Practical implementation of a path replanning system for an AUV operating in dynamic environments

Autonomous Underwater Vehicles (AUVs) are increasingly used for a wide variety of missions with extended durations. Due to the limited onboard battery capacities of typical AUVs, the energy shortage of AUVs is a relevant issue that requires further research to solve. Efficient motion and path planning is one of the key factors for completing long duration missions. Traditional pre-generative planners are inadequate to adapt to unknown and dynamically varying ocean environments. This is particularly evident for long-duration missions, in which a vehicle may unexpectedly encounter adverse ocean currents and suffer energy shortages. As the technological development in recent years continues to improve the onboard computational power of AUVs, it is timely to explore the potential of a sophisticated path planner for improving the operability, efficiency and endurability of AUVs. This thesis focuses on the development of a practical online path planner to improve the performance of an AUV operating in dynamic environments. The particle swarm optimization (PSO) algorithm and its variants are suitable for the application of online path planning in dynamic environments because they can maintain a large pool of solutions that can be used for replanning a vehicle path at any time throughout the mission. Nonetheless, there are concerns about the practicability of PSO based path planners such as the convergence of particles at local minima due to limited time for online planning, as well as their computational loads when implemented in an actual vehicle. Therefore, a preliminary review was first conducted to identify the strengths and weaknesses of existing PSO-based algorithms for solving the AUV path planning problem. A pre-generative AUV path planner was developed to solve the offline path planning problem of an AUV operating in a turbulent and cluttered ocean environment. The path planner successfully generated safe and feasible time-optimal paths that can exploit ocean currents to improve the AUV’s performance. Based on the preliminary review, a new approach was proposed to improve the performance of a PSO-based path planner by using selective hybridization of differential evolution (DE). The novel algorithms can conduct the DE operation selectively on particles to enhance the particles’ searching ability and resistance to local minima without inflating the computational cost. According to the results of Monte Carlo simulations and Kruskal-Wallis tests, the proposed algorithms outperformed other algorithms in the tested scenarios. To improve the path planner’s search efficiency, which is the most critical attribute for an online path planner, constrained optimization was applied to formulate the path planner. The search domain was modelled using the polar coordinate system and a combination of hard and soft constraints. This ensured the compliance of paths with the vehicular constraints and facilitated the placement of path nodes to improve search efficiency. Different types of constraints were analysed to identify the optimal constraint setting that produced the highest search efficiency. Using a novel PSO-based algorithm, an online AUV path planner that employed a path replanning scheme was proposed to address the path planning problem of an AUV operating in a dynamic and unexplored ocean environment. The proposed path replanner can continuously refine a safe and feasible time-optimal path for an unknown environment based on the feedback from its onboard sensors. The performance and robustness of the path replanner were verified in the Monte Carlo simulations that used the REMUS 100 AUV model. Next, the path replanner was implemented in an AUV system by using an open-source system architecture, MOOS-IvP. The implemented path replanner was verified in a hardware in-the-loop (HIL) test of an Explorer AUV. The path replanner seamlessly worked in conjunction with the hardware on the test platform in real time to continuously generate a time-optimal path. Based on the experiments that involved different sensor configurations and test scenarios, the path replanner demonstrated its scalability for missions that required different setups of onboard sensors and for AUVs of different sizes. It also showed its versatility to accept different current profile data for the generation of time-optimal paths. The resultant path replanner developed in this thesis is practical and promotes ease of applications in field operations of AUVs. It contributes to enabling an AUV to achieve a higher level of autonomy and hence improve its competence in missions with longer durations