Jaerock Kwon

University of Michigan - Dearborn, Electrical and Computer Engineering, Faculty Member

Followers

Following

Co-authors

Public Views

David Seamon

Kansas State University

Remo Caponi

University of Cologne

Armando Marques-Guedes

UNL - New University of Lisbon

Gary Feinman

Field Museum

Roshan Chitrakar

Nepal College of Information Technology

Lev Manovich

Graduate Center of the City University of New York

Jeffrey Too Chuan Tan

Nankai University

Monish Chatterjee

University of Dayton

Shruthi Rajagopalan

SRM UNIVERSITY

Bogdan Gabrys

University of Technology Sydney

Interests

Uploads

Papers by Jaerock Kwon

Connectome, Mouse

Springer eBooks, 2022

End-to-End Learning-based Self-Driving Control Imitating Human Driving

2021 International Conference on Information and Communication Technology Convergence (ICTC), Oct 20, 2021

In recent years, End-to-End learning-based self-driving cars have been actively researched. Unlik... more In recent years, End-to-End learning-based self-driving cars have been actively researched. Unlike conventional methods, neural networks are trained to drive like human drivers by mapping directly from sensory data to control commands. In this paper, we propose a neural network architecture for recognizing visual information and controlling the steering and speed of the vehicle like humans.

JoVE Video Dataset

Journal of Visualized Experiments, Dec 9, 2011

인공 랜드마크 기반의 이동 로봇 측위 기법

A Physics-Informed Machine Learning Approach for Estimating Lithium-Ion Battery Temperature

IEEE Access, 2022

The physics-informed neural network (PINN) has drawn much attention as it can reduce training dat... more The physics-informed neural network (PINN) has drawn much attention as it can reduce training data size and eliminate the need for physics equation identification. This paper presents the implementation of a PINN with adaptive normalization in the loss function to predict lithium-ion battery cell temperature. In particular, the PINN was trained with the actual battery test data, and a lumped capacitance lithium-ion battery thermal relationship was applied to the loss function with the addition of a pre-layer and connection layer to the neural network architecture. The PINN architecture shows the most accurate battery temperature prediction compared with the fully connected neural network (FCN) and its variants evaluated in this study. The proposed PINN architecture has a mean square prediction error of 0.05 • C with a limited number of training data and without battery thermal model identification. INDEX TERMS Physics-informed neural network, lithium-ion battery, battery temperature.

Download

Image Prediction for Lane Following Assist using Convolutional Neural Network-based U-Net

Current autonomous driving systems compute steering and throttle control commands by running perc... more Current autonomous driving systems compute steering and throttle control commands by running perception-decision-action pipeline at high frequency. Although human drivers cannot react or control the vehicles as quickly as the autonomous driving softwares, most drivers control their vehicles to stay in lane unless they intend to break away from the lane. According to forward internal model theory, human can choose an optimal action for the best outcome by internally simulating all the possible consequences of various actions. This means that humans drivers choose the optimal motor commands for lane following based on their internal simulation of near-future lane changes. This paper proposes a convolutional neural network-based U-Net as a state estimator for forward internal model-based lane following assist. This state estimator can predict the lane image of near-future based on current lane image and driving status data, such as speed and steering angle. This paper also explains how time difference between current lane image and the next one to be predicted will affect the training and prediction output of the estimator.

Indoor Localization and Navigation for a Mobile Robot Equipped with Rotating Ultrasonic Sensors Using a Smartphone as the Robot's Brain

IGI Global eBooks, 2020

Identifying the current location of a robot is a prerequisite for robot navigation. To localize a... more Identifying the current location of a robot is a prerequisite for robot navigation. To localize a robot, one popular way is to use particle filters that estimate the posterior probabilistic density of a robot's state space. But this Bayesian recursion approach is computationally expensive. Most microcontrollers in a small mobile robot cannot afford it. The authors propose to use a smartphone as a robot's brain in which heavy-duty computations take place whereas an embedded microcontroller on the robot processes rudimentary sensors such as ultrasonic and touch sensors. In their design, a smartphone is wirelessly connected to a robot via Bluetooth by which distance measurements from the robot are sent to the smartphone. Then the smartphone takes responsible for computationally expensive operations like executing the particle filter algorithm. In this paper, the authors designed a mobile robot and its control architecture to demonstrate that the robot can navigate indoor environment while avoiding obstacles and localize its current position.

Download

Fully Automated Image Preprocessing for Feature Extraction from Knife-edge Scanning Microscopy Image Stacks - Towards a Fully Automated Image Processing Pipeline for Light Microscopic Images

Bioimaging, Nov 21, 2016

Knife-Edge Scanning Microscopy (KESM) stands out as a fast physical sectioning approach for imagi... more Knife-Edge Scanning Microscopy (KESM) stands out as a fast physical sectioning approach for imaging tissues at sub-micrometer resolution. To implement high-throughput and high-resolution, KESM images a tissue ribbon on the knife edge as the sample is being sectioned. This simultaneous sectioning and imaging approach has following benefits: (1) No image registration is required. (2) No manual job is required for tissue sectioning, placement or microscope imaging. However spurious pixels are present at the left and right side of the image, since the field of view of the objective is larger than the tissue width. The tissue region needs to be extracted from these images. Moreover, unwanted artifacts are introduced by KESM's imaging mechanism, namely: (1) Vertical stripes caused by unevenly worn knife edge. (2) Horizontal artifacts due to vibration of the knife while cutting plastic embedded tissue. (3) Uneven intensity within an image due to knife misalignment. (4) Uneven intensity levels across images due to the variation of cutting speed. This paper outlines an image processing pipeline for extracting features from KESM images and proposes an algorithm to extract tissue region from physical sectioning-based light microscope images like KESM data for automating feature extraction from these data sets.

Download

Remote Health Monitoring Device for the Elderly

In the USA, the oldest of the baby-boomer generation just started to reach retirement age a few y... more In the USA, the oldest of the baby-boomer generation just started to reach retirement age a few years ago. In the years to come, a much larger percentage of the population will be dependent on others for their daily care. There are not many low-cost monitoring devices available on the market to help caregivers keep dependent elderly people comfortable and healthy in a home setting without a monthly service charge. The device presented in this paper addresses this challenge to provide caregivers an emergency alert system for the elderly based on monitoring of their breathing activities and room temperature measurements. The device also allows the dependents to make on demand request for assistance. The remote communication is enabled through the cellular telephone services; so there is no special or additional subscription services needed. This is essential to make the device more affordable for the elderly.

Download

Designing a cost-effective Autonomous Vehicle Control System Kit (AVCS Kit)

2017 IEEE AFRICON, 2017

Autonomous vehicles are equipped with a variety of instrumentations, sensors and controls dependi... more Autonomous vehicles are equipped with a variety of instrumentations, sensors and controls depending upon the mission of the target vehicle. One of the biggest challenges in developing autonomous vehicles system is their cost. Often times, researchers with limited budget do not have the resources to acquire realistic autonomous vehicle test platforms for carrying out their experiments on. Most researchers become limited to conduct their research in simulation environments or use small toy cars that do not accurately represent the real system. This paper describes the design of a low-cost Autonomous Vehicle Control System Kit (AVCS Kit) using inexpensive microcontrollers, encoders, motor controllers, LIDAR, and camera. The main focus of the project was to determine how much capability could be achieved on designing the AVCS Kit to transform a normal kids'-ride-on electric car to an autonomous vehicle research platform. As presented in the paper, with the use of commercially available components we were able to achieve a good performance system at a minimal cost. In the last section, the paper demonstrates the application of the AVCS kit for an autonomous vehicle lane-following and obstacle avoidance research experiment that was developed using this platform.

Fully Automated Image Preprocessing for Feature Extraction from Knife-edge Scanning Microscopy Image Stacks - Towards a Fully Automated Image Processing Pipeline for Light Microscopic Images

Knife-Edge Scanning Microscopy (KESM) stands out as a fast physical sectioning approach for imagi... more Knife-Edge Scanning Microscopy (KESM) stands out as a fast physical sectioning approach for imaging tissues at sub-micrometer resolution. To implement high-throughput and high-resolution, KESM images a tissue ribbon on the knife edge as the sample is being sectioned. This simultaneous sectioning and imaging approach has following benefits: (1) No image registration is required. (2) No manual job is required for tissue sectioning, placement or microscope imaging. However spurious pixels are present at the left and right side of the image, since the field of view of the objective is larger than the tissue width. The tissue region needs to be extracted from these images. Moreover, unwanted artifacts are introduced by KESMâs imaging mechanism, namely: (1) Vertical stripes caused by unevenly worn knife edge. (2) Horizontal artifacts due to vibration of the knife while cutting plastic embedded tissue. (3) Uneven intensity within an image due to knife misalignment. (4) Uneven intensity l...

Download

Lateral and Longitudinal Motion Control of Autonomous Vehicles using Deep Learning

2019 IEEE International Conference on Electro Information Technology (EIT), 2019

The Current trend of the automotive industry combined with active research by the major tech comp... more The Current trend of the automotive industry combined with active research by the major tech companies has proven that self-driving vehicles are the future. The biggest challenge for self-driving cars is autonomous lateral and longitudinal control. An end-to-end model seems very promising in providing a complete software stack for autonomous driving. The work described in this paper focuses on how a deep learning technique is utilized for implementing both lateral and longitudinal control of vehicles. The open racing car simulator (TORCS) is used for developing and testing the implementation. Two separate neural networks were trained that can predict the vehicle speed and steering based on the road trajectory. Such an approach serves as a foundation towards building a system that utilizes artificial intelligence to analyze the environment and determine what the vehicle speed should be rather than following a set of predetermined rules.

OPEMI: Online Performance Evaluation Metrics Index for Deep Learning-Based Autonomous Vehicles

IEEE Access

Vision-based autonomous driving is rapidly growing. There are, however, presently no agreed-upon ... more Vision-based autonomous driving is rapidly growing. There are, however, presently no agreed-upon metrics for assessing how well deep neural network (DNN) models perform in driving. To compare novel approaches and architectures to existing ones, some researchers employed a mean error between labeled and predicted values in a test dataset and others presented a new metric that is designed to match their requirements. The discrepancy in the usage of various performance metrics and lack of objective metrics to judge the driving performance were our primary motives for developing a feasible solution. In this study, we propose online performance evaluation metrics index (OPEMI), an integrated metric that can evaluate the driving capabilities of autonomous driving models in various driving scenarios. To evaluate driving performance precisely and objectively, OPEMI incorporates several variables, including driving control stability, driving trajectory stability, journey duration, travel distance, success rate, and speed. To demonstrate the validity of OPEMI, we first confirmed that the prediction accuracy has a weak correlation with driving performance. Then, we have discussed the constraints in the existing driving performance metrics in certain circumstances, and their failure to assess the driving models. Finally, we conducted experiments with four popular DNN models and two in-house models under three different driving scenarios (generic, urban, and racing). The results show that the proposed evaluation metric, OPEMI, realistically displays driving performance and demonstrates its validity in various driving scenarios. INDEX TERMS Autonomous driving, deep learning, end-to-end learning, online/offline performance metrics. I. INTRODUCTION Recent advances in deep learning (DL) have encouraged research on vision-based longitudinal and lateral control for autonomous driving. Researchers have proposed novel learning techniques and deep learning architectures to improve the driving performance of autonomous vehicles (AVs). There exist two types of metrics to evaluate vision-based control for AVs: prediction accuracy offline metrics and driving The associate editor coordinating the review of this manuscript and approving it for publication was Vlad Diaconita. performance online metrics. The majority of comparative DNN model studies for AVs use prediction accuracy metrics (offline) [1], [2], [3], [4]. Driving performance metrics (online) have often been used to measure how far an AV drives without human intervention [5], [6], [7]. Fig. 1 illustrates the details of the two approaches. To evaluate new techniques and architectures comparatively, researchers have used a mean errors such as mean absolute error (MAE), mean squared error (MSE), and root mean squared error (RMSE), between labeled and predicted values in a test dataset [2], [3] or suggested a new metric

Download

An Integrated Approach of Threat Analysis for Autonomous Vehicles Perception System

IEEE Access

Automated vehicles are a revolutionary step in mobility, providing a safe and convenient riding e... more Automated vehicles are a revolutionary step in mobility, providing a safe and convenient riding experience while keeping the human-driving task minimal to none. Therefore, these intelligent vehicles are equipped with sophisticated perception sensors (e.g., cameras and radars), high-performance computers, artificial intelligence (AI)-driven algorithms, and connectivity with other internet-of-things (IoT) devices. This makes autonomous vehicles (AVs) a special kind of cyber-physical system (CPS) that is moving at speed in highly interactive and dynamic environments (e.g., public roads). Thus, AV is a potential target for cyber attackers to weaponize, compromising safety and mobility on the road. The first step in addressing this problem is to have a robust threat modeling framework that can address the evolving cyber-physical threats, especially to AV applications. In this regard, two areas are studied in this paper: the common practice of threat modeling in automotive and the ISO/SAE 21434 standard, and sensors and machine learning (ML) algorithms for AV perception systems and potential cyber-physical attacks. A comparative threat analysis for an AV perception system with the ISO/SAE 21434 standard and a system-theoretic process analysis for security (STPA-Sec) approach is also demonstrated in this paper. Based on the analysis, this paper proposes a robust threat analysis and risk assessment framework with mathematical modeling to identify cyber-physical threats to AV perception systems that are critical for the driving behaviors and complex interactions of AVs in their operational design domain.

Download

Short Survey of Vehicular Communication Technology

2018 IEEE International Conference on Electro/Information Technology (EIT), 2018

With the advancements in technology, the hazards of driving are expected to go down. There are a ... more With the advancements in technology, the hazards of driving are expected to go down. There are a number of safety features that are enabled by sensors and included in modern cars in order to make them safer. There are systems in place in the USA, Japan, and the European Union that are used for deployment of intelligent vehicle technology. Many scholars worldwide are actively doing research in vehicular communication (V2X) trying to make it practically applicable. The research shows the challenges and improvements that could be made in vehicular communication. The goal of this effort is to make driving conditions reliably safer in all weather conditions and at all times of the day and the year. This survey looks at the existing as well as developing technology in V2X.

Self‐sustainable wireless sensor network for low‐temperature application

Microwave and Optical Technology Letters, 2021

While environmental monitoring system is an exemplary application for wireless sensor network (WS... more While environmental monitoring system is an exemplary application for wireless sensor network (WSN), its energy‐constraint still remains unresolved issue and needs to be addressed. In this work, a self‐sustainable WSN is proposed for low‐temperature application. To resolve the energy‐constraint issue, a wireless power transfer (WPT) technique, called conformal strongly coupled magnetic resonant (CSCMR) technique, is adopted. At 90 MHz, the RF power transmission efficiency is 84% with 60 mm distance between transmitting and receiving resonators. For the 90 MHz RF power, the rectifier efficiency for the RF to DC power conversion is 69% and 76 at 25°C and −20°C, respectively. For an energy storage unit, a supercapacitor is chosen for the low‐temperature application. With the excellent temperature characteristic of the supercapacitor, the effectiveness of the supercapacitor is verified as the energy storage unit operating at −20°C. It is shown that the WSN proposed in this work can operate in perpetuity as long as the circuit components in the WSN do not have any defects.

Deep-Learning Based Fault Events Analysis in Power Systems

Energies

The identification of fault types and their locations is crucial for power system protection/oper... more The identification of fault types and their locations is crucial for power system protection/operation when a fault occurs in the lines. In general, this involves a human-in-the-loop analysis to capture the transient voltage and current signals using a common format for transient data exchange for power systems (COMTRADE) file. Then, protection engineers can identify the fault types and the line locations after the incident. This paper proposes intelligent and novel methods of faulty line and location detection based on convolutional neural networks in the power system. The three-phase fault information contained in the COMTRADE file is converted to an image file and extracted adaptively by the proposed CNN, which is trained by a large number of images under various kinds of fault conditions and factors. A 500 kV power system is simulated to generate different types of electromagnetic fault transients. The test results show that the proposed CNN-based analyzer can classify the fault...

Download

Impact of Data Sampling Methods on the Performance of Data-driven Parameter Identification for Lithium ion Batteries

IFAC-PapersOnLine, 2021

Remote Health Monitoring Device for the Elderly Development of an Affordable and Easy-to-use Breathing Monitoring Device

Download

Simulation Framework for Development and Testing of Autonomous Vehicles

2020 IEEE International IOT, Electronics and Mechatronics Conference (IEMTRONICS), 2020

The benefits of Autonomous Driving (AD) and Advanced Driver Assistance Systems (ADAS) include pro... more The benefits of Autonomous Driving (AD) and Advanced Driver Assistance Systems (ADAS) include providing critical capabilities that would help to reduce accidents due to driver errors, improve safety in traffic, and enhance driving efficiency. The goal of this paper is to utilize the Robot Operating System (ROS) and Gazebo 3D simulator for realizing a fully immersive software-in-the-loop (SIL) simulation framework, which could be utilized for the virtual design, verification, and validation of AD/ADAS features. It supports designing and implementing a software platform and methodology to validate different features for AD/ADAS with a high degree of realism. The paper will also discuss the configuration of the sensors, such as the camera, LiDAR, and RADAR, that are later used as part of a sensor fusion technique that shows the usability of the developed environment.

Connectome, Mouse

Springer eBooks, 2022

End-to-End Learning-based Self-Driving Control Imitating Human Driving

2021 International Conference on Information and Communication Technology Convergence (ICTC), Oct 20, 2021

JoVE Video Dataset

Journal of Visualized Experiments, Dec 9, 2011

인공 랜드마크 기반의 이동 로봇 측위 기법

A Physics-Informed Machine Learning Approach for Estimating Lithium-Ion Battery Temperature

IEEE Access, 2022

Download

Image Prediction for Lane Following Assist using Convolutional Neural Network-based U-Net

Indoor Localization and Navigation for a Mobile Robot Equipped with Rotating Ultrasonic Sensors Using a Smartphone as the Robot's Brain

IGI Global eBooks, 2020

Download

Fully Automated Image Preprocessing for Feature Extraction from Knife-edge Scanning Microscopy Image Stacks - Towards a Fully Automated Image Processing Pipeline for Light Microscopic Images

Bioimaging, Nov 21, 2016

Download

Remote Health Monitoring Device for the Elderly

Download

Designing a cost-effective Autonomous Vehicle Control System Kit (AVCS Kit)

2017 IEEE AFRICON, 2017

Fully Automated Image Preprocessing for Feature Extraction from Knife-edge Scanning Microscopy Image Stacks - Towards a Fully Automated Image Processing Pipeline for Light Microscopic Images

Knife-Edge Scanning Microscopy (KESM) stands out as a fast physical sectioning approach for imagi... more Knife-Edge Scanning Microscopy (KESM) stands out as a fast physical sectioning approach for imaging tissues at sub-micrometer resolution. To implement high-throughput and high-resolution, KESM images a tissue ribbon on the knife edge as the sample is being sectioned. This simultaneous sectioning and imaging approach has following benefits: (1) No image registration is required. (2) No manual job is required for tissue sectioning, placement or microscope imaging. However spurious pixels are present at the left and right side of the image, since the field of view of the objective is larger than the tissue width. The tissue region needs to be extracted from these images. Moreover, unwanted artifacts are introduced by KESMâs imaging mechanism, namely: (1) Vertical stripes caused by unevenly worn knife edge. (2) Horizontal artifacts due to vibration of the knife while cutting plastic embedded tissue. (3) Uneven intensity within an image due to knife misalignment. (4) Uneven intensity l...

Download

Lateral and Longitudinal Motion Control of Autonomous Vehicles using Deep Learning

2019 IEEE International Conference on Electro Information Technology (EIT), 2019

OPEMI: Online Performance Evaluation Metrics Index for Deep Learning-Based Autonomous Vehicles

IEEE Access

Download

An Integrated Approach of Threat Analysis for Autonomous Vehicles Perception System

IEEE Access

Download

Short Survey of Vehicular Communication Technology

2018 IEEE International Conference on Electro/Information Technology (EIT), 2018

Self‐sustainable wireless sensor network for low‐temperature application

Microwave and Optical Technology Letters, 2021

Deep-Learning Based Fault Events Analysis in Power Systems

Energies

Download

Impact of Data Sampling Methods on the Performance of Data-driven Parameter Identification for Lithium ion Batteries

IFAC-PapersOnLine, 2021

Remote Health Monitoring Device for the Elderly Development of an Affordable and Easy-to-use Breathing Monitoring Device

Download

Simulation Framework for Development and Testing of Autonomous Vehicles

2020 IEEE International IOT, Electronics and Mechatronics Conference (IEMTRONICS), 2020

An Enhanced Classification Technique for Mitigating Unexpected Noise Intrusions in Autonomous Vehicles

by Aydin Zaboli and Jaerock Kwon

2024 IEEE Transportation Electrification Conference and Expo (ITEC), 2024

Autonomous vehicle, which must trust the judgment of the entire system without human participatio... more Autonomous vehicle, which must trust the judgment of the entire system without human participation, have many disadvantages. Traffic sign classification is one of those disadvantages and a challenge to overcome. Cybersecurity breaches or changes in the surrounding environment are critical obstacles to the traffic sign classification function. Therefore, this paper proposes an image reconstruction method based on mask autoencoder to improve classification performance. Additionally, classification accuracy is also improved by extracting characters from traffic signs. A case study proves that image classification accuracy improves even when various abnormalities are added to the input of traffic sign classification performance.

Object-focused Risk Evaluation of AI-driven Perception Systems in Autonomous Vehicles

by Aydin Zaboli and Jaerock Kwon

2024 IEEE Transportation Electrification Conference and Expo (ITEC)

One of the primary motivations for autonomous vehicle (AV) technology is to reduce road accidents... more One of the primary motivations for autonomous vehicle (AV) technology is to reduce road accidents compared to human-driven cars. This necessitates having robust perception systems to detect and classify objects correctly in real-time environments. Various factors, including the complexity of the scene, the type of object, the capability of the perception sensors, and the performance of AI-based algorithms, can affect its robustness. Furthermore, vulnerabilities in these factors can be exploited as cyber-physical attacks. Hence, this paper presents a novel mathematical model for system-level risk evaluation of AV perception systems that incorporates the relevant objects for AV applications and the machine learning (ML) algorithms used to detect and classify them. This model is adapted from the ISO/SAE 21434 threat analysis and risk assessment (TARA) model with an enhancement in impact rating and attack feasibility assessment. Additionally, a case study for impact rating is demonstrated with real data from traffic crashes where the most important objects are impacted. Also, the effect of the robustness of the detection algorithm on attack feasibility assessment is illustrated with some AI/ML-based state-of-the-art detection algorithms used in AVs.

Jaerock Kwon

Related Authors

Uploads

Papers by Jaerock Kwon

Log In