Unmanned underwater vehicles (UUVs), including autonomous
underwater vehicles (AUVs) and remotely operated vehicles (ROVs), are used to perform
various tasks regarding ocean and seafloor investigations. One of the key factors to allow
UUVs to complete underwater activities, especially near the seafloor, is the precise control of
UUVs. The development of UUV control rules is mostly based on UUV dynamic models.
However, such dynamic model essentially contains a number of unknown hydrodynamic
parameters. Thus, in order to achieve precise control of UUVs, it is important to have an
effective and accurate system identification method. System identification for a UUV is to
estimate the UUV’s hydrodynamic parameters using experimental data. Planar motion
mechanisms (PMMs), onboard sensors and vision technologies are three main approaches to
acquire data from UUV motion experiments.
The PMM-based identification method requires that motion experiments have to be
conducted using an original or scaled UUV in a large water tank (Avila et al., 2012,
Nakamura et al., 2013, Nomoto and Hattori, 1986). With the measured forces and torques
Page 1 of 19
Cambridge University Press
Journal of Navigation
For Review Only
2
exerted on a UUV, the UUV’s hydrodynamic parameters can be calculated through signal
processing techniques. Even though the PMM based method is the most straightforward
approach for UUV system identification, it is expensive and its accuracy is significantly
affected by the scaling of UUVs.
Depending on the functions of sensors, the onboard sensor-based method can obtain a
UUV’s position data (Caccia et al., 2000, Smallwood and Whitcomb, 2003) or velocity and
acceleration data (Avila et al., 2013, Farrell and Clauberg, 1993, Martin and Whitcomb, 2014,
Valeriano-Medina et al., 2013). Through the measured data and different algorithms, such as
the Kalman filter and least squares (LS), the UUV’s hydrodynamic parameters can be
identified. The onboard sensor-based method is highly cost-effective and repeatable, and is
particularly suitable for the UUVs whose payload and configuration must change to satisfy
the requirements of different tasks (Caccia and Veruggio, 2000).
The vision technology-based method can obtain accurate position data of a UUV, and due
to its low-cost feature, it is an appealing approach to build the navigation system of a UUV
(Gracias et al., 2003, Negahdaripour and Xu, 2002). More importantly, either through an
onboard camera (Ridao et al., 2004) or a camera outside of a water tank (Chen, 2008), the
vision technology-based identification method is both accurate and low-cost.
This paper presents a new method, the Laser Line Scanning for Hydrodynamic Parameter
Identification (LSHPI), which integrates laser line scanning, decoupled dynamics, and
evolutionary optimization to identify the hydrodynamic parameters of an AUV. The concept
of the laser line scanning technique of the LSHPI is based on the method proposed in (Wang
and Cheng, 2007). To extract information from photographs, a widely used approach is the
stereovision technique, which is classified as a passive vision technique and works well
unless the photographs have only smoothly textured areas, repetitive structures, or unclear
images. To overcome this limitation and increase the image signal-to-noise ratio (SNR) is to
adopt an active vision technique, which projects structured light onto the scene and infers
detailed information of various features from the distortion of the structured light in the
image. A light stripe generated by a laser source is a commonly adopted structured light
pattern. Moreover, as for acquiring real-world position information from images, which
directly relates to camera calibration for underwater experiments, the LSHPI adopts the
method proposed in (Wang and Cheng, 2007). The method only uses a board to implement
the calibration scheme, which is easier than other approaches that require a rigid control
frame. In addition, the LSHPI also has the following advantages: (1) it has a high sampling
rate; (2) it is cost-effective; (3) it has high spatial resolution; (4) once the camera calibration
has been done, it does not require the re-calibration for the camera; (5) it does not require the
accurate dimensions of the experimental area.