• Journal of Advanced Navigation Technology
• /
• v.27 no.2
• /
• pp.182-189
• /
• 2023

Depth estimation is a key technology in 3D map generation for autonomous driving of vehicles, robots, and drones. The existing sensor-based method has high accuracy but is expensive and has low resolution, while the camera-based method is more affordable with higher resolution. In this study, we propose self-attention-based unsupervised monocular depth estimation for UAV camera system. Self-Attention operation is applied to the network to improve the global feature extraction performance. In addition, we reduce the weight size of the self-attention operation for a low computational amount. The estimated depth and camera pose are transformed into point cloud. The point cloud is mapped into 3D map using the occupancy grid of Octree structure. The proposed network is evaluated using synthesized images and depth sequences from the Mid-Air dataset. Our network demonstrates a 7.69% reduction in error compared to prior studies.

• The Journal of The Korea Institute of Intelligent Transport Systems
• /
• v.20 no.6
• /
• pp.313-330
• /
• 2021

In an autonomous driving system equipped with a camera and lidar, depth completion techniques enable dense depth estimation. In particular, using self-supervised learning it is possible to train the depth completion network even without ground truth. In actual autonomous driving, such depth completion should have very short latency as it is the input of other algorithms. So, rather than complicate the network structure to increase the accuracy like previous studies, this paper focuses on network latency. We design a U-Net type network with RegNet encoders optimized for GPU computation. Instead, this paper presents several techniques that can increase accuracy during the process of self-supervised learning. The proposed techniques increase the robustness to unreliable lidar inputs. Also, they improve the depth quality for edge and sky regions based on the semantic information extracted in advance. Our experiments confirm that our model is very lightweight (2.42 ms at 1280x480) but resistant to noise and has qualities close to the latest studies.

• Journal of Korean Society of Transportation
• /
• v.28 no.4
• /
• pp.107-118
• /
• 2010

Wireless communication technologies including vehicle-to-vehicle (V2V) and vehicle-to-infrastructure (V2I) enable the development of more sophisticated and effective traffic information systems. This study presents a method to estimate vehicular travel times in a vehicular ad hoc network (VANET) environment. A novel feature of the proposed method is estimating individual vehicle travel times through advanced on-board units in each vehicle, referred to as self-estimated travel time in this study. The method uses travel information including vehicle position and speed at each given time step transmitted through the V2V and V2I communications. Vehicle trajectory data obtained from the VISSIM simulator is used for evaluating the accuracy of estimated travel times. Relevant technical issues for successful field implementation are also discussed.

• Ocean Systems Engineering
• /
• v.7 no.2
• /
• pp.157-177
• /
• 2017

This research is based on the concept of safety airbag to design a self-rescue system for the autonomous underwater vehicle (AUV) using micro inertial sensing module. To reduce the possibility of losing the underwater vehicle and the difficulty of searching and rescuing, when the AUV self-rescue system (ASRS) detects that the AUV is crashing or encountering a serious collision, it can pump carbon dioxide into the airbag immediately to make the vehicle surface. ASRS consists of 10-DOF sensing module, sensing attitude algorithm and air-pumping mechanism. The attitude sensing modules are a nine-axis micro-inertial sensor and a barometer. The sensing attitude algorithm is designed to estimate failure attitude of AUV properly using sensor calibration and extended Kalman filter (SCEKF), feature extraction and backpropagation network (BPN) classify. SCEKF is proposed to be used subsequently to calibrate and fuse the data from the micro-inertial sensors. Feature extraction and BPN training algorithms for classification are used to determine the activity malfunction of AUV. When the accident of AUV occurred, the ASRS will immediately be initiated; the airbag is soon filled, and the AUV will surface due to the buoyancy. In the future, ASRS will be developed successfully to solve the problems such as the high losing rate and the high difficulty of the rescuing mission of AUV.

• Journal of the Society of Naval Architects of Korea
• /
• v.57 no.5
• /
• pp.287-296
• /
• 2020

The present study aims to improve the accuracy of the maneuvering simulations based on captive model test results. To derive the hydrodynamic coefficients in a self-propelled condition, a mathematical maneuvering model using a whole vehicle model was established. Captive model tests were carried out using the Vertical Planar Motion Mechanism (VPMM) equipment. A motor controller was used to control the constant propeller revolution rate during pure motion tests. The resistance tests, self-propulsion tests, static drift tests, and VPMM tests were performed in the towing tank of Seoul National University. When the vertical drift angle changes, the gravity load on the sensors were changed. The hydrodynamic forces were deduced by subtracting the gravity load from the measured forces. The hydrodynamic coefficients were calculated using the least-square method. The simulation of the turning circle test was compared with the free-running model test result, and the error of the turning radius was 8.3 % compared to the free-running model test.

• The Transactions of The Korean Institute of Electrical Engineers
• /
• v.66 no.9
• /
• pp.1416-1423
• /
• 2017

Lane detection technology is one of the most important issues on car safety and self-driving capability of autonomous vehicle. This paper introduces an accurate lane detection scheme based on OCC(Optical Camera Communication) for moving vehicles. For lane detection of moving vehicles, the streetlights and the front camera of the vehicle were used for a transmitter and a receiver, respectively. Based on the angle information of multiple streetlights in a captured image, the distance from sidewalk can be calculated using non-linear regression analysis. Simulation results show that the proposed scheme shows robust performance of accurate lane detection.

• Proceedings of the Korean Society for Noise and Vibration Engineering Conference
• /
• 2007.05a
• /
• pp.263-263
• /
• 2007

In the future, self-contained sensors and processing units will need on-board, renewable power supplies to be truly autonomous. One way of supplying such power is through energy harvesting, processes by which ambient forms of energy are converted into electricity. One energy harvesting technique involves converting kinetic energy, in the form of vibrations, into electrical energy through the use of piezoelectric materials. Researchers are currently investigating how piezoelectric materials can be used to harvest power. This study examines the use of auxiliary structures, consisting of a mechanical fixture and a lead zirconate/lead titanate (PZT) piezoelectric element, which can be attached to any boundary conditions vibrating beam of the any boundary conditions. Adjusting various boundary conditions of these structures can maximize the strain induced in the attached PZT element and improve power output.

• The Journal of the Korea institute of electronic communication sciences
• /
• v.12 no.3
• /
• pp.493-498
• /
• 2017

Recently, traffic data is analyzed for efficiency of bus operation, D2D(: Door to Door) service, and self-driving of public transportation. However, various studies have been carried out to predict the delay time of public transportation, especially buses, but the research to date has been insufficient due to limitations of simple analysis and data acquisition. In this study, delay time estimation is performed by collecting and processing data such as day of the week, weather, and time of day based on bus operation information. The proposed method in this paper can be applied to autonomous public transport and public traffic control system by improving the accuracy by adding variables in the future.

• International Journal of Fuzzy Logic and Intelligent Systems
• /
• v.6 no.2
• /
• pp.161-166
• /
• 2006

Localization of mobile agent within a sensing network is a fundamental requirement for many applications, using networked navigating systems such as the sonar-sensing system or the visual-sensing system. To fully utilize the strengths of both the sonar and visual sensing systems, This paper describes a networked sensor-based navigation method in an indoor environment for an autonomous mobile robot which can navigate and avoid obstacle. In this method, the self-localization of the robot is done with a model-based vision system using networked sensors, and nonstop navigation is realized by a Kalman filter-based STSF(Space and Time Sensor Fusion) method. Stationary obstacles and moving obstacles are avoided with networked sensor data such as CCD camera and sonar ring. We will report on experiments in a hallway using the Pioneer-DX robot. In addition to that, the localization has inevitable uncertainties in the features and in the robot position estimation. Kalman filter scheme is used for the estimation of the mobile robot localization. And Extensive experiments with a robot and a sensor network confirm the validity of the approach.

• The Journal of The Korea Institute of Intelligent Transport Systems
• /
• v.21 no.3
• /
• pp.83-95
• /
• 2022

This paper proposes a system that performs object detection and distance estimation for application to autonomous vehicles. Object detection is performed by a network that adjusts the split grid to the input image ratio using the characteristics of the recently actively used deep learning model YOLOv4, and is trained to a custom dataset. The distance to the detected object is estimated using a bounding box and homography. As a result of the experiment, the proposed method improved in overall detection performance and processing speed close to real-time. Compared to the existing YOLOv4, the total mAP of the proposed method increased by 4.03%. The accuracy of object recognition such as pedestrians, vehicles, construction sites, and PE drums, which frequently occur when driving to the city center, has been improved. The processing speed is approximately 55 FPS. The average of the distance estimation error was 5.25m in the X coordinate and 0.97m in the Y coordinate.