CN113869148A - Method of trigger type task management system based on autonomous underwater vehicle - Google Patents

Method of trigger type task management system based on autonomous underwater vehicle Download PDF

Info

Publication number
CN113869148A
CN113869148A CN202111066141.XA CN202111066141A CN113869148A CN 113869148 A CN113869148 A CN 113869148A CN 202111066141 A CN202111066141 A CN 202111066141A CN 113869148 A CN113869148 A CN 113869148A
Authority
CN
China
Prior art keywords
path
target
target object
real
module
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Pending
Application number
CN202111066141.XA
Other languages
Chinese (zh)
Inventor
于菲
刘继鑫
何波
沈钺
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Qingdao Pengpai Ocean Exploration Technology Co ltd
Original Assignee
Qingdao Pengpai Ocean Exploration Technology Co ltd
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Qingdao Pengpai Ocean Exploration Technology Co ltd filed Critical Qingdao Pengpai Ocean Exploration Technology Co ltd
Priority to CN202111066141.XA priority Critical patent/CN113869148A/en
Publication of CN113869148A publication Critical patent/CN113869148A/en
Pending legal-status Critical Current

Links

Images

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/24—Classification techniques
    • G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/04—Architecture, e.g. interconnection topology
    • G06N3/045—Combinations of networks
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/04—Architecture, e.g. interconnection topology
    • G06N3/048—Activation functions
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/08—Learning methods

Abstract

The invention discloses a method of a trigger type task management system based on an autonomous underwater vehicle, wherein the task management system comprises an online identification module and a real-time path planning module, the online identification module carries an improved neural network W-ShuffleNet to identify an ocean target online, and the real-time path planning module realizes real-time planning of a task navigation path according to information identified by the online identification module; according to the scheme, the improved lightweight network W-ShuffleNet is carried, a wide activation method is combined, the access flexibility is improved, a real-time path planning module is combined to determine global path planning points, a feasible path set is established, the target position is confirmed again according to a feasible path point set conversion equation, path re-planning is carried out, and finally more accurate target positioning and task management are achieved.

Description

Method of trigger type task management system based on autonomous underwater vehicle
Technical Field
The invention belongs to the technical field of submarine investigation, and particularly relates to a task management system and method based on an autonomous underwater vehicle.
Background
Oceanographic surveys have important value in the fields of oceanography, biology and geophysical. Thus, perception, modeling, sampling, and prediction of the environment in marine surveys is a challenging task. The AUV has the advantages of wide range of motion, good maneuverability, good safety, intellectualization and the like, and becomes a necessary tool for completing various underwater tasks, such as marine survey, marine environment prediction, marine structure evaluation, marine sampling and the like.
The traditional ocean survey adopts a highly integrated autonomous unit system, can carry out ocean sampling in an economic and effective mode, and has the key idea that a set of fixed sensors is integrated with a plurality of AUVs, each AUV carries a proper load, and data are collected on a set path. For example, a marine survey using a fixed sensor (e.g., a sonobuoy) or using a mobile sensor (e.g., a towed array), and finally deducing from the large amount of data collected whether the target is present in the area and tracking it to perform the marine survey. However, since the environmental survey task requires a long continuous deployment, the AUV cluster work increases costs.
At present, when an AUV is used for carrying out a marine survey task, a route is planned or designed on line in advance, the algorithm mode is single, the algorithm belongs to an off-line algorithm, once the task is started, the change cannot be carried out, the AUV can only move according to a specified track and sample the sea by using a sensor, and the AUV lacks initiative.
In order to reduce the energy consumption of the AUV, identify a target object in real time during marine investigation and perform accurate positioning, and reduce human cost, a new task management system is urgently needed to be provided for managing the investigation of an AUV task stage (identification and path planning) on marine environment so as to trigger a corresponding mechanism according to the real-time marine environment, ensure that only interested areas are investigated after online identification, and reduce time waste to optimize paths irrelevant to tasks.
Disclosure of Invention
The invention provides a task management system and method based on an autonomous underwater vehicle to solve the defects in the prior art, so as to autonomously identify and accurately position a target area and realize more accurate task management.
The invention is realized by adopting the following technical scheme: a method of a trigger type task management system based on an autonomous underwater vehicle comprises an online identification module and a real-time path planning module, wherein the online identification module carries an improved neural network W-ShuffleNet to identify an ocean target online, and the real-time path planning module realizes real-time planning of a task navigation path according to information identified by the online identification module; the method comprises the following steps:
step A, obtaining a sample data set: acquiring a corresponding target based on an AUV carrying a side scan sonar, taking a picture containing a target object as a positive sample and a picture not containing the target object as a negative sample, and dividing all the positive samples and the negative samples into a training set, a verification set and a test set in proportion;
b, training a data set: b, training the improved neural network W-ShuffleNet carried by the online identification module based on the sample data set obtained in the step A to obtain a trained online identification module;
step C, target identification: the underwater vehicle navigates under the initial path of the real-time path planning module, and the acquired side scan sonar data is preprocessed by using the trained online recognition module so as to recognize the target object, wherein the recognition principle is as follows:
step C1, preprocessing acquired sonar data such as converting acoustic data into pixel data, cutting images, enhancing images and the like;
step C2, inputting the preprocessed image into a trained W-ShuffleNet model, extracting deep texture information and characteristics in the image, and judging whether the image is a target object;
step C3, when the target object is identified, the output of the model is 1, and the position of the target point is calculated; when the target object is not identified, the output of the model is 0;
step D, after the online identification module identifies the target object, sending the identification result to the real-time path planning module, carrying out re-identification and confirmation on the position of the target object by the real-time path planning module, screening to determine whether the target exists in the area, and if the target exists in the area, changing the original track by the AUV; if not, returning to the initial default track.
Further, the step D is specifically realized by the following steps:
(1) firstly, a path point of a global task area receives and identifies a target object and a given longitude and latitude point;
in the task area, a path point set is set, and a state vector at the moment k is defined as:
Figure BDA0003258353110000021
wherein x iskAnd ykIn the form of a global coordinate position,
Figure BDA0003258353110000022
and
Figure BDA0003258353110000023
taking a point P0 which is closest to the current position of the AUV in geometric distance as an original point as a state vector of a coordinate position, taking a straight line where a long side of a task area is located as an x-axis, and taking a straight line where a short side of the task area is located as a y-axis to establish a rectangular coordinate system xoy;
assuming that the target is traveling at a constant speed, the global points are therefore:
xk＝Fxk-1
wherein F is a state transition matrix;
(2) judging the received identification information during the course, and changing the original global path after receiving the target object;
after the online identification module identifies the target object, changing the path points of the original global planning, and moving to the target point position provided by the online identification module, wherein the set of the changed path points is as follows:
Figure BDA0003258353110000024
wherein, f (x) is a transformation equation of the feasible path point set at the current moment, namely, the original default global path point set is changed, and the path point set is confirmed again aiming at the target position;
(3) and after triggering, replanning the original default path, wherein the AUV considers that the area is worthy of investigation until the task is finished, or the triggering mechanism is closed, and the AUV returns to the default baseline.
Further, in the step C2, when determining whether the target object is the target object, two operations of widely activating and calculating with the same complexity are performed, specifically:
(1) broad activation: converting pixel data of the image into corresponding array information, expanding a data channel before the array information enters an active layer, and extracting the characteristics and deep information of the acoustic image:
assuming the mapping is W1 and the width of the residual net block before activation is W2, then:
W2＝W1*r
wherein r is an expansion factor, and advanced and wide activation is realized before each residual error network;
(2) calculations of the same complexity are kept:
Figure BDA0003258353110000031
where W1 is the mapping and W2 is the width of the residual net block before activation.
Compared with the prior art, the invention has the advantages and positive effects that:
according to the scheme, the AUV is ensured to investigate the interested region only by improving the design of the online identification module and the real-time path planning module, the online identification module carries an improved lightweight network W-ShuffleNet, and by combining a wide activation method, the access flexibility can be improved, additional advantages in the aspects of safety, privacy and energy consumption are obtained, better feature representation in the identification process is effectively ensured, and errors are reduced; the real-time path planning module confirms the target position again according to the feasible path point set conversion equation by determining the global path planning points and establishing a feasible path set, and performs path re-planning to finally realize more accurate target positioning and task management.
Drawings
FIG. 1 is a schematic diagram of a task management system and a marine survey process according to an embodiment of the invention;
FIG. 2 is a schematic diagram of the principle of the extensive activation method based on ShuffleNet in the embodiment of the present invention;
FIG. 3 is a schematic diagram of a real-time path planning module according to an embodiment of the present invention;
FIG. 4 is a schematic diagram of a simulation experiment result for verifying an online identification module;
FIG. 5 is a schematic diagram of a simulation experiment result of a verification real-time path planning module;
FIG. 6 is a diagram illustrating experimental results of a sea trial according to an embodiment of the present invention.
Detailed Description
In order to make the above objects, features and advantages of the present invention more clearly understood, the present invention will be further described with reference to the accompanying drawings and examples. In the following description, numerous specific details are set forth in order to provide a thorough understanding of the present invention, however, the present invention may be practiced in other ways than those described herein, and thus, the present invention is not limited to the specific embodiments disclosed below.
Task management system MMS: session Management System;
an online identification module: OR module, online recommendation;
a real-time path planning module: an RTPP module, real-time path playing;
broad activation: wide activation;
the embodiment discloses a method for a trigger type task management system based on an autonomous underwater vehicle, and as shown in fig. 1, the task management system comprises an online identification module and a real-time path planning module, wherein the online identification module carries an improved lightweight network W-shuffle net to identify an ocean target online; the real-time path planning module realizes real-time planning of the task navigation path according to the information identified by the online identification module, and is triggered only in the interested area, so that the working efficiency of the AUV is improved.
In the scheme, in order to meet the real-time processing capability of the AUV side-scan sonar, an improved neural network W-ShuffleNet is carried by the online identification module in consideration of the movement of an underwater vehicle and the environment of limited resources, and the improved network is a lightweight network and can support the online identification of the side-scan sonar image. And the idea of wide activation is combined in the ShuffleNet structure, so that a large amount of effective information can be better reserved, the hidden information of the low-quality side-scan sonar image can be fully mined, and the identification precision can be effectively improved.
The method specifically comprises the following steps:
step A, obtaining a sample data set: aiming at task requirements, a sample data set needs to be obtained firstly, a model is trained, an AUV carrying a side scan sonar is used specifically to collect corresponding targets, the targets such as sand waves and artificial fish reefs use a picture containing the targets as a positive sample, a picture without the targets as a negative sample, and all the positive and negative samples are according to 6: 2: 2, dividing the ratio into a training set, a verification set and a test set;
step B, training a data set: based on the improved neural network W-ShuffleNet training, the method can extract deep information of a target object in a picture and keep the same complexity of a calculation method aiming at the condition of high noise of an acoustic image, and is specifically realized as follows:
the image is input into the network, pixel information of the image is converted into array information through convolution operation, and the array channel information is expanded before RELU activation of the deep neural network, wherein the expansion aims to ensure that image characteristics cannot be fully reserved due to the fact that channels of original ShuffleNet are shallow before each activation layer, and aiming at the conditions of large noise of an acoustic image and insufficient color information, wide activation is adopted for reserving more image texture information, and deeper information is reserved. However, since the network is widely expanded and the computational load is increased, the same complexity is maintained in the calculation, and the load on the hardware system is reduced in order to maintain the same amount of calculation.
And (4) training the training set once, using the verification set to verify until the accuracy of the verification set is high and stable, stopping training, using the test set to test the trained model at the moment, and selecting the model to load into the AUV after obtaining high accuracy.
Step C, target object identification: loading the model trained in the last step into an AUV (autonomous Underwater vehicle), acquiring sonar data in real time by the AUV, inputting the sonar data into an online identification model for identification, wherein the module is called an OR module and comprises the following specific steps:
(1) and preprocessing the acquired sonar data such as converting the acoustic data into pixel data, cutting the image, enhancing the image and the like.
(2) And inputting the processed image into a trained W-ShuffleNet model, extracting deep texture information and characteristics in the image, and identifying whether the image is a target object.
The model converts the pixel data of the image into corresponding array information, expands the data channel before the array information enters the activation layer, extracts the characteristics and deep information of the acoustic image, and the part is widely activated.
The extensive activation is mainly to expand the features before activation and reduce the features of the residual identity mapping path at the same time, as shown in fig. 2B: assume that the mapping is W1, and the width of the residual net block before activation is W2. In most networks, W1 ═ W2, where the parameter is 2 × W12In FIG. 2B (a), the conventional residual network is widely activated, as shown in FIG. 2B (b)
W2＝W1*r
Wherein r is a spreading factor. I.e. advanced wide activation before each residual network, ensuring that more information in the upper layer of the network is received.
After wide activation, the model has higher identification accuracy, but the calculation burden of an AUV hardware platform is increased, and the network is loaded in an embedded system of the AUV, so the network has high requirements on calculation amount and real-time performance, requires low calculation complexity and high real-time performance, increases the complexity after wide activation, reduces the real-time performance, and causes delay in the aspects of subsequent path decision and the like. Therefore, the W-ShuffleNet model carries out calculation with the same complexity, so that the AUV has higher accuracy on the target and the calculation amount of the AUV hardware platform cannot be increased;
the network original parameters are:
2*W1*W2＝2*W12*r2
extensive activation, while reducing the loss of data characteristics of the network due to excessive depth, also increases the computational load of the parameters. When the input size is fixed, in order to have the same complexity, as in fig. 2b (c), the present embodiment performs the following calculation:
Figure BDA0003258353110000051
through the transformation, the feature extraction can be better carried out after the wide activation is introduced, the same parameter calculation amount of the network before the improvement is ensured, the accuracy is ensured, the low calculation complexity is also ensured, and the rapid and efficient identification is achieved.
(3) When the target object is identified, the output of the model is 1, and the position of the target point is calculated; when the target object is not recognized, the model output is 0.
And D, when the OR module aims at the condition that the identification result is 1, sending the identification result to the RTPP module, activating the RTPP module at the moment, changing an initial default path, carrying out re-identification and confirmation on the position of a provided target point by the RTPP module, and screening to determine whether the area has a target according to a judgment mechanism:
when planning a path, the method specifically comprises the following steps:
(1) firstly, a path point of a global task area receives and identifies a target object and a given longitude and latitude point;
in a task area, a path point set is set, traversal investigation is carried out on the area according to a weeder mode by default, and a state vector at the moment k is defined as:
Figure BDA0003258353110000061
wherein x iskAnd ykIn the form of a global coordinate position,
Figure BDA0003258353110000062
and
Figure BDA0003258353110000063
and taking a point P0 which is closest to the current position of the AUV in geometric distance as an origin point for a state vector of the coordinate position, taking a straight line where the long side of the task area is located as an x-axis, and taking a straight line where the short side of the task area is located as a y-axis to establish a rectangular coordinate system xoy.
Assuming that the target is traveling at a constant speed, the global points are therefore:
xk＝Fxk-1
where F is the state transition matrix
Figure BDA0003258353110000064
θ is the angle of counterclockwise rotation of the target about the origin.
(2) Judging the received identification information during the course, and changing the original global path after receiving the target object;
when the OR module identifies an object, it is determined whether the object is classified as being generated by an object of interest, by an ambiguous interferer, OR by clutter.
When the OR module identifies that the target exists around the AUV track, the route points of the original global planning are changed, the route points go to the position of the target point provided by the OR module, and the changed route point set is as follows:
Figure BDA0003258353110000065
and f (x) is a transformation equation of the feasible path point set at the current moment, namely, the original default global path point set is changed, and the target position is confirmed again.
In this embodiment, assuming that the OR module initially identifies a target with a probability of accuracy α, when the probability of accuracy of identifying the target again heading to the target position is β, the set gating threshold is γ, and after identifying the target each time, the target is screened:
Figure BDA0003258353110000071
(1) when the reliability of alpha and beta is greater than gamma, and C is 1, the area is considered to be 1, the reliability of the existing target object is high, namely the AUV marks the target with interest, and the AUV returns to the original default path again to continue detection;
(2) and when the alpha and the beta credibility are both smaller than gamma, or one of the alpha and the beta credibility is smaller than gamma, the 1 identified in the previous step is considered as false identification, and at the moment, the AUV returns to the original default path again to continue the detection.
The AUV collects marine environment information, optimizes the path of the collected information by using an online data processing technology, and decides an optimal path. For the recognition result of the OR module, it is very important that the AUV selects an optimal path that makes the AUV have the least time consumption and the best efficiency in order to improve the efficiency of the underwater autonomous safe task execution. After the AUV identifies that an interested target exists around the track, considering the complexity of the external environment and the dynamic characteristics of the AUV, the fact that the AUV can perform self-adaptive learning in a continuous space is very important, and therefore a real-time path planning algorithm is adopted, as shown in FIG. 3, an optimal path is evaluated and selected based on the minimum energy consumption principle to correct an original task path, and the AUV continues the process until the task is completed;
(3) after triggering, replanning the original default path, wherein the AUV considers that the area is worthy of investigation until the task is finished, or the triggering mechanism is closed, and the AUV returns to the default baseline;
in the phase, the OR module continuously identifies the environment, and acquires the ocean information acquired in the AUV driving process again until a new round of 'initial judgment' is entered.
The hardware implementation platform of the task management system in this embodiment is as follows:
taking Sailfish-324AUV autonomously developed in China ocean university underwater vehicle laboratories and intelligent perception and machine learning laboratories as an example, verification is performed based on a side scan sonar data set (sand waves and artificial reefs) acquired by the AUV. Wherein, the training loss, training precision and verification precision based on the ShuffleNet and W-ShuffleNet side scan sonar sand wave data set and the artificial fish reef data set are shown in figure 4, figure 4A is the test result of sand waves, figure 4B is the test result of artificial fish reef, it can be seen that W-ShuffleNet has better loss convergence and higher accuracy than ShuffleNet, and the performance is greatly improved.
Simulation of the RTPP module is shown in fig. 5, fig. 5A is a conventional method, fig. 5B is a method proposed by the present invention, where i-iv are trigger points, i.e. the module is triggered only at the location where there is an object, and only the object of interest is detected. The actual sea test result is shown in fig. 6, the identification capability of the OR module for the sand waves of the target object and the trigger type path planning capability of the RTPP module are verified, and the sea test result proves that, when a target is identified, the AUV changes the original default path and only detects the target of interest.
The above description is only a preferred embodiment of the present invention, and not intended to limit the present invention in other forms, and any person skilled in the art may apply the above modifications or changes to the equivalent embodiments with equivalent changes, without departing from the technical spirit of the present invention, and any simple modification, equivalent change and change made to the above embodiments according to the technical spirit of the present invention still belong to the protection scope of the technical spirit of the present invention.

Claims (3)

1. The method is characterized in that the task management system comprises an online identification module and a real-time path planning module, the online identification module carries an improved neural network W-ShuffleNet to identify an ocean target online, and the real-time path planning module realizes real-time planning of a task navigation path according to information identified by the online identification module; the method comprises the following steps:
step A, obtaining a sample data set: acquiring a corresponding target based on an AUV carrying a side scan sonar, taking a picture containing a target object as a positive sample and a picture not containing the target object as a negative sample, and dividing all the positive samples and the negative samples into a training set, a verification set and a test set in proportion;
b, training a data set: b, training the improved neural network W-ShuffleNet carried by the online identification module based on the sample data set obtained in the step A to obtain a trained online identification module;
step C, target identification: the underwater vehicle navigates under the initial path of the real-time path planning module, and the acquired side scan sonar data is preprocessed by using the trained online recognition module so as to recognize the target object, wherein the recognition principle is as follows:
step C1, preprocessing acquired sonar data such as converting acoustic data into pixel data, cutting images, enhancing images and the like;
step C2, inputting the preprocessed image into a trained W-ShuffleNet model, extracting deep texture information and characteristics in the image, and judging whether the image is a target object;
step C3, when the target object is identified, the output of the model is 1, and the position of the target point is calculated; when the target object is not identified, the output of the model is 0;
step D, after the online identification module identifies the target object, sending the identification result to the real-time path planning module, carrying out re-identification and confirmation on the position of the target object by the real-time path planning module, screening to determine whether the target exists in the area, and if the target exists in the area, changing the original track by the AUV; if not, returning to the initial default track.
2. The method for autonomous underwater vehicle based triggered task management system of claim 1, characterized in that: the step D is specifically realized by the following steps:
(1) firstly, a path point of a global task area receives and identifies a target object and a given longitude and latitude point;
in the task area, a path point set is set, and a state vector at the moment k is defined as:
Figure FDA0003258353100000011
the method comprises the steps of taking a point which is closest to the current position of the AUV in geometric distance as an original point, taking a straight line where a long side of a task area is located as an x axis, and taking a straight line where a short side of the task area is located as a y axis to establish a rectangular coordinate system xoy, xkAnd ykIn the form of a global coordinate position,
Figure FDA0003258353100000012
and
Figure FDA0003258353100000013
a state vector that is a coordinate position;
assuming that the target is traveling at a constant speed, the global points are therefore:
xk＝Fxk-1
wherein F is a state transition matrix;
(2) judging the received identification information during the course, and changing the original global path after receiving the target object;
after the online identification module identifies the target object, changing the path points of the original global planning, and moving to the target point position provided by the online identification module, wherein the set of the changed path points is as follows:
Figure FDA0003258353100000021
wherein, f (x) is a transformation equation of the feasible path point set at the current moment, namely, the original default global path point set is changed, and the path point set is confirmed again aiming at the target position;
(3) and after triggering, replanning the original default path, wherein the AUV considers that the area is worthy of investigation until the task is finished, or the triggering mechanism is closed, and the AUV returns to the default baseline.
3. The method for autonomous underwater vehicle based triggered task management system of claim 5, characterized in that: in the step C2, when determining whether the target object is the target object, two operations of widely activating and maintaining the same complexity are specifically:
(1) broad activation: converting pixel data of the image into corresponding array information, expanding a data channel before the array information enters an active layer, and extracting the characteristics and deep information of the acoustic image:
assuming the mapping is W1 and the width of the residual net block before activation is W2, then:
W2＝W1*r
wherein r is an expansion factor, and advanced and wide activation is realized before each residual error network;
(2) calculations of the same complexity are kept:
Figure FDA0003258353100000022
where W1 is the mapping and W2 is the width of the residual net block before activation.
CN202111066141.XA 2021-09-13 2021-09-13 Method of trigger type task management system based on autonomous underwater vehicle Pending CN113869148A (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202111066141.XA CN113869148A (en) 2021-09-13 2021-09-13 Method of trigger type task management system based on autonomous underwater vehicle

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202111066141.XA CN113869148A (en) 2021-09-13 2021-09-13 Method of trigger type task management system based on autonomous underwater vehicle

Publications (1)

Publication Number Publication Date
CN113869148A true CN113869148A (en) 2021-12-31

Family

ID=78995358

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202111066141.XA Pending CN113869148A (en) 2021-09-13 2021-09-13 Method of trigger type task management system based on autonomous underwater vehicle

Country Status (1)

Country Link
CN (1) CN113869148A (en)

Cited By (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN115586777A (en) * 2022-11-04 2023-01-10 广西壮族自治区水利电力勘测设计研究院有限责任公司 Unmanned ship remote measurement control method for water depth measurement

Cited By (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN115586777A (en) * 2022-11-04 2023-01-10 广西壮族自治区水利电力勘测设计研究院有限责任公司 Unmanned ship remote measurement control method for water depth measurement

Similar Documents

Publication Publication Date Title
Paull et al. AUV navigation and localization: A review
US20190033447A1 (en) Systems and methods for detecting objects in underwater environments
CN103033817B (en) Obstruction automatic recognition system for collision preventing of large-scale autonomous underwater vehicle (AUV)
Huang et al. Efficient seafloor classification and submarine cable route design using an autonomous underwater vehicle
Palomeras et al. Automatic target recognition for mine countermeasure missions using forward-looking sonar data
CN109213204B (en) AUV (autonomous underwater vehicle) submarine target searching navigation system and method based on data driving
Williams et al. Adaptive underwater sonar surveys in the presence of strong currents
Williams AUV-enabled adaptive underwater surveying for optimal data collection
CN105787886A (en) Multi-beam image sonar-based real-time image processing method
Pyo et al. Beam slice-based recognition method for acoustic landmark with multi-beam forward looking sonar
Rahmati et al. Slam-based underwater adaptive sampling using autonomous vehicles
CN113869148A (en) Method of trigger type task management system based on autonomous underwater vehicle
Bagnitckii et al. A survey of underwater areas using a group of AUVs
Alla et al. Vision-based Deep Learning algorithm for Underwater Object Detection and Tracking
Loebis et al. Review of multisensor data fusion techniques and their application to autonomous underwater vehicle navigation
Li et al. Real-time underwater target detection for AUV using side scan sonar images based on deep learning
Foresti et al. A voting-based approach for fast object recognition in underwater acoustic images
Praczyk et al. Concept and first results of optical navigational system
Zhang et al. Object detection and tracking method of AUV based on acoustic vision
Inzartsev et al. Detection and inspection of local bottom objects with the help of a group of special-purpose AUVs
CN112880678A (en) Unmanned ship navigation planning method in complex water area environment
Maurelli et al. Integrated MCM missions using heterogeneous fleets of AUVs
Stuckey Navigational error reduction of underwater vehicles with selective bathymetric SLAM
Sung et al. Crosstalk noise detection and removal in multi-beam sonar images using convolutional neural network
Bykova et al. Safe navigation algorithm for autonomous underwater vehicles

Legal Events

Date Code Title Description
PB01 Publication
PB01 Publication
SE01 Entry into force of request for substantive examination
SE01 Entry into force of request for substantive examination