JP6400437B2 - Floating body positioning system and floating body positioning method - Google Patents

Description

  The present invention relates to a technique for positioning a floating body.

  In the caisson installation work by the winch method, a plurality of winches are installed on the caisson floated on the water, the wire rope fed from the winch is connected to the sinker block installed on the bottom of the water, and the caisson is positioned by operating the winch. However, when the operator gets on the caisson and operates the winch, the caisson is shaken by the waves, so it is difficult to ensure the safety of the operator. In order to solve this problem, it has been proposed to perform caisson positioning by remote control (for example, Patent Document 1). An example of remote operation is that the worker monitors the position and attitude of the caisson using a satellite positioning device, video camera, etc., and sends a control signal indicating the amount of movement from the work ship to the winch on the caisson. Move the caisson to the position.

JP 2004-238914 A

However, in the case of remote operation, the time required for positioning the caisson tends to be longer than when the operator operates the winch on the caisson. The following two points can be considered as the cause.
(1) Compared with the case where the operator operates the winch on the caisson, it is difficult to grasp the relationship between the movement amount of the winch and the actual change amount of the position and posture of the caisson.
(2) The relationship between the amount of movement of the winch and the actual amount of change in the position and orientation of the caisson varies depending on weather conditions, sea conditions, the size of the caisson, and the like.
For these reasons, it has been difficult for an operator to specify the amount of movement of the winch that causes a desired change in the position or posture of the caisson. In addition, this problem is common to positioning of all floating bodies such as ships and buoys as well as caisson.
Therefore, the present invention provides a technique that can easily position a floating body.

  The present invention provides an operation means for performing an operation for changing the position or orientation of a floating body, a detection means for detecting a change amount of the position or orientation of the floating body, and an operation by the operation means for an input layer in a neural network. Input the amount of change detected by the detection means to the output layer in the neural network, the input of the operation amount to the input layer and the input of the change amount to the output layer In the process of iterating, the calculating means for calculating at least one of the coupling load or the threshold value in the neural network, and the amount of movement by the operating means that causes a desired amount of change with respect to the position or posture of the floating body, An operation amount specifying means for specifying using a neural network, and the operation means for operating with the specified operation amount. Providing floating positioning system comprising an instruction unit that instructs the.

  The present invention also provides an operation means for performing an operation for changing the position or orientation of a floating body, a detection means for detecting a change amount of the position or orientation of the floating body, and the detection means for an input layer in a neural network. Is input to the output layer in the neural network, the input of the amount of change to the input layer, and the input of the amount of operation to the output layer. In the process of repeating the above, the calculation means for calculating at least one of the coupling load or the threshold value in the neural network, and the movement amount by the movement means that causes a desired change amount with respect to the position or posture of the floating body. , The operation amount specifying means for specifying using the neural network, and the operation so as to operate with the specified operation amount Providing floating positioning system comprising an instruction unit that instructs the work unit.

  In the above configuration, the operation amount specifying unit acquires the change amount output from the output layer by inputting the arbitrary operation amount to the input layer, and the acquired change amount, By the operation means that generates a desired change amount with respect to the position or posture of the floating body by using a desired change amount with respect to the position or posture of the floating body and the arbitrary movement amount input to the input layer. The operation amount may be specified.

  In the above configuration, the calculation unit repeatedly calculates at least one of the combined load or the threshold value from the start of positioning of one floating body to the end of positioning, and the instruction unit performs positioning from the start of positioning of the one floating body. A neural network including at least one of the combined load and the threshold value calculated until the end may be used only in the positioning process of the one floating body.

  In the above configuration, when the positioning positions of the N-th floating body positioning and the (N + 1) -th floating body positioning are within a predetermined range and the positioning date / time is within a predetermined range, the calculating means The combined load or the threshold value calculated at the end of the floating body positioning process may be used as the initial value of the combined load or the threshold value at the time of the N + 1th floating body positioning start.

  Further, the present invention inputs an operation amount by an operation means for performing an operation for changing the position or posture of the floating body to the input layer in the neural network, and the position or position of the floating body to the output layer in the neural network. In the process of inputting the amount of change detected by the detecting means for detecting the amount of change in posture, and repeating the input of the operation amount to the input layer and the input of the change amount to the output layer, A calculation step for calculating at least one of a combined load and a threshold value, and an operation for specifying an operation amount by the operation means that causes a desired change amount for the position or posture of the floating body using the neural network An amount specifying step and an instruction step for instructing the operation means to operate at the specified operation amount. It provides a floating body positioning method and a-up.

  Further, the present invention inputs the amount of change detected by the detecting means for detecting the amount of change of the position or orientation of the floating body to the input layer in the neural network, and the position of the floating body to the output layer in the neural network. Alternatively, in the process of inputting an operation amount by an operation means for performing an operation for changing the posture, and repeating the input of the change amount to the input layer and the input of the operation amount to the output layer, A calculation step for calculating at least one of a combined load and a threshold value, and an operation for specifying an operation amount by the operation means that causes a desired change amount for the position or posture of the floating body using the neural network An amount specifying step and an instruction step for instructing the operation means to operate at the specified operation amount. It provides a floating body positioning method and a-up.

  According to the present invention, the floating body can be easily positioned.

Top view of caisson 1 The figure which shows the computer 5 and its peripheral device Diagram showing the hierarchical structure model of a neural network Flow chart showing processing executed by computer 5 Diagram showing the hierarchical structure model of a neural network Flow chart showing processing executed by computer 5

<First Embodiment>
An example for carrying out the present invention will be described. In this embodiment, a caisson positioning operation is assumed. A caisson is an example of a floating body according to the present invention.
FIG. 1 is a plan view of the caisson 1. FIG. 2 is a diagram showing the computer 5 and its peripheral devices. The computer 5 is installed in a place other than the caisson 1, such as a work boat or a land command room. On the upper part of the caisson 1, winches 2A, 2B, 2C, 2D, a positioning signal receiving device 3, and a relay device 4 are installed. The winches 2A, 2B, 2C, and 2D are collectively referred to as winches 2. A sinker block 23 is installed on the water bottom around the caisson 1. The wire rope 21 fed out from the winch 2 is hung around a pulley 22 installed on the upper part of the caisson 1 and connected to a sinker block 23. The positioning signal receiving device 3 receives a positioning signal from a navigation satellite of GNSS (Global Navigation Satellite System). The winch 2 and the positioning signal receiving device 3 are connected to the relay device 4 by wired or wireless communication means.

  The computer 5 includes a control unit 51, a storage unit 52, and an external IF (Interface) unit 53. The control unit 51 includes an arithmetic device such as a CPU (Central Processing Unit) and a storage device such as a ROM (Read Only Memory) and a RAM (Random Access Memory). The ROM stores firmware that describes the procedure for starting up hardware and an OS (Operating System). The RAM is used for storing data when the CPU executes an operation. The storage unit 52 includes, for example, a hard disk storage device, and stores an OS, application programs, and the like.

  An input device 54, a display device 55, a printing device 56, and a wireless communication device 57 are connected to the external IF unit 53. The input device 54 is, for example, a keyboard or a mouse. The display device 55 is, for example, a liquid crystal display panel. The printing device 56 is a printer, for example. The wireless communication device 57 performs wireless communication with the relay device 4. The computer 5 transmits a control signal to the winch 2 via the wireless communication device 57 and the relay device 4. The winch 2 performs the operation of winding and feeding the wire rope 21 in accordance with this control signal, and the position or posture of the caisson 1 is changed by this operation. That is, the winch 2 is an example of an operation unit that performs an operation for changing the position or posture of the floating body.

  The positioning signal received from the navigation satellite by the positioning signal receiving device 3 is transmitted to the computer 5 via the relay device 4 and the wireless communication device 57. The computer 5 calculates the position or orientation (azimuth and inclination) of the caisson 1 based on this positioning signal, and detects the amount of change in the position or orientation of the caisson 1 from the calculated value. That is, the computer 5 and the positioning signal receiving device 3 are an example of a detecting unit that detects the amount of change in the position or posture of the floating body.

  When positioning the caisson 1 by remote control, it is difficult to grasp the relationship between the movement amount of the winch 2 and the change amount of the position or posture of the caisson 1 as compared with the case where the operator operates the winch 2 on the caisson 1. . In addition, the relationship between the movement amount of the winch 2 and the change amount of the position or posture of the caisson 1 varies depending on weather conditions, sea conditions, caisson size, and the like. Therefore, it has been difficult for the operator to specify the amount of movement of the winch that causes a desired amount of change in the position or posture of the caisson.

  Therefore, in this embodiment, the amount of change in the position or orientation of the caisson 1 is detected using GNSS or the like, the operation amount by the winch 2 is input to the input layer in the neural network, and the output layer in the neural network is In the process of inputting the detected change amount and repeating the input of the operation amount to the input layer and the input of the change amount to the output layer, at least one of the coupling load and the threshold value in the neural network is calculated. Then, an operation amount by the winch 2 that causes a desired change amount with respect to the position or posture of the caisson 1 is specified using a neural network, and the winch 2 is instructed to operate with the specified operation amount. In specifying the amount of movement, an arbitrary amount of movement is input to the input layer to obtain the amount of change output from the output layer, and the amount of change obtained and the amount of change desired for the position or orientation of the caisson 1 Then, using the movement amount input to the input layer, the movement amount by the winch 2 that causes a desired change amount with respect to the position or posture of the caisson 1 is specified. Specifically, it is as follows.

  FIG. 3 is a diagram illustrating a hierarchical structure model of a neural network. The neural network in the present embodiment is a three-layer hierarchical model including an input layer, an intermediate layer, and an output layer, and uses a sigmoid function as a response function. The movement amount of the winch 2 is input to the input layer, and the estimated change amount of the position and posture of the caisson 1 is output from the output layer. In the following description, for simplification of the drawing, only the values in the horizontal plane (x coordinate, y coordinate, azimuth θ) are shown for the position and orientation of the caisson 1; Coordinate values and tilt angles are also included.

  FIG. 4 is a flowchart showing processing executed by the computer 5. The control unit 51 of the computer 5 executes this process according to the procedure described by the application program stored in the storage unit 52. In the figure, the neural network is abbreviated as ANN (Artificial Neural Network).

In step S01, the control unit 51 calculates the movement amount D _n caisson 1 from mounting target position and the current position. The installation target position is determined in advance. The current position is calculated using the positioning signal received by the positioning signal receiver. The movement amount D _n is represented by the difference between the coordinate value of the installation target position and the coordinate value of the current position.

In step S02, the control unit 51 determines whether or not the movement amount Dn is greater than or _equal to a set value. This set value is a threshold value for determining whether or not the positioning of the caisson 1 is completed. If the movement amount D _n is greater than or _equal to the set value (step S02: YES), the movement of the caisson 1 is necessary, so the processing of the control unit 51 proceeds to step S03. On the other hand, when the movement amount D _n is less than the set value (step S02: NO), the positioning of the caisson 1 is completed, and the control unit 51 ends the process.

In step S03, the control unit 51 inputs the operation amount A _n of the winch 2 to the neural network. Specifically, inputs winch 2A, 2B, 2C, the input layer an operation amount A _n provisional arbitrarily defined for 2D as values I1, I2, I3, I4.

In step S05, the control unit 51 obtains the movement amount _{An + 1} of the winch 2 up to the installation target position. For the winches 2A, 2B, 2C, and 2D, the operation amount _{An + 1} to the installation target position is obtained by the following equation.
A _{n + 1} = D _n × A _n / M _n

In step S06, the control unit 51 inputs the movement amount _{An + 1} of the winch 2 up to the installation target position to the neural network. Specifically, the motion amount _{An + 1} obtained in step S05 for the winches 2A, 2B, 2C, and 2D is input to the input layer as values I1, I2, I3, and I4.

In step S07, the control unit 51 estimates the change amount M _{n + 1} of the position of the caisson 1. Specifically, the amount of change in the position (x coordinate, y coordinate) and posture (azimuth θ) of the caisson 1 is output from the output layer as values O1, O2, and O3 as the estimated change amount M _{n + 1} .

In step S08, the control unit 51 operates the winch 2 by the operation amount _{An + 1} . Specifically, the control unit 51 transmits a control signal indicating the operation amount _{An + 1} to the winches 2A, 2B, 2C, and 2D.

In step S09, the control unit 51 inputs a teacher signal T _n determined from the positioning result to the neural network. Specifically, the control unit 51 obtains a change amount of the position or posture of the caisson 1 based on the positioning signal received from the positioning signal receiving device, and inputs this change amount as the teacher signal T _n .

Second Embodiment
In the present embodiment, the change amount of the position or orientation of the caisson 1 is detected using GNSS or the like, the detected change amount is input to the input layer in the neural network, and the winch 2 is input to the output layer in the neural network. In the process of repeating the input of the change amount to the input layer and the input of the operation amount to the output layer, at least one of the coupling load and the threshold value in the neural network is calculated. Specifically, it is as follows.

  FIG. 5 is a diagram showing a hierarchical structure model of a neural network. The neural network in the present embodiment is a three-layer hierarchical model including an input layer, an intermediate layer, and an output layer, and uses a sigmoid function as a response function. A target change amount of the position and orientation of the caisson 1 is input to the input layer, and an estimated operation amount of the winch 2 is output from the output layer. In the following description, for simplification of illustration, only the values in the horizontal plane (x coordinate, y coordinate, azimuth θ) are shown for the position and orientation of the caisson 1, but in the actual positioning, the vertical direction Coordinate values and tilt angles are also included.

FIG. 6 is a flowchart showing processing executed by the computer 5.
Steps S21 and S22 are the same as steps S01 and 02 in the first embodiment.
In step S23, the control unit 51 inputs the change amount M _n of the position or orientation of the caisson 1 to the neural network. Specifically, the change amounts of the position (x coordinate, y coordinate) and posture (azimuth θ) of the caisson 1 are input to the input layer as values I1, I2, and I3 as the change amount M _n .

In step S24, the control unit 51 estimates the operation amount A _n winch 2. Specifically, winch 2A, 2B, 2C, and outputs from the output layer the operation amount A _n as a value O1, O2, O3, O4 for 2D.

In step S25, the control unit 51 operates the winch 2 by the operation amount A _n. Specifically, the control unit 51 transmits a control signal indicating the operation amount _An to the winches 2A, 2B, 2C, and 2D.

In step S26, the control unit 51 inputs the amount of change M _{n + 1} obtained from the positioning result to the neural network. Specifically, the control unit 51 obtains the change amount M _{n + 1} of the position or orientation of the caisson 1 based on the positioning signal received from the positioning signal receiving device, and uses this change amount M _{n + 1} as the values I1 and I2. , I3 is input to the input layer.

In step S27, the control unit 51 estimates the movement amount _{An + 1} of the winch 2. Specifically, the operation amount _{An + 1} is output from the output layer as the values O1, O2, O3, and O4 for the winches 2A, 2B, 2C, and 2D.

In step S28, the control unit 51, the back propagation operation amount A _n as a teacher signal T _n to operation amount A _{n + 1} modifies the parameters.

<Modification>
The above embodiment may be modified as follows. A plurality of modified examples may be combined.
<1>
The relationship between the movement amount of the winch and the actual movement amount of the caisson varies depending on weather conditions, sea conditions, caisson size, and the like. Accordingly, at least one of the combined load or the threshold is repeatedly calculated from the start of positioning of one caisson to the end of positioning, and at least one of the combined load or threshold calculated from the start of positioning of one caisson to the end of positioning. A neural network including one may be used only in the positioning process of one caisson.

<2>
If conditions related to the location and time of the previous and current positioning are within the specified range, the combined load or threshold calculated based on the previous positioning is used as the initial value of the combined load or threshold at the current positioning. It may be used. That is, when the positioning positions of the N-th floating body positioning and the (N + 1) -th floating body positioning are within a predetermined range and the positioning date / time is within a predetermined range, the calculation is performed at the end of the N-th floating body positioning process. The combined load or threshold value may be used as the initial value of the combined load or threshold value at the start of the N + 1th floating body positioning.

<3>
The number of layers of the hierarchical structure model of the neural network may be any number. That is, a hierarchical structure model provided with a plurality of intermediate layers may be used.

<4>
The present invention may be applied to any floating body such as a ship, a buoy (buoy), a rig, a mega float, and an offshore wind power generator.

<5>
In the above-described embodiment, an example in which the computer 5 is installed in a place other than the caisson 1 such as a work ship or an onshore command room has been shown, but the computer 5 may be installed on the caisson 1.

1 Caisson, 2, 2A, 2B, 2C, 2D winch, 21 wire rope, 22 pulley, 23 sinker block, 3 positioning signal receiving device, 4 relay device, 5 computer, 51 control unit, 52 storage unit, 53 external IF unit 54 input device 55 display device 56 printing device 57 wireless communication device

Claims (7)

An operation means for performing an operation for changing the position or posture of the floating body;
Detecting means for detecting a change amount of the position or posture of the floating body before and after the operation ;
Detected by the detecting means when an arbitrary amount of motion is given to the operating means for a neural network that outputs the amount of change in position or orientation from the output layer when the amount of motion in the operating means is input to the input layer A calculation means for inputting a change amount as a teacher signal and calculating at least one of a coupling load or a threshold value in the neural network by back propagation ;
Using the neural network that includes at least one of the combined load or the threshold value calculated by the calculating means , which is the amount of movement by the moving means that causes a desired amount of change in the position or orientation of the floating body. An operation amount specifying means for specifying
A floating body positioning system comprising: instruction means for instructing the operation means to operate at the specified operation amount. An operation means for performing an operation for changing the position or posture of the floating body;
Detecting means for detecting a change amount of the position or posture of the floating body before and after the operation ;
When input to the input layer of the variation amount detected by said detection means, against the neural network to output an operation amount of the operating means from the output layer, the operation when the amount of change is detected by said detecting means A calculation means for inputting an operation amount given to the means as a teacher signal, and calculating at least one of a coupling load or a threshold in the neural network by backpropagation ;
Using the neural network that includes at least one of the combined load or the threshold value calculated by the calculating means , which is the amount of movement by the moving means that causes a desired amount of change in the position or orientation of the floating body. An operation amount specifying means for specifying
A floating body positioning system comprising: instruction means for instructing the operation means to operate at the specified operation amount. The operation amount specifying means includes:
Acquiring the change amount output from the output layer by inputting the arbitrary operation amount with respect to the input layer,
Using the obtained amount of change, the amount of change desired for the position or posture of the floating body, and the amount of movement input to the input layer, the amount of change desired for the position or posture of the floating body is obtained. The floating body positioning system according to claim 1, wherein an operation amount by the operation means to be generated is specified. The calculation means repeatedly calculates at least one of the combined load or the threshold value from the start of positioning of one floating body to the end of positioning,
Said instructing means, the claims used only in at least one neural network including one, the process of positioning the one floating body of the coupling load or the threshold calculated for the end position from the positioning start of the one floating body The floating body positioning system according to any one of 1 to 3. When the positioning positions of the N-th floating body positioning and the (N + 1) -th floating body positioning are within a predetermined range and the positioning date / time is within a predetermined range, the calculation means ends the N-th floating body positioning process. The floating body positioning system according to any one of claims 1 to 4, wherein the joint load or the threshold value calculated in step S3 is used as an initial value of the joint load or the threshold value when the N + 1th floating body positioning is started. A floating body positioning method using a neural network that outputs a change amount of the position or orientation from an output layer when an operation amount in an operation means for performing an operation for changing the position or orientation of the floating body is input to an input layer ,
A detection step of detecting a change amount of the position or posture of the floating body before and after the operation;
When the movement amount in the movement means is input to the input layer, the change amount detected when an arbitrary movement amount is given to the movement means to the neural network that outputs the change amount of the position or orientation from the output layer. Calculating as a teacher signal, and calculating at least one of the connection weight or threshold in the neural network by backpropagation ,
Using the neural network that includes at least one of the combined load calculated by the calculation step or the threshold value as an operation amount that causes a desired amount of change in the position or orientation of the floating body. The operation amount specifying step to be specified by
A floating body positioning method comprising: an instruction step for instructing the operation means to operate with the specified operation amount. When the change amount detected by the detection means for detecting the change amount of the position or posture of the floating body is input to the input layer before and after the operation by the operation means for performing the operation for changing the position or posture of the floating body , the operation A floating body positioning method using a neural network that outputs an operation amount in the means from an output layer,
The against the neural network, the operation quantity given to said operation means when the change amount by the detection means is detected input as a teacher signal, among the coupling load or thresholds in the neural network by backpropagation A calculation step for calculating at least one of the above;
Using the neural network that includes at least one of the combined load calculated by the calculation step or the threshold value as an operation amount that causes a desired amount of change in the position or orientation of the floating body. The operation amount specifying step to be specified by
A floating body positioning method comprising: an instruction step for instructing the operation means to operate with the specified operation amount.

Images