness. With this intuition people are able to apply a nearly rules based approach to their driving. With a transition to autonomous driving, these intuitive skills need to be taught to the system which makes perception is the most fundamental and critical task. One of the major challenges for autonomous vehicles is accurately knowing the position of the vehicle relative to the world frame. Currently, this is achieved by utilizing expensive sensors such as a differential GPS which provides centimeter accuracy, or by using computationally taxing algorithms to attempt to match live input data from LiDARs or cameras to previously recorded data or maps. Within this paper an algorithm and accompanying hardware stack is proposed to reduce the computational load on the localization of the robot relative to a prior map. The principal of the software stack is to leverage deep learning and powerful filters to perform classification of landmark objects within a scan of the LiDAR. These landmarks…

Based on FOT data of a Chinese automobile company, this paper aims to study the practical role of lane departure warning system. The data of this automobile company collects a total of 32.29 hours of test data, including vehicle control, lane line and other relevant information, FOT data included both test groups and contrast groups. This paper designs research questions for the development purpose of LDW system: whether the LDW system can affect driver behavior or vehicle performance to improve road safety. To solve this problem, a hypothesis is proposed: due to the role of LDW system, in the test group and contrast group, the driving safety of the test group is higher than that of the benchmark group. According to the research hypothesis, three analysis indexes of the test are determined and defined: the number of road deviation, the time of road deviation and the maximum distance of road deviation, which are collectively referred to as safety and benefit indexes. Through data screening and processing, a total of 302 test conditions and 589 contrast…