Zhi Quan Leong

This paper proposes a methodology for the generation and registration of three-dimensional data sets to support an adaption of Teach and Repeat path following for an Autonomous Underwater Vehicle (AUV) equipped with a multibeam sonar system. The goal of this system is to enable an AUV to generate a topological map of a path consisting of locally consistent sub maps and to re-follow this path using newly collected data. For AUVs traversing long distances without external navigational aids, this methodology would allow robust return-to-home capability, specifically in remote and harsh environments such as beneath ice.

This paper proposes a methodology for the generation and registration of three-dimensional data sets to support an adaption of teach and repeat path following for an Autonomous Underwater Vehicle (AUV) equipped with a multibeam sonar. The goal of this system is to enable an AUV to generate a topological map of a path consisting of locally consistent sub maps and to re-follow this path using newly collected data. For AUVs traversing long distances without external navigational aids, this methodology would allow robust return-to-home capability, specifically in remote and harsh environments such as beneath ice.

Computational fluid dynamics is used to analyze the influence of the horseshoe vortex on the wake features of a simplified geometry representing an underwater vehicle sail (i.e. Rood wing). The sail wake features are of interest as they influence the performance of the downstream components of an underwater vehicle such as the aft appendages and propeller. This paper uses the Rood wing, a generic wing body, mounted on a flat plate as its low aspect ratio is comparable to the underwater vehicle sail and there are substantial published experimental data for validation. Two main simulation schemes were adopted in this paper, i.e. the Reynolds-averaged Navier–Stokes (RANS) and hybrid RANS–large Eddy simulation (LES) incorporating several turbulence models. Both schemes were also examined in their ability to predict the downstream wake features as the literature available to date have primarily focused only on the near-field flow features around the wing root. Three main parameters were investigated including the pressure distribution along the wing’s body, the mean streamwise velocity, and its root mean square fluctuation at three different downstream planes, two in the near field and one in the far field. Results show that the RANS and the hybrid RANS–LES models are capable of predicting the wing-body pressure distribution and the paths of the horseshoe vortex (HSV) as it moves downstream with acceptable numerical dissipation. It was found that different models provided higher accuracy when compared to the experiment depending on the downstream location of the plane. The re-normalization group k-epsilon model with enhanced wall treatment (RNG KE-EN) model captured the wake properties with the highest accuracy within the near field, while further downstream (in the far field), the scale adaptive simulation (SAS) model predicted the flow field with the highest accuracy followed by the RNGKE-EN model.

This paper presents the numerical study on the hydrodynamic coefficients of the submarine hull form “DARPA SUBOFF” when deeply submerged and near the free surface when travelling straight ahead and at a range of drift angles. The Computational Fluid Dynamics (CFD)-based numerical model was validated with experimental data existing in the public domain for the SUBOFF, travelling over a range of drift angles when deeply submerged at a constant speed and in a straight line near the free surface over a range of speeds. The free surface effect on the hydrodynamic behaviour of the SUBOFF was then investigated throughout a range of speeds, drift angles and submergence depths. Results show that the effects of the free surface diminish rapidly with submergence and the near free surface hydrodynamic behaviour of the SUBOFF is highly Froude number dependent.

Non-parametric system identification with Gaussian Processes for underwater vehicles is explored in this research with the purpose of modelling autonomous underwater vehicle (AUV) dynamics with low amount of data. Multi-output Gaussian processes and its aptitude to model the dynamic system of an underactuated AUV without losing the relationships between tied outputs is used. The simulation of a first-principles model of a Remus 100 AUV is employed to capture data for the training and validation of the multi-output Gaussian processes. The metric and required procedure to carry out multi-output Gaussian processes for AUV with 6 degrees of freedom (DoF) is also shown in this paper. Multi-output Gaussian processes are compared with the popular technique of recurrent neural network show that Multi-output Gaussian processes manage to surpass RNN for non-parametric dynamic system identification in underwater vehicles with highly coupled DoF with the added benefit of providing a measurement of confidence.

This paper examines the methodology for the Reynolds Averaged Navier-Stokes (RANS) based Computational Fluid Dynamics (CFD) prediction of the flow around a propeller operating in a uniform flow. The work investigates the influence of grid and physics setup on the prediction of propeller propulsion properties (i.e. thrust (Kt) and torque (Kq) coefficients). The investigation includes the selection of the turbulence model, near wall gridding and state condition to achieve predictions comparable to experimental measurements. The prediction was validated against the experimental data from the open water propeller test conducted at the Australian Maritime College Cavitation Tunnel (AMCCT). The results show the CFD predictions were in good agreement with the experimental measurements (i.e. less than 5% at up to J = 1.1 and 10% at up to J = 1.19 for Kt and Kq). This confirms the presented methodology can be adopted to provide high fidelity prediction for the hydrodynamic characteristics of the propeller.

Underwater vehicles are employed in the exploration of dynamic environments where tuning of a specific controller for each task would be time-consuming and unreliable as the controller depends on calculated mathematical coefficients in idealised conditions. For such a case, learning task from experience can be a useful alternative. This paper explores the capability of probabilistic inference learning to control autonomous underwater vehicles that can be used for different tasks without re-programming the controller. Probabilistic inference learning uses a Gaussian process model of the real vehicle to learn the correct policy with a small number of real field experiments. The use of probabilistic reinforced learning looks for a simple implementation of controllers without the burden of coefficients calculation, controller tuning or system identification. A series of computational simulations were employed to test the applicability of model-based reinforced learning in underwater vehicles. Three simulation scenarios were evaluated: waypoint tracking, depth control and 3D path tracking control. The 3D path tracking is done by coupling together a line-of-sight law with probabilistic inference for learning control. As a comparison study LOS-PILCO algorithm can perform better than a robust LOS-PID. The results shows that probabilistic model based reinforced learning is a possible solution to motion control of underactuated AUVs as can generate capable policies with minimum quantity of episodes.