Hybrid Balance Artificial Potential Field Navigation System For An Autonomous Surface Vessel In Riverine Environment

• Main Author: Jianhong, Mei
• Format: Thesis
• Language: English
• Published: 2019
• Subjects: T Technology; TK1-9971 Electrical engineering. Electronics. Nuclear engineering
• Online Access: http://eprints.usm.my/48209/1/Hybrid%20Balance%20Artificial%20Potential%20Field%20Navigation%20System%20For%20An%20Autonomous%20Surface%20Vessel%20In%20Riverine%20Environment.pdf; http://eprints.usm.my/48209/

The demands of Autonomous Surface Vessels (ASVs) for applications such as river bathymetry survey and environmental monitoring are increasing rapidly. However, it is still relatively challenging for the ASVs platform to navigate autonomously due to factors such as unknown and unstructured waterway, and the presence of static and dynamic obstacles. The ASV platform needs some level of autonomy and intelligence in order to make reasonable decisions and risk analysis for safe autonomous navigation. There are two issues related to ASV autonomous riverine navigation; river environment modelling and autonomous path planning and obstacles avoidance. Thus, the objectives of the research are: to develop a riverbanks identification algorithm for ASV navigation; and to develop a marine traffic rules compliant navigation and obstacles avoidance algorithm for ASV in the unstructured riverine environment. The riverbanks are selected as the visual cues for the river tracking. The issues of recognising the riverbanks include factors such as color variation with the light condition, water reflection and the complex scene of plants on the riverbanks. In order to overcome these issues, a Color Segmentation Constrained Hough Transform Algorithm is proposed. The results show that the proposed method identified all the riverbanks successfully. To evaluate the performance of the proposed method, the average and variance error deviation are calculated. The Euclidean distances of detected lines from ground truth are used to compare the accuracy of the proposed method. The average error deviation of the proposed method, color segmentation method, Hough Transform method are 3.145 pixel, 16.736 pixel and 27.507 pixel, respectively. The variance error deviation of the three methods are 0.099, 5.467 and 19.749, respectively. For the river tracking problem, a balance control scheme is proposed in order to achieve simultaneous river tracking and obstacles avoidance. The proposed Hybrid Balance-Artificial Potential Field (APF) method is a method that does not utilize the GPS information for the river navigation which means that it is suitable for the case without known river map. Static and dynamic obstacles in the river are used to verify the proposed balance-APF method. The simulation results show that the Hybrid Balance-APF method successfully achieved simultaneous river tracking and obstacles avoidance. In addition, convention on the International Regulations for Preventing Collisions at Sea (COLREGs) is integrated into the ASV navigation system, which makes the ASV able to abide by the standard marine traffic rules. From the adaptation with COLREGs requirements, the ASV platform can navigate safely from typical riverine encounter such as static and dynamic obstacles avoidance, head-on and overtaking encounter. In summary, feasible autonomous riverine environment navigation system for ASV has been successfully developed.