PROJECTS
Embedded Files
0. Summary
0. Summary
Military Autonomous Tractor-Trailer System
Multi-agent autonomous tractor-trailer system for military snow removal operation
Project Leader, '19 - '22
Collision Avoidance of Autonomous Vehicle Using Robust Potential Function & Nonlinear MPC
Stable and robust collision avoidance algorithms based on the iLQR control and the proposed uncertainty-integrated potential function
'22 - '23
3D/4D Radar-based ADAS & Autonomous Driving Systems Integration
Integration of ADAS into autonomous driving systems
Project Leader, '22 - '24
Desnowing Algorithms Using Camera & LiDAR Fusion
Camera & LiDAR fusion for real-time desnowing
Project Leader, '22 - '23
Image Super-Resolution w/ Deep Learning Model Compression
Enhance dental X-ray image by increasing resolution and removing various noise
Project Leader, '21 - '22
1. Military Autonomous Tractor-Trailer System (Project Leader, '19 - '22)
1. Military Autonomous Tractor-Trailer System (Project Leader, '19 - '22)
Overview: Multi-agent autonomous tractor-trailer system for military snow removal operation
Platform: 1) Full-scale 6 wheels tractor-4 wheels trailer systems, 2) small-scale vehicle
Research Institute: 1) KAIST, 2) Republic of Korea Air Force, 3) Korea Institute of Machinery & Materials (KIMM), and 4) Treeze Engineering
① Tractor-Trailer Path Tracking
① Tractor-Trailer Path Tracking
6 wheels tractor-4 wheels trailer error dynamics for path tracking & trailer-stabilization
LQR control
Verification: Full-scale vehicle, Trucksim
② Leader-Follower Formation Control
② Leader-Follower Formation Control
High-speed leader-follower formation control
4 wheels vehicle dynamics & wheel dynamics
Feedback linearization-based control
Verification: Carsim
③ Waypoint Formation Control
③ Waypoint Formation Control
Follower yaw-stabilizing formation control
Lyapunov stability-based follower stabilizing reference generation
GPS/IMU-based EKF w/ bias compensation
Verification: Small-scale vehicle (ROS, GPS, IMU, AGX Xavier), Matlab
④ Architecture Design for
H/W & Communication
④ Architecture Design for
H/W & Communication
Small-scale vehicle (ROS, GPS, IMU, AGX Xavier)
2. Collision Avoidance of Autonomous Vehicle Using Robust Potential Function & Nonlinear MPC ('22 - '23)
2. Collision Avoidance of Autonomous Vehicle Using Robust Potential Function & Nonlinear MPC ('22 - '23)
Overview: Stable and robust collision avoidance algorithms based on the iLQR control and the proposed uncertainty-integrated potential function
Research Institute: KAIST
① Sensor Uncertainty Modeling
① Sensor Uncertainty Modeling
Sensor (radar) uncertainty area formulated by a multivariate normal distribution with a position measurement uncertainty
Integrate velocity uncertainty into position uncertainty using the delta method and constant velocity model
② Uncertainty-Integrated
Potential Function
② Uncertainty-Integrated
Potential Function
Propose surface-morphing function that smoothly transitions shape of isopotential contour from object shape to uncertainty area
Propose uncertainty-integrated potential function incorporating 1) surface-morphing function, 2) uncertainty area, and 3) Yukawa potential function.
③ iLQR (nonlinear MPC) for Collision Avoidance & Path TrackingUn
③ iLQR (nonlinear MPC) for Collision Avoidance & Path TrackingUn
Adopt iterative-LQR (iLQR) to address high non-linearity in potential function and vehicle dynamics
The cost function of iLQR consists of
1) potential function for collision avoidance
& 2) reference path for path trackingVerification: Carsim
3. 3D/4D Radar-based ADAS & Autonomous Driving Systems Integration (Project Leader, '22 - '24)
3. 3D/4D Radar-based ADAS & Autonomous Driving Systems Integration (Project Leader, '22 - '24)
Overview: Integration of ADAS into autonomous driving systems
*Advanced Driver Assistance Systems (ADAS)Platform: 1) IONIQ EV, 2) Daedong Golf Cart
Research Institute: 1) KAIST, 2) Future EV (FEV), and 3) Daedong
① Architecture Design for
ADAS & Autonomous Driving
① Architecture Design for
ADAS & Autonomous Driving
System integration: Reference generation for ADAS & autonomous driving
Sensor: 3D/4D radar, stereo camera, RTK-GPS, AHRS, wheel encoder
Actuator: CAN bus (steering wheel, acceleration pedal)
② Object-behvaior
Adaptive Emergency Braking
② Object-behvaior
Adaptive Emergency Braking
Propose safety distance adaptive to uncertain braking duration of the object vehicle → Improve driving comfort while performing collision avoidance
Verification: ROS-based Full-scale vehicle (IONIQ EV, golf cart), Carsim
③ Road Condition
Adaptive Emergency Braking
③ Road Condition
Adaptive Emergency Braking
Estimate maximum feasible deceleration considering 1 ~ 3 for calculating safety distance
1. Road geometry modeling (slope, curve)
2. Road friction estimation using friction-slip curve and recursive least square
3. Braking performance
Verification: Carsim
④ Road geometry adaptive ROI for emergency braking
④ Road geometry adaptive ROI for emergency braking
Minimize false alaram of emergency braking by proposed adaptive ROI
The proposed ROI is adjusted by road geometry and ego's motion
Verification: ROS-based Full-scale vehicle (IONIQ EV) w/ 3D radar
⑤ Ego motion Estimation
⑤ Ego motion Estimation
GPS/IMU-based EKF w/ bias compensation
Ego velocity estimation using 4D radar and random sampling consensus (RANSAC)
Verification: ROS-based Full-scale vehicle (IONIQ EV) w/ Oculii 4D radar
⑥ Multiple Object Tracking (MOT)
⑥ Multiple Object Tracking (MOT)
Estimate the motion states of surrounding multiple objects w/ point cloud
Clustering: DBSCAN
Prediction: IMM & EKF
Data association: PDA
Verification: ROS-based Full-scale vehicle (IONIQ EV) w/ Oculii 4D radar & Ouster LiDAR
4. Desnowing Algorithms Using Camera & LiDAR Fusion (Project Leader, '22 - '23)
4. Desnowing Algorithms Using Camera & LiDAR Fusion (Project Leader, '22 - '23)
Overview: Camera & LiDAR fusion for real-time desnowing
Research Institute: 1) KAIST and 2) KIMM
① Desnowing Algorithms
① Desnowing Algorithms
Camera Desnowing : Hampel Filter + PWC Net
LiDAR Desnowing : LIOR Filter + RIS Filter
② Camera and LiDAR Fusion System Design
② Camera and LiDAR Fusion System Design
Camera & LiDAR signal synchronization
Camera & LiDAR calibration (RANSAC & PNP) and point cloud mapping
5. Image Super-Resolution w/ Deep Learning Model Compression (Project Leader, '21 - '22)
5. Image Super-Resolution w/ Deep Learning Model Compression (Project Leader, '21 - '22)
Overview: Enhance dental X-ray image by increasing resolution and removing various noise
Research Institute: 1) KAIST and 2) Digiray
① Image super-resolution architecture design
① Image super-resolution architecture design
Preprocessing : CLAHE for image contrast enhancement
Processing : Image super-resolution by BSRGAN
② Real-time network implementation with deep learning model compression
② Real-time network implementation with deep learning model compression
Implementation of IPC communication on Windows for real-time deep learning network operation (*IPC: Inter-Process Communication)
Model compression : Apply TensorRT (quantization, graph compiler, and pruning) for inference acceleration speed
Page updated
Google Sites
Report abuse