Dae-Kwan Ko
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Research Subjects
Research Interests
Dae-Kwan Ko has a total of 9 co-authors. Here are the top co-authors by number of shared publications:
- Soo-Chul Lim
Department of Mechanical, Robotics and Energy Engineering, Dongguk University, Seoul, South Korea
6 shared publications - Kang-Won Lee
Department of Mechanical, Robotics and Energy Engineering, Dongguk University, Seoul, South Korea
3 shared publications Dong Han Lee
1 shared publication- Dong-Han Lee
Mechanical, Robotics, and Energy Engineering, Dongguk University, 34942 Seoul Korea (the Republic of) 04620
1 shared publication - Jee-Hwan Ryu
Department of Civil and Environmental Engineering, KAIST, Daejeon, South Korea
1 shared publication - Kang-Il Yoon
Department of Mechanical, Robotics, and Energy Engineering, Dongguk University, Seoul, Korea
1 shared publication
Latest Publications
- 26 March 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Robotics and Automation Letters
- Vol. 11 (5) , 6106-6113
ThermoAct: Thermal-Aware Vision-Language-Action Models for Robotic Perception and Decision-Making
In recent human-robot collaboration environments, there is a growing focus on integrating diverse sensor data beyond visual information to enable safer and more intelligent task execution. Although thermal data can be crucial for enhancing robot safety and operational efficiency, its integration has been relatively overlooked in prior research. This letter proposes a novel Vision-Language-Action (VLA) framework that incorporates thermal information for robot task execution. The proposed system leverages a Vision-Language Model (VLM) as a high-level planner to interpret complex natural language commands and decompose them into simpler sub-tasks. This approach facilitates efficient data collection and robust reasoning for complex operations. Unlike conventional methods that rely solely on visual data, our approach integrates thermal information, enabling the robot to perceive physical properties and proactively ensure environmental safety. Experimental results from real-world task scenarios validate the feasibility of our proposed framework, suggesting its potential to enhance task success rates and safety compared to existing vision-based systems.
- 26 March 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Robotics and Automation Letters
- Vol. 11 (5) , 6106-6113
ThermoAct: Thermal-Aware Vision-Language-Action Models for Robotic Perception and Decision-Making
In recent human-robot collaboration environments, there is a growing focus on integrating diverse sensor data beyond visual information to enable safer and more intelligent task execution. Although thermal data can be crucial for enhancing robot safety and operational efficiency, its integration has been relatively overlooked in prior research. This letter proposes a novel Vision-Language-Action (VLA) framework that incorporates thermal information for robot task execution. The proposed system leverages a Vision-Language Model (VLM) as a high-level planner to interpret complex natural language commands and decompose them into simpler sub-tasks. This approach facilitates efficient data collection and robust reasoning for complex operations. Unlike conventional methods that rely solely on visual data, our approach integrates thermal information, enabling the robot to perceive physical properties and proactively ensure environmental safety. Experimental results from real-world task scenarios validate the feasibility of our proposed framework, suggesting its potential to enhance task success rates and safety compared to existing vision-based systems.
- 6 August 2024
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Intelligent Transportation Systems
- Vol. 25 (11) , 16676-16686
Latency-Free Driving Scene Prediction for On-Road Teledriving With Future-Image-Generation
Teledriving could serve as a practical solution for handling unforeseen situations in autonomous driving. However, the latency of transmission networks remains a prominent concern. Despite advancements like 5G networks, delays in remote driving scenes cannot be entirely eradicated, potentially leading to unwanted incidents. While a few attempts have been made to address this issue by predicting the future driving scenes, these efforts have been restricted in their ability to accurately foresee clear and relevant driving scenarios. This study presents a method to predict a latency-free future driving scene. Unlike prior approaches, our method incorporates the command signal of a remote driver into the prediction network, as well as the past driving video frames and vehicle status. As a result, we can accurately predict relevant and clear latency-free future driving scenes. A combination of convolutional long short-term memory (ConvLSTM) and generative adversarial networks (GAN) was utilized in a deep neural network to predict the future driving scenes based on latency. The dataset was gathered from on-road teledriving experiments, with a maximum vehicle speed of 53 km/h and a driving route length of approximately 1.3 km. The dataset used to train the deep neural network was gathered from on-road teledriving experiments. The proposed method can estimate the future driving scene for up to 0.5 s, surpassing the performance of both baseline video prediction methods and a method that does not utilize the input command of the driver.
- 6 August 2024
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Intelligent Transportation Systems
- Vol. 25 (11) , 16676-16686
Latency-Free Driving Scene Prediction for On-Road Teledriving With Future-Image-Generation
Teledriving could serve as a practical solution for handling unforeseen situations in autonomous driving. However, the latency of transmission networks remains a prominent concern. Despite advancements like 5G networks, delays in remote driving scenes cannot be entirely eradicated, potentially leading to unwanted incidents. While a few attempts have been made to address this issue by predicting the future driving scenes, these efforts have been restricted in their ability to accurately foresee clear and relevant driving scenarios. This study presents a method to predict a latency-free future driving scene. Unlike prior approaches, our method incorporates the command signal of a remote driver into the prediction network, as well as the past driving video frames and vehicle status. As a result, we can accurately predict relevant and clear latency-free future driving scenes. A combination of convolutional long short-term memory (ConvLSTM) and generative adversarial networks (GAN) was utilized in a deep neural network to predict the future driving scenes based on latency. The dataset was gathered from on-road teledriving experiments, with a maximum vehicle speed of 53 km/h and a driving route length of approximately 1.3 km. The dataset used to train the deep neural network was gathered from on-road teledriving experiments. The proposed method can estimate the future driving scene for up to 0.5 s, surpassing the performance of both baseline video prediction methods and a method that does not utilize the input command of the driver.
- 11 July 2023
- Published by Springer Nature in International Journal of Control, Automation, and Systems
- Vol. 21 (7) , 2387-2397
Real-time Video Prediction Using GANs With Guidance Information for Time-delayed Robot Teleoperation
- 11 July 2023
- Published by Springer Nature in International Journal of Control, Automation, and Systems
- Vol. 21 (7) , 2387-2397
Real-time Video Prediction Using GANs With Guidance Information for Time-delayed Robot Teleoperation
- 9 August 2022
- Published by Elsevier in Expert Systems with Applications
- Vol. 211
Vision-based interaction force estimation for robot grip motion without tactile/force sensor
- 9 August 2022
- Published by Elsevier in Expert Systems with Applications
- Vol. 211
Vision-based interaction force estimation for robot grip motion without tactile/force sensor
- 7 July 2021
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Robotics and Automation Letters
- Vol. 6 (4) , 6640-6646
Toward Vision-Based High Sampling Interaction Force Estimation With Master Position and Orientation for Teleoperation
In this study, a vision-based high sampling rate interaction force estimation method is proposed for teleoperation systems that uses master position and orientation information without using physical sensors such as force/torque (F/T) or tactile sensors. The proposed method uses red-green-blue (RGB) images, six-axis robot pose and motor current data, gripper position and current data, as well as master position and orientation information as inputs without requiring force sensors. To estimate the interaction forces, a deep neural network composed of densely connected convolutional network (DenseNet) and long short-term memory (LSTM) is proposed. The database was created by operators using grip and picking motions to interact with 10 objects over a teleoperation system. In addition, we compared the proposed method with different deep learning networks that used different sets of inputs. The results show that the proposed model can estimate 1 kHz interaction force based on 60 Hz images and 1 kHz master inputs. Moreover, the results indicate that the master position and orientation information are useful in estimating the interaction force at a high sampling rate through the result of the change in the network input.
- 7 July 2021
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Robotics and Automation Letters
- Vol. 6 (4) , 6640-6646
Toward Vision-Based High Sampling Interaction Force Estimation With Master Position and Orientation for Teleoperation
In this study, a vision-based high sampling rate interaction force estimation method is proposed for teleoperation systems that uses master position and orientation information without using physical sensors such as force/torque (F/T) or tactile sensors. The proposed method uses red-green-blue (RGB) images, six-axis robot pose and motor current data, gripper position and current data, as well as master position and orientation information as inputs without requiring force sensors. To estimate the interaction forces, a deep neural network composed of densely connected convolutional network (DenseNet) and long short-term memory (LSTM) is proposed. The database was created by operators using grip and picking motions to interact with 10 objects over a teleoperation system. In addition, we compared the proposed method with different deep learning networks that used different sets of inputs. The results show that the proposed model can estimate 1 kHz interaction force based on 60 Hz images and 1 kHz master inputs. Moreover, the results indicate that the master position and orientation information are useful in estimating the interaction force at a high sampling rate through the result of the change in the network input.
- 1 March 2021
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Industrial Informatics
- Vol. 17 (3) , 1978-1986
Continuous Image Generation From Low-Update-Rate Images and Physical Sensors Through a Conditional GAN for Robot Teleoperation
When a robot is teleoperated, its operator control is based on transmitted images. Network limitations and/or a remote distance usually cause delays or interruptions of the image transmission, which is one of the reasons for the instability of teleoperation systems. In this work, we propose a high-update-rate image generation method using past low update image and current grip position and electrical motor current of gripper received by sensors during teleoperation via a conditional generative adversarial network. The main challenge is that such a network can generate current high-update-rate images from past low-update-rate one, the current high-update-rate grip force, and the grip angle. We equipped a robot gripper with a camera and a grip force sensor and collected a large dataset of robot vision, grip force, and grip angle sequences; objects with deformation, including irregular deformation, and rigid objects were tested in the experiment to verify the possibility of high-update-rate image generation under various grip conditions. We found that the proposed network allows the generation of current images with high update rate. Here, we present the results of real-time image generation from delayed images and sensor data along with the root-mean-square value, the peak signal-to-noise ratio, and the structural similarity index measure and compare the images generated by the proposed network with those from other networks.
- 1 March 2021
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Industrial Informatics
- Vol. 17 (3) , 1978-1986
Continuous Image Generation From Low-Update-Rate Images and Physical Sensors Through a Conditional GAN for Robot Teleoperation
When a robot is teleoperated, its operator control is based on transmitted images. Network limitations and/or a remote distance usually cause delays or interruptions of the image transmission, which is one of the reasons for the instability of teleoperation systems. In this work, we propose a high-update-rate image generation method using past low update image and current grip position and electrical motor current of gripper received by sensors during teleoperation via a conditional generative adversarial network. The main challenge is that such a network can generate current high-update-rate images from past low-update-rate one, the current high-update-rate grip force, and the grip angle. We equipped a robot gripper with a camera and a grip force sensor and collected a large dataset of robot vision, grip force, and grip angle sequences; objects with deformation, including irregular deformation, and rigid objects were tested in the experiment to verify the possibility of high-update-rate image generation under various grip conditions. We found that the proposed network allows the generation of current images with high update rate. Here, we present the results of real-time image generation from delayed images and sensor data along with the root-mean-square value, the peak signal-to-noise ratio, and the structural similarity index measure and compare the images generated by the proposed network with those from other networks.