Author: Zheng Zeng
Zeng, Zheng, 2015 Evolutionary Path Planner with Shell Space Decomposition for Autonomous Underwater Vehicles in Ocean Environments, Flinders University, School of Computer Science, Engineering and Mathematics
This electronic version is made publicly available by Flinders University in accordance with its open access policy for student theses. Copyright in this thesis remains with the author. This thesis may incorporate third party material which has been used by the author pursuant to Fair Dealing exceptions. If you are the owner of any included third party copyright material and/or you believe that any material has been made available without permission of the copyright owner please contact email@example.com with the details.
Autonomous underwater vehicles (AUVs) have been proposed for a large spectrum of applications, ranging from environmental monitoring, ocean floor mapping, search and rescue operations, tracking of multiple targets, surveillance and reconnaissance, etc. Many of these applications consume much time lasting from days to weeks, and cover large areas of hundreds to thousands of square kilometres. Oceanographic processes, such as currents show great variability over such large expanses and will evolve over such durations. In most existing applications, AUVs are typically deployed from surface vessels with support personnel for deployment, piloting and recovery. During the course of the mission, the support vessel will shadow the AUV and provide much of the higher level decision making processes needed to deal with changes of oceanographic processes. The cost of keeping the support vessel on standby is generally by far the most significant component of the mission cost. Recently, there has been growing interest in developing long range AUVs with increased autonomy to conduct science missions over longer periods without supervision, thereby reducing mission costs and extending their applicability. Many recent efforts towards improving AUV persistence focus on improvements in battery technologies, on-board power management, reduced hotel load, drag, efficient propulsion, navigation and guidance systems, and overall system reliability. In particular, an effective and versatile path planning system is of crucial importance to the safe, successful and efficient completion of long range missions. Rather than depend on the support of a manned surface vehicle, an AUV could be launched from shore where upon a path planning system could be used to generate a trajectory that exploits the ocean energy taking use of the favourable currents to propel the vehicle and lead the vehicle to a remote work site, perform a survey, and then return to shore completely on its own. This will greatly diminish the costs for AUV operations since there is no need for a manned support vessel and vehicle operations can be monitored on-shore, thereby enhancing the affordability of AUVs to science and industry. In order to facilitate extended range, the AUV may be used in combination with an autonomous surface vehicle (ASV). The ASV will provides the AUV with localization support, and the AUV will periodically rendezvous and dock with the ASV to recharge and upload data and download instructions. This thesis proposes several path planning and re-planning techniques for applications involving either a single vehicle or teams of vehicles operating in a dynamic, cluttered, and uncertain ocean environment. This thesis presents a methodology for formulating the AUV path planning problem in the context of the environmental constraints. The turbulent, cluttered and uncertain environments modelled here incorporate strong currents field, irregularly shaped terrains and obstacles, the position of which may be dynamic and uncertain. The B-Spline based quantum-behaved particle swarm optimization (QPSO) path planning technique, introduced in this thesis, combines the main advantages of previously published approaches. These include smooth curvature paths represented by the Spline to accommodate constraints imposed by the manoeuvrability of the vehicle; and a QPSO algorithm enables the path planner to obtain a more optimized trajectory than the A*, rapidly exploring random tree (RRT), the conventional genetic algorithm (GA) and particle swarm optimization (PSO) derived trajectories. A new shell space decomposition (SSD) scheme is then proposed to increase the searching efficiency of the B-Spline based QPSO path planner. This scheme decomposes the search space into shell regions radiating out from the starting point to the destination, and one or more control points generating the Splines are placed within each of these regions. This arrangement gives more freedom to the placement of the control points, but still restricts the search space for each control point to its respective regions to save computation time. The SSD scheme has been integrated with a QPSO based path planner and tested to find an optimal trajectory for an AUV navigating through a variable ocean environment in the presence of obstacles. Subsequently, the generic SSD scheme is extended to account for the case of AUV operating in a spatiotemporal ocean environment. A dynamic SSD strategy is developed and incorporated with the on-line planning system that adapts and regenerates the trajectory during the course of the mission using continuously updated current profiles from on-board sensors. The next part of this thesis introduces path planning problems involving multiple autonomous marine vehicles (AMVs). The focus of the work is the problem of organising simultaneous arrival for multiple AMVs in the presence of variable ocean currents, irregularly shaped terrains and dynamic obstacles. A distributed shell space decomposition (DSSD) scheme that directly derives from the SSD concept is developed. The proposed QPSO-DSSD path planner is integrated with a novel optimized mass-centre rendezvous point selection scheme to identify the optimal rendezvous position, along with an optimal operational speed scheme to improve the performance of simultaneousness arrival at the rendezvous point, and reduce power consumption to complete the mission. This research accommodates the current and future needs of persistent presence of AUVs. The new autonomous planning techniques developed in this thesis contribute to improve the capability of the AUVs to have longer mission durations as well as higher levels of autonomy.
Keywords: Autonomous marine vehicles; Evolutionary algorithm; Optimization; Particle swarm optimization; Path planning; Space decomposition; Dynamic path re-planning; Spatiotemporal current map
Subject: Engineering thesis
Thesis type: Doctor of Philosophy
School: School of Computer Science, Engineering and Mathematics
Supervisor: Karl Sammut