Anfeng Jiang
Research Subjects
Anfeng Jiang has a total of 14 co-authors. Here are the top co-authors by number of shared publications:
- Aiqiang Pan
State Grid Shanghai Municipal Electric Power Company, China
3 shared publications - Yingying Zhao
Electric Power Science Research Institute, State Grid Shanghai Municipal Electric Power Company, Shanghai, China
2 shared publications - Yubo Fan
State Grid Shanghai Municipal Electric Power Company, China
2 shared publications - Yue Tian
State Grid Shanghai Municipal Electric Power Company, China
2 shared publications - Zhe Liu
State Grid Shanghai Municipal Electric Power Company, China
2 shared publications - Bengang Wei
State Grid Shanghai Electric Power Research Institute, Shanghai 200437, China
1 shared publication
Latest Publications
- Open Access
- 24 May 2024
- Published by Association for Computing Machinery (ACM)
- p. 376-380
Trajectory Acceleration Based Reinforcement Learning Motion Planning For the Robotic Arm of Live Working Robots
The motion planning of the robotic arm is the core element for efficient and reliable completion of live working tasks by live working robots. The deep learning algorithms used in the past for motion planning of the robotic arm of live working robots can achieve motion planning in high-dimensional space, but these algorithms use neural network motion planners with slow training speed, low success rate of motion planning, and high convergence rounds. In response to this issue, this paper proposes a trajectory acceleration-based reinforcement learning motion planning method for the robotic arm of live working robots. A motion planning model for the robotic arm of live working robots is established through software, and the relevant parameters of the reinforcement learning motion planning method are trained. The success rate and convergence rate of motion planning are calculated under different trajectory guidance and reward function designs. The calculation results show that the motion planning method proposed in this paper can significantly improve the planning success rate and convergence speed under trajectory guidance and dense reward functions.
- 24 May 2024
- Published by Association for Computing Machinery (ACM)
- p. 376-380
Open AccessTrajectory Acceleration Based Reinforcement Learning Motion Planning For the Robotic Arm of Live Working Robots
The motion planning of the robotic arm is the core element for efficient and reliable completion of live working tasks by live working robots. The deep learning algorithms used in the past for motion planning of the robotic arm of live working robots can achieve motion planning in high-dimensional space, but these algorithms use neural network motion planners with slow training speed, low success rate of motion planning, and high convergence rounds. In response to this issue, this paper proposes a trajectory acceleration-based reinforcement learning motion planning method for the robotic arm of live working robots. A motion planning model for the robotic arm of live working robots is established through software, and the relevant parameters of the reinforcement learning motion planning method are trained. The success rate and convergence rate of motion planning are calculated under different trajectory guidance and reward function designs. The calculation results show that the motion planning method proposed in this paper can significantly improve the planning success rate and convergence speed under trajectory guidance and dense reward functions.
- Open Access
- 24 November 2023
- Published by Association for Computing Machinery (ACM)
- p. 334-338
Research on Visual Localization Method for Substation Live Working Robot Based on Deep Learning
The use of visual systems to judge the surrounding environment and locate work targets is a key link for efficient and reliable completion of live working tasks by substation live working robots. The traditional feature object detection algorithms used by live working robots in the past cannot meet the recognition and positioning accuracy requirements of live working in complex working environments of substations. Therefore, this article proposes a visual positioning method for substation live working robots based on deep learning models. It uses 10000 annotated complex substation scene photos containing various types of equipment as the training set, trains the visual positioning method proposed in this article, tests 2000 photos of complex substation scenes containing various equipment, and analyzes the recognition and positioning results.
- 24 November 2023
- Published by Association for Computing Machinery (ACM)
- p. 334-338
Open AccessResearch on Visual Localization Method for Substation Live Working Robot Based on Deep Learning
The use of visual systems to judge the surrounding environment and locate work targets is a key link for efficient and reliable completion of live working tasks by substation live working robots. The traditional feature object detection algorithms used by live working robots in the past cannot meet the recognition and positioning accuracy requirements of live working in complex working environments of substations. Therefore, this article proposes a visual positioning method for substation live working robots based on deep learning models. It uses 10000 annotated complex substation scene photos containing various types of equipment as the training set, trains the visual positioning method proposed in this article, tests 2000 photos of complex substation scenes containing various equipment, and analyzes the recognition and positioning results.
- 25 December 2020
- Published by Institute of Electrical and Electronics Engineers (IEEE)
- p. 321-325
Intelligent Control Method of Live Working Robot Based on Cloud and Edge Computing Terminal
The existing single machine working mode of live working robots cannot meet the computational power requirements of artificial intelligence high-precision algorithm, and it is difficult to adapt to the complex working scene, and cannot be applied to the scene where a large number of robots work at the same time to generate massive data management and control. In view of the existing shortcomings, this paper analyzes and compares the working modes of existing power grid inspection robots and live working robots, and proposes an intelligent management and control method for live working robots based on cloud and edge computing terminals. By deploying different parts of deep learning algorithm in cloud and edge computing terminal, the problems existing in the previous working mode are solved. A typical example of live working robot scene in substation is analyzed. The results show that the intelligent control method proposed in this paper can meet the computational power requirements of artificial intelligence deep learning, neural network coordinated control, recognition and control model self-learning, self-renewal and other high-precision algorithm, and can be applied to the scene of complex environment or a large number of robots working at the same time to produce massive data control.
- 25 December 2020
- Published by Institute of Electrical and Electronics Engineers (IEEE)
- p. 321-325
Intelligent Control Method of Live Working Robot Based on Cloud and Edge Computing Terminal
The existing single machine working mode of live working robots cannot meet the computational power requirements of artificial intelligence high-precision algorithm, and it is difficult to adapt to the complex working scene, and cannot be applied to the scene where a large number of robots work at the same time to generate massive data management and control. In view of the existing shortcomings, this paper analyzes and compares the working modes of existing power grid inspection robots and live working robots, and proposes an intelligent management and control method for live working robots based on cloud and edge computing terminals. By deploying different parts of deep learning algorithm in cloud and edge computing terminal, the problems existing in the previous working mode are solved. A typical example of live working robot scene in substation is analyzed. The results show that the intelligent control method proposed in this paper can meet the computational power requirements of artificial intelligence deep learning, neural network coordinated control, recognition and control model self-learning, self-renewal and other high-precision algorithm, and can be applied to the scene of complex environment or a large number of robots working at the same time to produce massive data control.
- 1 October 2014
- Published by Institute of Electrical and Electronics Engineers (IEEE)
- p. 681-687
Experimental and analytical studies of the effects of multiple lightning impulse currents on metal-oxide arrester blocks
The multiple lightning strokes are the main cause of distribution arresters failures. In order to evaluate performance of distribution lines arresters more realistically, the experimental and analytical studies of the effects of multiple lightning impulse currents on metal-oxide arrester blocks are carried out. The multiple impulse current generator developed by Shanghai Jiao Tong University is adopted, the generator could output four impulse current representing one first stroke current, two subsequent stroke current and one continuing current of interval stroke, respectively. The variable interval time between the two impulses could achieve as low as 1 ms. The effects of block dimension, interval time between impulse currents, the sequences of impulse currents and the amplitude of impulse currents are studied. The results show larger amplitude of multiple impulse current, less interval time between impulse currents and smaller geometric parameter of arrester block are the causes of blocks failure. Surface flashover and energy absorption are two failure characteristics under multiple impulse current. The DC 1mA reference voltage is more sensitive than the residual voltage of blocks under multiple impulse current. The multiple impulse currents tests are suitable for performance evaluation of distribution lines arresters under natural lightning stroke.
- 1 October 2014
- Published by Institute of Electrical and Electronics Engineers (IEEE)
- p. 681-687
Experimental and analytical studies of the effects of multiple lightning impulse currents on metal-oxide arrester blocks
The multiple lightning strokes are the main cause of distribution arresters failures. In order to evaluate performance of distribution lines arresters more realistically, the experimental and analytical studies of the effects of multiple lightning impulse currents on metal-oxide arrester blocks are carried out. The multiple impulse current generator developed by Shanghai Jiao Tong University is adopted, the generator could output four impulse current representing one first stroke current, two subsequent stroke current and one continuing current of interval stroke, respectively. The variable interval time between the two impulses could achieve as low as 1 ms. The effects of block dimension, interval time between impulse currents, the sequences of impulse currents and the amplitude of impulse currents are studied. The results show larger amplitude of multiple impulse current, less interval time between impulse currents and smaller geometric parameter of arrester block are the causes of blocks failure. Surface flashover and energy absorption are two failure characteristics under multiple impulse current. The DC 1mA reference voltage is more sensitive than the residual voltage of blocks under multiple impulse current. The multiple impulse currents tests are suitable for performance evaluation of distribution lines arresters under natural lightning stroke.