Zhang, Tongpo
(2023)
Scheduling Optimization And Coordination With Target Tracking Under Heterogeneous Networks In Automated Guided Vehicles (AGVs).
PhD thesis, University of Liverpool.
|
Text
PhD thesis_Tongpo Zhang.pdf - Author Accepted Manuscript Download (17MB) | Preview |
Abstract
Throughout the development of the multi-AGV systems, prevailing research directions contain improving the performance of individual AGV, optimizing the coordination of multiple AGVs, and enhancing the efficiency of communication among AGVs. Current researchers tend to pay attention to one research direction at a time. There is a lack of research on the overall AGV system design that tackles multiple critical design aspects of the system. This PhD research addresses four key factors of the AGV system which are AGV prototypes, target tracking algorithms, AGVs scheduling optimization and the communication of a multi-AGV system. Extensive field experiments and algorithm optimization are implemented. Comprehensive literature review is conducted to identify the research gap. The proposed solutions cover the following three aspects of the AGV system design including communication between AGVs, AGVs scheduling and computer vision in AGVs. For AGV communication, a network selection optimization algorithm is presented. An improved method for preventing convolutional neural network (CNN) immune from backdoor attack to ensure a multi-AGV system's communication security is presented. Meanwhile, a transmission framework for a multi-AGV system is presented. Those methods are used to establish a safe and efficient multi-AGV system's communication environment. For AGV scheduling, a multi-robot planning algorithm with quadtree map division for obstacles of irregular shape is presented. In addition, a scheduling optimization platform is presented. These methods are used to make a multi-AGV system have a shorter time delay and decrease the possibility of collision in a multi-robot system.Meanwhile, a scheduling optimization method based on the combination of a handover rule and the A* algorithm is proposed. The system properties that may affect the scheduling performance are also discussed. Finally, the overall performance of the newly integrated scheduling system is compared with other scheduling systems to validate its superiority and shortcomings in different corresponding work scenarios. Computer vision in AGV is investigated in detail. To improve an individual AGV's performance, an improved Camshift Algorithm has been proposed and applied to AGV prototypes. Furthermore, three deep learning models are tested under specific environments. In addition, based on the designed algorithm, the AGV prototype is able to make a convergent prediction of the pixels in the target area after the first detection of the object. Relative coordinates of the target can be located more accurately in less time. As tested in the experiments, the system architecture and new algorithm lead to reduced hardware cost, shorter time delay, improved robustness, and higher accuracy in tracking. With the three design aspects in mind, a novel method for real-time visual tracking and distance measurement is proposed. Tracking and collision avoidance functions are tested in the designed multi-AGV prototype system. Detailed design procedure, numerical analysis of the measurement data and recommendations for further improvement of the system design are presented.
| Item Type: | Thesis (PhD) |
|---|---|
| Uncontrolled Keywords: | AGV, computer vision, heterogeneous network, deep learning, path planning, collision avoidance |
| Divisions: | Faculty of Science and Engineering > School of Electrical Engineering, Electronics and Computer Science |
| Depositing User: | Symplectic Admin |
| Date Deposited: | 25 Aug 2023 09:50 |
| Last Modified: | 25 Aug 2023 09:51 |
| DOI: | 10.17638/03171347 |
| Supervisors: |
|
| URI: | https://livrepository.liverpool.ac.uk/id/eprint/3171347 |
Dimensions
Dimensions
