Glider Position Research Articles

Artificial Neural Network for Glider Detection in a Marine Environment by Improving a CNN Vision Encoder

Despite major economic and technological advances, much of the ocean remains unexplored, which has led to the use of remotely operated vehicles (ROVs) and gliders for surveying. ROVs and underwater gliders are essential for ocean data collection. Gliders, which control their own buoyancy, are particularly effective unmanned platforms for long-term observations. The traditional method of recovering the glider on a small boat is a risky operation and depends on the skill of the workers. Therefore, a safer, more efficient, and automated system is needed to recover them. In this study, we propose a lightweight artificial neural network for underwater glider detection that is efficient for learning and inference. In order to have a smaller parameter size and faster inference, a convolutional neural network (CNN) vision encoder in an artificial neural network splits an image of a glider into a number of elongated patches that overlap to better preserve the spatial information of the pixels in the horizontal and vertical directions. Global max-pooling, which computes the maximum over all the spatial locations of an input feature, was used to activate the most salient feature vectors at the end of the encoder. As a result of the inference of the glider detection models on the test dataset, the average precision (AP), which indicates the probability that an object is located within the predicted bounding box, shows that the proposed model achieves AP = 99.7%, while the EfficientDet-D2 model for comparison of detection performance achieves AP = 69.2% at an intersection over union (IOU) threshold of 0.5. Similarly, the proposed model achieves an AP of 78.9% and the EfficientDet-D2 model achieves an AP of 50.5% for an IOU threshold of 0.75. These results show that accurate prediction is possible within a wide range of recall for glider position inference in a real ocean environment.

Development of hybrid neural network and current forecasting model based dead reckoning method for accurate prediction of underwater glider position

Buoyancy-driven underwater gliders are considered to be advanced platforms for large-scale ocean exploration. However, it is greatly affected by ocean currents, and traditional dead reckoning navigation methods are inadequate in accurately predicting its own position, causing great difficulties in underwater target detection. To solve this problem, a navigation estimation method based on the current prediction model and a convolutional neural network-long short-term memory hybrid neural network is proposed in this paper to predict the position of underwater gliders. Analysis of the experimental results shows that the hybrid neural network model trained with underwater glider sea trial data and simulated motion data can predict the glider speed more accurately than the current-assisted dead reckoning navigation, and can predict the relative position of the glider more accurately with the updated feedback from the current prediction model. The test results show that the data-driven prediction method can greatly help to predict the position of underwater gliders in the absence of other underwater positioning and navigation equipment.

Multi-objective optimization for control parameters of underwater gliders considering effect of uncertain input errors

In actual application, the energy utilization rate of underwater glider directly affects the total voyage range. When underwater glider is used for executing exploration mission for a fixed point, the position that the glider resurfaces should be accurate enough. In this paper, we employ a multi-objective optimization method to determine the control parameters values that can maximize the position accuracy that the glider resurfaces and the energy utilization rate simultaneously. Especially, the optimization of this paper considers the effect of uncertain input errors. The control parameters include the net buoyancy adjustment amount and the movable mass block translation amount. The input errors include the control parameters errors, the motion depth error and the current. Based on the dynamic model of an underwater glider, we propose the calculation model and evaluation flow that are used for analyzing the glider position accuracy and energy utilization rate, considering the effect of uncertain input errors. Besides, a combinatorial experimental design method is proposed to calculate the performance evaluation parameters under different control parameters values. Then the radial basis function neural network is employed to establish the surrogate models of performance evaluation parameters to participate in the optimization calculation, which can improve the optimization efficiency. After optimization calculation based on the non-dominated sorting genetic algorithm II, we obtain a Pareto optimal set consisting of 257 sets of non-dominated solutions. Finally, the selection rule of optimal control parameters values is given, and the optimization results are validated under 3 sets of solutions. This research may be valuable for the improvement of the glider work quality.

Standoff tracking control of underwater glider to moving target

The standoff tracking control of underwater glider to moving targets has non-negligible potentials in oceanic observations. To perform such operations, this study presents an integrated standoff tracking method that focuses on dealing with the coupled motions and under-actuated characteristics of the gliders as well as offsetting the dilemma caused by lacking of navigation system. The proposed approach contains guidance algorithm, basic controller, and reinforcement learning optimizer to improve the tracking accuracy. The verified simulation model for the OUC-III glider is utilized to develop the algorithm and evaluate the effectiveness of tracking. The adaptive navigation law called Lyapunov guidance vector field method generates the desired heading in this framework. The guidance only requires positioning information that is accessible through dead-reckoning. For the deficiency that the glider’s state controlled by the proportional integral differential controller can not immediately reach the required state, the reinforcement learning optimizer is introduced. The gliding status constitutes optimizer input while the compensating rotation signals are outputted to actuate the movable mass. The reward function of the optimizer is designed for maintaining the glider position error within 1 m with respect to the standoff circle. On this basis, the objective is set to maximize the rewards and consequently to guarantee the best accuracy. Three reinforcement learning algorithms are adopted and tested in our proposed framework. In the training phase, the training speeds of composite methods and pure reinforcement learning algorithms are compared. To test the stability, we design different types of moving targets, standoff radium and disturbance in the validation simulation, which have not appeared in the training. This study can provide patterns and solutions for the standoff tracking applications of underwater gliders. Also, it enlarges the glider’s scope of deployment based on the existing hardware.

A restricted circle based position keeping strategy for the wave glider

The wave glider is a class of wave-propelled, persistent, unmanned ocean surface vehicle. Due to the characteristics of low speed and weak maneuvering, it is challenging to control the wave glider accurately under the environmental disturbances. This paper focuses on the position keeping of the wave glider in open sea. A novel robust position keeping guidance strategy based on the restricted circle is firstly developed. Subsequently, the position keeping performance of the wave glider under different sea conditions is simulated using an eight-degree-of-freedom mathematical model. The influences of various factors on the positioning capability are investigated, including waves, current, restricted radius of circle, and control mode after entering the restricted circle. The results show that the proposed position keeping algorithm can limit the motion area of the wave glider to a certain range. Besides, under a given sea state, a smaller restricted circle ensures a better positioning accuracy, but requires more energy consumption. Waves and current both have significant impacts on the positioning accuracy and energy consumption for the wave glider position keeping.

Underwater glider localization using broadband source transmissions in the Canada Basin

Two Seaglider autonomous underwater vehicles were deployed as mobile receiving platforms in a 2016–2017 tomography experiment in the Canada Basin. The introduction of gliders into a tomography experiment has the advantage of adding acoustic data receptions at many depths and ranges relative to moored acoustic sources. The challenge in interpreting these data is the lack of statistics at any given location and the uncertainty in the glider position during a dive, the latter resulting in a fundamental ambiguity between the position and sound speed. The acoustic arrival matching localization technique, performed in post-processing, has been employed to position the instruments while underwater using transmissions from moored acoustic sources. This technique capitalizes on the broadband nature of the tomography sources and estimates latitude and longitude while relying upon the vehicle’s pressure sensor for depth. Acoustic ranging was also performed onboard the gliders during the deployment, and estimated ranges from the moored sources were transmitted via Iridium satellite link every few hours when the glider surfaced. Positions resulting from acoustic arrival matching localization will be compared with results based on semi-real time acoustic ranging. Results will be discussed in the context of reviewing the state of the art in underwater glider localization.

The predictability of acoustic receptions on gliders in the Arctic

Acoustic transmissions from shallow sources in the Arctic Ocean can propagate several hundred kilometers due to the presence of an Arctic acoustic duct. Receptions of these long-range transmissions are complex patterns of arrivals which are strongly dependent upon upper ocean sound-speed structure. In addition to measuring sound-speed parameters, gliders equipped with acoustic recorders can measure these arrivals and can complement moored receptions, providing data at many ranges with respect to the moored sources. There is a higher degree of uncertainty in glider position compared with moored receivers, but localization can be improved in post-processing with enhanced acoustic predictability. Two acoustic Seagliders were deployed for a short pilot study in late Summer 2016 in the vicinity of an array of acoustic tomography sources with frequencies on the order of 250 Hz in the Arctic Ocean in anticipation of a longer deployment in Summer 2017. Source receptions recorded on the gliders are compared with acoustic predictions based on sound-speed profiles generated from environmental data collected on the gliders themselves to begin to understand the predictability of transmissions received on gliders in an Arctic environment. Acoustic predictions are analyzed for receptions both within the acoustic duct and at depths exceeding the acoustic duct.

Ocean gliders as tools for acoustic tomography

Ocean gliders are capable and increasingly ubiquitous platforms for collecting oceanographic data such as temperature and salinity measurements. These data alone are useful for ocean acoustic tomography applications, but gliders equipped with acoustic receivers can also record transmissions from acoustic sources, potentially providing additional paths for tomographic inversions. There are challenges associated with the use of gliders as tomography receivers, however, notably the uncertainty in underwater glider position, which can lead to ambiguity between sound-speed and glider position. Glider-based acoustic receptions from moored tomography sources can provide range measurements to aid in subsea glider localization. These data can be used in post-processing for precise glider positioning and in real-time for glider navigation. The current state of the art will be reviewed, and preliminary results will be presented from an experiment in the Arctic Ocean, which seeks to use moored acoustic tomography source signals for both subsea glider positioning and real-time navigation, including navigation under ice. [Work sponsored by ONR.]

Localization and Subsurface Position Error Estimation of Gliders Using Broadband Acoustic Signals at Long Range

Broadband acoustic source transmissions recorded on Seagliders at ranges up to 700 km are used to estimate subsurface glider position. Because the sources transmitted at 9-min intervals the glider moved appreciably between source receptions. Source–glider ranges estimated from acoustic arrivals were combined using least squares analysis to estimate glider position and velocity during each reception period. The analysis was applied to 387 sets of source transmissions using three different flight models of glider subsurface motion for initial position input values. The offsets between the position estimated from the flight models and the acoustically derived position resulting from the inversions were 600–900-m root mean square (rms) depending upon the model and input parameters. The offsets were tripled if the positions from the flight models were not corrected for a dive-averaged current (DAC). Estimates of a posteriori errors ranged from 78–105-m rms and from 9.1–11.6-cm/s rms for glider position and velocity, respectively. Data residuals were on the order of 50-m rms, a dramatic reduction from 178-m rms, which was documented for the case neglecting the motion of the glider between subsequent source transmissions (Van Uffelen , J. Acoust. Soc. Amer., vol. 134, pp. 3260–3271, 2013). Overall horizontal glider speed was estimated to be approximately 21-cm/s rms.

Acoustic localization of seagliders in the Philippine Sea using broadband transmissions from moored acoustic sources

Seagliders obtain precise GPS positioning at surfacing events between dives, but can be underwater for several hours and travel several kilometers during a single dive, during which time less is known about the position of the glider. Four Seagliders deployed in the Philippine Sea from November 2010 to April 2011 were equipped with Acoustic Recorder Systems, enabling them to record over 2000 transmissions from five moored broadband acoustic sources at ranges up to 700 km and depths up to 1000 m. These sources sequentially transmitted linear FM sweeps from approximately 200 to 300 Hz at 9-min intervals. Source-receiver ranges, determined from travel-time offsets of the received pattern of acoustic arrival peaks, are combined for sets of source transmissions using the stochastic inverse method of least squares to estimate the position of the glider at the time of acoustic reception. Because the source transmissions are separated in time and because the glider is in constant motion, it can move appreciably between source receptions. A linear model of glider motion is incorporated into the least-squares analysis to account for this motion and to simultaneously estimate glider velocity. These acoustically derived estimates of glider position are compared with the positions estimated by a glider kinematic model.

Estimating uncertainty in subsurface glider position using transmissions from fixed acoustic tomography sources

Four acoustic Seagliders were deployed in the Philippine Sea November 2010 to April 2011 in the vicinity of an acoustic tomography array. The gliders recorded over 2000 broadband transmissions at ranges up to 700 km from moored acoustic sources as they transited between mooring sites. The precision of glider positioning at the time of acoustic reception is important to resolve the fundamental ambiguity between position and sound speed. The Seagliders utilized GPS at the surface and a kinematic model below for positioning. The gliders were typically underwater for about 6.4 h, diving to depths of 1000 m and traveling on average 3.6 km during a dive. Measured acoustic arrival peaks were unambiguously associated with predicted ray arrivals. Statistics of travel-time offsets between received arrivals and acoustic predictions were used to estimate range uncertainty. Range (travel time) uncertainty between the source and the glider position from the kinematic model is estimated to be 639 m (426 ms) rms. Least-squares solutions for glider position estimated from acoustically derived ranges from 5 sources differed by 914 m rms from modeled positions, with estimated uncertainty of 106 m rms in horizontal position. Error analysis included 70 ms rms of uncertainty due to oceanic sound-speed variability.

Toward subsurface positioning of gliders using fixed acoustic tomography sources

Acoustic Seagliders can be positioned precisely using GPS at the surface, but are underwater and unable to utilize GPS for up to 9 hours at a time as they dive to depths of up to 1000 m. During this time, a kinematic model estimates the position of the glider. Four acoustic Seagliders were deployed in the Philippine Sea November 2010 - April 2011, and received transmissions from five broadband acoustic tomography sources moored in the region. Over 2000 acoustic receptions were recorded at ranges up to 700 km from the moored sources. Measured acoustic arrival peaks were unambiguously associated with ray arrivals predicted using the model-estimated glider position at the time of reception and a mean sound-speed profile. Estimates of source-receiver range uncertainty were calculated from statistics of travel-time offsets between the measured arrivals and the eigenray dispersion patterns. The uncertainty in range between the source and the modeled glider position during a dive is estimated to be 639 m (426 ms) rms disregarding the effects of ocean sound-speed variability, which are anticipated to be on the order of 70 ms rms. The range uncertainty is attributed primarily to advection of the glider by unobserved ocean currents.