Search Results

Machine learning methods, such as decision trees and deep neural networks, are becoming increasingly important and useful for data analysis in various scientific fields including dynamics and control, signal processing, pattern recognition, fluid mechanics, and chemical synthesis, etc. For future engine design and performance optimization, there is an urgent need for a robust predictive model which could capture the major combustion properties such as autoignition and flame propagation of multicomponent fuels under a wide range of engine operating conditions, without massive experimental measurement or computational efforts. It will be shown that these long-held limitations and challenges related to complex fuel combustion and engine research could be readily solved by implementing machine learning methods.

This work presents a method for estimating human body orientation using a combination of convolutional neural network (CNN) and stereo camera in real time. The approach uses the CNN model to predict certain human body keypoints then transforms these points into a 3D space using the stereo vision system to estimate the body orientations. The CNN module is trained to estimate the shoulders, the neck and the nose positions, detecting of three points is required to confirm human detection and provided enough data to translate the points into 3D space.

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.

This work presents a method to estimate the human body orientation using 2D images from a person view; the challenge comes from the variety of human body poses and appearances. The method utilizes OpenPose neural network as a human pose detector module and depth sensing module. The modules work together to extract the body orientation from 2D stereo images. OpenPose is proven to be efficient in detecting human body joints, defined by COCO dataset, OpenPose can detect the visible body joints without being affected by backgrounds or other challenging factors. Adding the depth data for each point can produce rich information to the process of 3D construction for the detected humans. This 3D point’s setup can tell more about the body orientation and walking direction for example. The depth module used in this work is the ZED camera stereo system which uses CUDA for high performance depth computation.

In the automotive industry, a Malfunction Indicator Light (MIL) is commonly employed to signify a failure or error in a vehicle system. To identify the root cause that has triggered a particular fault, a technician or engineer will typically run diagnostic tests and analyses. This type of analysis can take a significant amount of time and resources at the cost of customer satisfaction and perceived quality. Predicting an impending error allows for preventative measures or actions which might mitigate the effects of the error. Modern vehicles generate data in the form of sensor readings accessible through the vehicle’s Controller Area Network (CAN). Such data is generally too extensive to aid in analysis and decision making unless machine learning-based methods are used. This paper proposes a method utilizing a recurrent neural network (RNN) to predict an impending fault before it occurs through the use of CAN data.

An existing challenge in current state-of-the-art autonomous vehicles is the process of safely transferring control from autonomous driving mode to manual mode because the driver may be distracted with secondary tasks. Such distractions may impair a driver’s situational awareness of the driving environment which will lead to fatal outcomes during a handover. Current state-of-the-art vehicles notify a user of an imminent handover via auditory, visual, and physical alerts but are unable to improve a driver’s situational awareness before a handover is executed. The overall goal of our research team is to address the challenge of providing a driver with relevant information to regain situational awareness of the driving task. In this paper, we introduce a novel approach to estimating a driver’s visual focus of attention using a 2D RGB camera as input to a Multi-Input Convolutional Neural Network with shared weights. The system was validated in a realistic driving scenario.

Controller Area Network (CAN), the de facto standard for in-vehicle networks, has insufficient security features and thus is inherently vulnerable to various attacks. To protect CAN bus from attacks, intrusion detection systems (IDSs) based on advanced deep learning methods, such as Convolutional Neural Network (CNN) and Recurrent Neural Network (RNN), have been proposed to detect intrusions. However, those models generally introduce high latency, require considerable memory space, and often result in high energy consumption. To accelerate intrusion detection and also reduce memory requests, we exploit the use of Binarized Neural Network (BNN) and hardware-based acceleration for intrusion detection in in-vehicle networks. As BNN uses binary values for activations and weights rather than full precision values, it usually results in faster computation, smaller memory cost, and lower energy consumption than full precision models.