|
AI-based Regression Analysis for Optimizing the Performance of Robot Manipulator Trajectory Tracking |
|
Chenwei Sun, Jivka Ovtcharova(FZI Research Center for Information Technology, Germany) |
| To exploit the potential of the robot manipulator in machining operations, it is significant to improve and optimize its trajectory tracking performance. In this paper, an AI-based method is proposed to reduce the robot tool center point (TCP) error under different external excitations and trajectories in the simulation environment. The robot input joints are compensated by using the fully connected neural network model to ensure that all measured joint values are as close as possible to the reference joint values. It has been tested that the TCP errors have been largely reduced in all translation directions with better vibration behaviours under all tested motion scenarios. |
|
|
Tidy-Up Tasks Using Trajectory-based Imitation Learning |
|
Doo-Jun Kim, HyunJun Jo, Jae-Bok Song(Korea University, Korea) |
| When performing reinforcement learning using a robot arm in the real environment, it is important to perform reinforcement learning safely and quickly. This is because unexpected behaviors during reinforcement learning and long-term learning can damage the robot arm or surrounding objects. In this study, trajectory-based imitation learning that suppresses unexpected situations and quickly learns the policies suitable for the robots is proposed by limiting the workspace to be explored through one human demonstration. Trajectory-based imitation learning consists
of two stages. |
|
|
A Strategy of Subsea Pipeline Identification with Sidescan Sonar based on YOLOV5 Model |
|
Yan Li, Meiyan Wu, Jiahong Guo, Yan Huang(Shenyang Institute of Automation, Chinese Academy of Sciences, China) |
| Accurate identification of pipelines is the basis and prerequisite for tracking and inspection of subsea pipelines with the help of autonomous unmanned vehicles. In this paper, we proposed a strategy based on a deep learning model YOLOV5 to extract the subsea pipeline from acoustic images acquired by a Side scan sonar (SSS). Considering the imaging mechanisms of SSS, the formed bar image by SSS in a short certain period is segmented into many sub-images. Subsequently, these sub-images are fed into a pre-trained identification model based on YOLOV5 to extract the subsea pipelines. This strategy ensures the subsea pipeline could be detected with low time consumption and satisfactory accuracy. |
|
|
Model Diet: A Simple yet Effective Model Compression for Vision Tasks |
|
Jongmin Lee, Elibol Armagan, Chong Nakyoung(JAIST, Korea) |
| Model Diet is a model compression algorithm which can be applied throughout various computer vision deep learning models. It's main features are ease of implementation and effectiveness on most of the CNN models which is popularly used recently. |
|
|
A Model-free Deep Reinforcement Learning Approach for Robotic Manipulators Path Planning |
|
Wenxing Liu, Hanlin Niu(The University of Manchester, United Kingdom), Muhammad Nasiruddin Mahyuddin(Universiti Sains Malaysia, Malaysia), Guido Herrmann, Joaquin Carrasco(The University of Manchester, United Kingdom) |
| In this research, a model-free off-policy actor critic based deep reinforcement learning method is proposed to solve the classical path planning problem of a UR5 robot arm. The proposed method not only guarantees that the joint angle of the UR5 robotic arm lies within the allowable range each time when it reaches the random target point, but also ensures that the joint angle of the UR5 robotic arm is always within the allowable range during the entire episode of training. A standard path planning method was implemented in Robot Operating System (ROS) successfully. |
|
|
Local Tetra Pattern and Its Benefits to Improve the Performance of Car and Pedestrian Detection Under Hostile Conditions |
|
Anh Linh Dang(Ho Chi Minh City University of Technology, Viet Nam), Tuyen Quang Nguyen(The University of Aizu, Japan), Tri Thien Cao(Ho Chi Minh City University of Science, Viet Nam), Vinh Quang Dinh(Vietnamese German University, Viet Nam), Vinh Dinh Nguyen(FPT University Can Tho, Viet Nam) |
| Traffic detection is a topic of great interest in recent years due to a high demand for better traffic detection systems. Existing traffic detection algorithms work well under ideal driving conditions, however their performance decreases under difficult conditions such as insufficient lighting and illumination. We propose a method that applies Local Tetra Pattern for data preprocessing, so as to improve the performance of deep learning models under said conditions. Our approach achieved better performance than the original raw-models while the changes in inference time are maintained within a negligible interval. |
|