CN115373383A - Autonomous obstacle avoidance method and device for garbage recovery unmanned boat and related equipment - Google Patents
Autonomous obstacle avoidance method and device for garbage recovery unmanned boat and related equipment Download PDFInfo
- Publication number
- CN115373383A CN115373383A CN202210831313.6A CN202210831313A CN115373383A CN 115373383 A CN115373383 A CN 115373383A CN 202210831313 A CN202210831313 A CN 202210831313A CN 115373383 A CN115373383 A CN 115373383A
- Authority
- CN
- China
- Prior art keywords
- data
- obstacle avoidance
- unmanned ship
- module
- autonomous
- 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
Links
Images
Classifications
-
- G—PHYSICS
- G05—CONTROLLING; REGULATING
- G05D—SYSTEMS FOR CONTROLLING OR REGULATING NON-ELECTRIC VARIABLES
- G05D1/00—Control of position, course or altitude of land, water, air, or space vehicles, e.g. automatic pilot
- G05D1/02—Control of position or course in two dimensions
- G05D1/0206—Control of position or course in two dimensions specially adapted to water vehicles
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01C—MEASURING DISTANCES, LEVELS OR BEARINGS; SURVEYING; NAVIGATION; GYROSCOPIC INSTRUMENTS; PHOTOGRAMMETRY OR VIDEOGRAMMETRY
- G01C21/00—Navigation; Navigational instruments not provided for in groups G01C1/00 - G01C19/00
- G01C21/005—Navigation; Navigational instruments not provided for in groups G01C1/00 - G01C19/00 with correlation of navigation data from several sources, e.g. map or contour matching
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01C—MEASURING DISTANCES, LEVELS OR BEARINGS; SURVEYING; NAVIGATION; GYROSCOPIC INSTRUMENTS; PHOTOGRAMMETRY OR VIDEOGRAMMETRY
- G01C21/00—Navigation; Navigational instruments not provided for in groups G01C1/00 - G01C19/00
- G01C21/20—Instruments for performing navigational calculations
- G01C21/203—Specially adapted for sailing ships
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01S—RADIO DIRECTION-FINDING; RADIO NAVIGATION; DETERMINING DISTANCE OR VELOCITY BY USE OF RADIO WAVES; LOCATING OR PRESENCE-DETECTING BY USE OF THE REFLECTION OR RERADIATION OF RADIO WAVES; ANALOGOUS ARRANGEMENTS USING OTHER WAVES
- G01S13/00—Systems using the reflection or reradiation of radio waves, e.g. radar systems; Analogous systems using reflection or reradiation of waves whose nature or wavelength is irrelevant or unspecified
- G01S13/86—Combinations of radar systems with non-radar systems, e.g. sonar, direction finder
- G01S13/867—Combination of radar systems with cameras
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01S—RADIO DIRECTION-FINDING; RADIO NAVIGATION; DETERMINING DISTANCE OR VELOCITY BY USE OF RADIO WAVES; LOCATING OR PRESENCE-DETECTING BY USE OF THE REFLECTION OR RERADIATION OF RADIO WAVES; ANALOGOUS ARRANGEMENTS USING OTHER WAVES
- G01S13/00—Systems using the reflection or reradiation of radio waves, e.g. radar systems; Analogous systems using reflection or reradiation of waves whose nature or wavelength is irrelevant or unspecified
- G01S13/88—Radar or analogous systems specially adapted for specific applications
- G01S13/93—Radar or analogous systems specially adapted for specific applications for anti-collision purposes
- G01S13/937—Radar or analogous systems specially adapted for specific applications for anti-collision purposes of marine craft
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01S—RADIO DIRECTION-FINDING; RADIO NAVIGATION; DETERMINING DISTANCE OR VELOCITY BY USE OF RADIO WAVES; LOCATING OR PRESENCE-DETECTING BY USE OF THE REFLECTION OR RERADIATION OF RADIO WAVES; ANALOGOUS ARRANGEMENTS USING OTHER WAVES
- G01S19/00—Satellite radio beacon positioning systems; Determining position, velocity or attitude using signals transmitted by such systems
- G01S19/38—Determining a navigation solution using signals transmitted by a satellite radio beacon positioning system
- G01S19/39—Determining a navigation solution using signals transmitted by a satellite radio beacon positioning system the satellite radio beacon positioning system transmitting time-stamped messages, e.g. GPS [Global Positioning System], GLONASS [Global Orbiting Navigation Satellite System] or GALILEO
- G01S19/40—Correcting position, velocity or attitude
- G01S19/41—Differential correction, e.g. DGPS [differential GPS]
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01S—RADIO DIRECTION-FINDING; RADIO NAVIGATION; DETERMINING DISTANCE OR VELOCITY BY USE OF RADIO WAVES; LOCATING OR PRESENCE-DETECTING BY USE OF THE REFLECTION OR RERADIATION OF RADIO WAVES; ANALOGOUS ARRANGEMENTS USING OTHER WAVES
- G01S19/00—Satellite radio beacon positioning systems; Determining position, velocity or attitude using signals transmitted by such systems
- G01S19/38—Determining a navigation solution using signals transmitted by a satellite radio beacon positioning system
- G01S19/39—Determining a navigation solution using signals transmitted by a satellite radio beacon positioning system the satellite radio beacon positioning system transmitting time-stamped messages, e.g. GPS [Global Positioning System], GLONASS [Global Orbiting Navigation Satellite System] or GALILEO
- G01S19/42—Determining position
Abstract
The invention provides an autonomous obstacle avoidance method, an autonomous obstacle avoidance device and related equipment of a garbage recycling unmanned ship, which are used for an autonomous obstacle avoidance system, wherein the autonomous obstacle avoidance system comprises a navigation positioning module, an environment sensing module, a path planning module and a driving module, and the autonomous obstacle avoidance method comprises the following steps: the navigation positioning module is used for acquiring satellite data and transmitting an observed value and site coordinate information of the satellite data to the unmanned ship through a data link; carrying out real-time carrier phase difference processing on the satellite data and the received data chain to obtain a positioning result; the environment sensing module is used for obtaining position information of a barrier and a target point position of the unmanned ship by fusing information measured by the millimeter wave radar and the monocular camera; and performing autonomous decision making through the path planning module, and driving the unmanned ship through the driving module to achieve automatic path planning and autonomous obstacle avoidance. The invention has the advantages of reasonable path planning, convenient obstacle avoidance, high accuracy and wide application range.
Description
Technical Field
The invention relates to the technical field of autonomous obstacle avoidance of unmanned boats, in particular to an autonomous obstacle avoidance method and device of a garbage recycling unmanned boat and related equipment.
Background
The unmanned surface vehicle is a novel carrier capable of executing tasks in various complex and unknown water surface environments under the condition of no human intervention, and has the advantages of small size, intellectualization, autonomy and the like. However, the existing domestic autonomous navigation unmanned ship system is not complete enough, and particularly, no significant breakthrough is made on the obstacle avoidance technology of the unmanned surface ship, so that the research on the key technology of autonomous obstacle avoidance of the unmanned surface ship has great significance for improving the autonomous intelligent level of the unmanned surface ship.
The existing unmanned ship can avoid obstacles independently, is an indispensable link for the unmanned ship to complete various water tasks, and is also a key for realizing intellectualization of the unmanned ship. The work development in the technical field of unmanned surface vehicle obstacle avoidance planning in China is late, the artificial potential field method is adopted to realize the global obstacle avoidance planning of unmanned surface vehicles, and the unmanned surface vehicles are likely to fall into the local optimal solution, so that the path planning is likely to fail. There are other ship collision avoidance algorithms such as genetic algorithm, particle swarm algorithm, ant colony algorithm, etc., but all have a problem of poor real-time performance. Because a model cannot be established in advance from sample data, a path optimization search process needs to be repeated in an application process, a large amount of useless calculation is caused, a dependent evaluation function is simple, the robustness under different water area application environments is poor, the obstacle avoidance effect is poor, and the application range is small.
Disclosure of Invention
Aiming at the defects of the related technologies, the invention provides an autonomous obstacle avoidance method, device and related equipment of the unmanned garbage recycling boat, which have the advantages of good positioning effect, reasonable path planning, convenient obstacle avoidance and high accuracy.
In order to solve the technical problem, in a first aspect, an embodiment of the present invention provides an autonomous obstacle avoidance method for a garbage collection unmanned ship, where the autonomous obstacle avoidance system is used for an autonomous obstacle avoidance system, the autonomous obstacle avoidance system includes a navigation positioning module, an environment sensing module, a path planning module, and a driving module, and the autonomous obstacle avoidance method includes the following steps:
satellite data are collected through the navigation positioning module, and the observed value and the site coordinate information of the satellite data are transmitted to the unmanned ship through a data link;
carrying out real-time carrier phase difference processing on the satellite data and the received data link to obtain a positioning result;
the environment sensing module is used for obtaining position information of a barrier and a target point position of the unmanned ship by fusing information measured by the millimeter wave radar and the monocular camera;
and performing autonomous decision making through the path planning module, and driving the unmanned ship through the driving module to achieve automatic path planning and autonomous obstacle avoidance.
Preferably, the obtaining a positioning result by performing real-time carrier phase difference processing on the satellite data and the received data link specifically includes the following sub-steps:
eliminating phase difference values in the data in an iterative process through a RANSAC algorithm;
carrying out plane fitting by utilizing a PCA method;
centimeter-level positioning results are obtained.
Preferably, the RANSAC algorithm specifically comprises the following sub-steps:
presetting a three-dimensional coordinate data set detected by an on-board sensor to be omega;
selecting a minimum sample subset meeting the conditions from the data set, and calculating the model parameters of the minimum subset as initial parameters;
calculating a difference between the Ω and the initial parameter;
comparing the difference value with a preset threshold value to obtain a comparison result, and screening out and deleting points which do not meet the conditions for judging the conditions according to the comparison result;
and repeating the processes, and continuously iterating to finally obtain a mathematical model parameter.
Preferably, the RANSAC algorithm yields a probability p that the result is useful, and the expression (1) of p is as follows:
p＝1-[1-ω n ] K …(1)；
wherein ω is the probability of data in a sample point set, n is the number of coordinate points required for model fitting once, K is the actual iteration number, and the expression (2) of K is as follows:
preprocessing by using RANSAC algorithm to obtain a point set, calculating a normal vector of a plane and distances d from all sampling points to a target plane by using a PCA method, wherein an expression (3) of d is as follows:
wherein x is i ,y i ,z i Is the coordinate of the sampling point, N is the number of the sampling points, x i ,y i ,z i The respective mean expressions (4) are as follows:
then d is obtained i The standard deviation δ, the expression (5) of the standard deviation δ is as follows:
when d is i <When the distance is 2 delta, the point is reserved, and an optimal two-dimensional fitting equation is obtained through continuous iteration; after the information of the surrounding environment of the unmanned ship is obtained and processed, the garbage on the water surface is identified by using a target detection algorithm yolov4, and then the target garbage is tracked by using a KCF algorithm.
Preferably, the obtaining of the position information of the obstacle and the unmanned ship target point location by the environment sensing module through fusing the information measured by the millimeter wave radar and the monocular camera specifically includes the following sub-steps:
acquiring environment information acquired by a millimeter wave radar and a monocular camera, wherein the environment information comprises point cloud data of the millimeter radar and image data of the monocular camera;
performing preset time synchronization processing on the point cloud data and the image data;
filtering and denoising the point cloud data and the image data respectively;
performing target detection on the image data subjected to noise reduction;
and performing projection transformation on the point cloud data and the image data, and performing data fusion positioning to obtain obstacle motion information and corresponding coordinate values in a three-dimensional space around the unmanned ship.
Preferably, the autonomous decision making by the path planning module, and the driving of the unmanned vehicle by the driving module to achieve automatic path planning and autonomous obstacle avoidance specifically include the following substeps:
establishing a preset unmanned ship operation model according to the autonomous decision;
fusing a deep learning neural network into a DPG strategy according to the unmanned ship operation model to obtain a DDPG algorithm;
according to the boat-carried data collected by the boat-carried sensors on the unmanned boat;
processing by adopting a speed obstacle method according to the ship-borne data;
obtaining a set of sample data according to the processing;
putting the sample data into an experience pool;
selecting a strategy action and making an evaluation;
and selecting a course angle according to the strategy action and the evaluation.
Preferably, the autonomous obstacle avoidance method further includes the following sub-steps:
the DDPG algorithm makes an error selection;
returning the DDPG algorithm to a last state;
the DDPG network outputs an action value and improves noise;
controlling the unmanned ship to execute corresponding actions according to the action values, and obtaining sample data;
and updating the DDPG network according to the sample data.
In a second aspect, an embodiment of the present invention further provides an autonomous obstacle avoidance device of a garbage collection unmanned ship, which is used for an autonomous obstacle avoidance system, where the autonomous obstacle avoidance system includes a navigation positioning module, an environment sensing module, a path planning module, and a driving module, and the autonomous obstacle avoidance device includes the following steps:
the acquisition unit is used for acquiring satellite data through the navigation positioning module and transmitting an observed value and site coordinate information of the satellite data to the unmanned ship through a data link;
the differential processing unit is used for carrying out real-time carrier phase differential processing on the satellite data and the received data chain to obtain a positioning result;
the position obtaining unit is used for obtaining position information of the barrier and the unmanned ship target point location by the environment sensing module through fusing information measured by the millimeter wave radar and the monocular camera;
and the driving unit is used for carrying out autonomous decision making through the path planning module and driving the unmanned ship to achieve automatic path planning and autonomous obstacle avoidance through the driving module.
In a third aspect, an embodiment of the present invention further provides a computer device, including a memory, a processor, and a computer program stored on the memory and executable on the processor, where the processor, when executing the computer program, implements the steps in the autonomous obstacle avoidance method for a garbage collection unmanned ship according to any one of the above embodiments.
In a fourth aspect, an embodiment of the present invention further provides a computer-readable storage medium, where a computer program is stored on the computer-readable storage medium, and when the computer program is executed by a processor, the steps in the autonomous obstacle avoidance method for a garbage collection unmanned ship according to any one of the foregoing embodiments are implemented.
Compared with the prior art, the navigation positioning module is used for acquiring satellite data and transmitting an observed value and site coordinate information of the satellite data to the unmanned ship through the data chain; carrying out real-time carrier phase difference processing on the satellite data and the received data link to obtain a positioning result; the environment sensing module is used for obtaining position information of a barrier and a target point position of the unmanned ship by fusing information measured by the millimeter wave radar and the monocular camera; and performing autonomous decision making through the path planning module, and driving the unmanned ship through the driving module to achieve automatic path planning and autonomous obstacle avoidance. By fusing data of the millimeter wave radar and picture information acquired by a camera, the surrounding environment can be fully sensed, accurate and high-precision target information and obstacle information are provided for the unmanned ship, and then the autonomous planning of the path is realized through a planning module; the path planning module is used for making an autonomous decision, an effective collision avoidance strategy is learned from a large amount of experience data, the stability and the accuracy of the collision avoidance strategy can still be maintained under an unknown water surface environment, a speed obstacle method is introduced to guide the training of the ddpg algorithm, and large noise is added in a failure area to improve the efficiency and the accuracy of algorithm training.
Drawings
The present invention will be described in detail below with reference to the accompanying drawings. The foregoing and other aspects of the invention will become more apparent and more readily appreciated from the following detailed description, taken in conjunction with the accompanying drawings. In the drawings:
fig. 1 is a flow chart of a method of autonomous obstacle avoidance for the unmanned garbage collection boat according to the present invention;
FIG. 2 is a flowchart illustrating a method specific to step S2 of the present invention;
FIG. 3 is a flowchart illustrating a method of step S3 according to the present invention;
FIG. 4 is a flowchart illustrating a method embodied in step S4 of the present invention;
fig. 5 is a flowchart of a method for autonomous obstacle avoidance of the unmanned garbage collection boat according to the present invention;
fig. 6 is a frame diagram of an autonomous obstacle avoidance system frame of the unmanned garbage collection boat of the present invention;
FIG. 7 is a flow chart of multi-sensor data fusion in accordance with the present invention;
FIG. 8 is a block diagram of the DDPG algorithm training aid of the speed obstacle method of the present invention;
FIG. 9 is a kinematic model of the unmanned boat of the present invention;
FIG. 10 is a schematic view of the angle between the barrier region and the unmanned surface vehicle of the present invention;
FIG. 11 is a network flow chart of the DDPG algorithm of the present invention;
FIG. 12 is a block diagram of the DDPG algorithm training aid of the velocity barrier method of the present invention;
fig. 13 is a block diagram of the autonomous obstacle avoidance apparatus of the unmanned garbage collection boat of the present invention;
FIG. 14 is a block diagram of a computer apparatus according to the present invention.
Detailed Description
The following detailed description of embodiments of the invention refers to the accompanying drawings.
The embodiments/examples described herein are specific embodiments of the present invention, are intended to be illustrative of the concepts of the present invention, are intended to be illustrative and exemplary, and should not be construed as limiting the embodiments and scope of the invention. In addition to the embodiments described herein, those skilled in the art will be able to employ other technical solutions which are obvious based on the disclosure of the claims and the specification of the present application, and these technical solutions include those which make any obvious replacement or modification of the embodiments described herein, and all of which are within the scope of the present invention.
Referring to fig. 1 to 8, fig. 1 is a flowchart illustrating a method of autonomous obstacle avoidance for a unmanned boat for garbage collection according to the present invention; FIG. 2 is a flowchart illustrating a method specific to step S2 of the present invention; FIG. 3 is a flowchart illustrating a method specific to step S3 of the present invention; FIG. 4 is a flowchart illustrating a method specific to step S4 of the present invention; fig. 5 is a flowchart of a method for autonomous obstacle avoidance of the unmanned garbage collection boat according to the present invention; fig. 6 is a frame diagram of an autonomous obstacle avoidance system frame of the unmanned garbage collection boat of the present invention; FIG. 7 is a flow chart of multi-sensor data fusion in accordance with the present invention; FIG. 8 is a block diagram of the DDPG algorithm training aid by the speed obstacle method of the present invention.
Example one
The invention provides an autonomous obstacle avoidance method of a garbage recycling unmanned ship, which is used for an autonomous obstacle avoidance system, wherein the autonomous obstacle avoidance system consists of an onboard industrial personal computer and a shore-end upper computer, the onboard industrial personal computer and the shore-end upper computer are in communication connection through a communication system, and the shore-end upper computer is used for inputting control quantity and measuring parameters in a mathematical direction, wherein the onboard industrial personal computer consists of four modules which are respectively a navigation positioning module, an environment sensing module, a path planning module and a driving module.
The autonomous obstacle avoidance method comprises the following steps:
s1, satellite data are collected through the navigation positioning module, and observed values and station coordinate information of the satellite data are transmitted to the unmanned ship through a data link.
And S2, carrying out real-time carrier phase difference processing on the satellite data and the received data chain to obtain a positioning result.
And S3, the environment sensing module obtains position information of the barrier and the unmanned ship target point location by fusing information measured by the millimeter wave radar and the monocular camera.
And S4, performing autonomous decision making through the path planning module, and driving the unmanned ship through the driving module to achieve automatic path planning and autonomous obstacle avoidance.
Specifically, through the method of the S1-S4, the navigation positioning module collects satellite data through RTK, and transmits the observed value and the station coordinate information to the unmanned ship through a data chain, and then performs real-time carrier phase difference processing on the collected satellite data and the received data chain to obtain a centimeter-level positioning result. The environment sensing module determines the positions of the target points of the barrier and the unmanned ship by fusing information measured by the millimeter wave radar and the monocular camera, and then the unmanned ship makes an autonomous decision by the path planning module, so that the unmanned ship has the functions of automatically planning a path and autonomously avoiding the barrier. By fusing data of the millimeter wave radar and picture information acquired by a camera, the surrounding environment can be fully sensed, accurate and high-precision target information and obstacle information are provided for the unmanned ship, and then the autonomous planning of the path is realized through a planning module; the path planning module is used for conducting autonomous decision making, an effective collision avoidance strategy is learned from a large amount of experience data, the stability and the accuracy of the collision avoidance strategy can still be maintained under an unknown water surface environment, meanwhile, a speed obstacle method is introduced to conduct training of the ddpg algorithm, large noise is added in a failure area to improve the efficiency and the accuracy of algorithm training, and the path planning is reasonable, the obstacle avoidance is convenient and the accuracy is high.
In this embodiment, step S2 specifically includes the following sub-steps:
and S21, eliminating phase difference values in the data in an iterative process through a RANSAC algorithm.
And S22, carrying out plane fitting by using a PCA method.
And S23, obtaining a centimeter-level positioning result.
The RANSAC algorithm is a method for estimating a mathematical model from a set of observed data including outliers in an iterative manner. The RANSAC algorithm fuses the idea of screening and deleting the unqualified data, so that the identification result can be more accurately obtained for the data sample with partial error data in many environments.
Specifically, coarse difference values in data are eliminated in an iterative process through a RANSAC algorithm, so that abnormal values in measured data can be greatly reduced, and a PCA method is used for carrying out plane fitting to replace a common least square method for fitting a plane. And obtaining a centimeter-level positioning result. Although the least square method can reduce errors in observation vectors, errors in coefficient matrixes are ignored, so that the accuracy of the fitted two-dimensional plane is poor, the following path planning step is influenced, and even the navigation safety of the unmanned ship is influenced.
Further, the RANSAC algorithm specifically comprises the following sub-steps:
presetting a three-dimensional coordinate data set detected by an on-board sensor to be omega; selecting a minimum sample subset meeting the conditions from the data set, and calculating a model parameter of the minimum subset as an initial parameter; calculating a difference between the Ω and the initial parameter; comparing the difference value with a preset threshold value to obtain a comparison result, screening out points which do not meet the conditions for judgment according to the comparison result, and deleting the points; and repeating the processes, and continuously iterating to finally obtain a mathematical model parameter.
Specifically, assuming that a three-dimensional coordinate data set detected by an on-board sensor is omega, a minimum sample subset meeting conditions is selected from the three-dimensional coordinate data set, model parameters of the minimum subset are calculated to serve as initial parameters, then, a difference value between the omega and the initial parameters is calculated, the difference value is compared with a set threshold value, and points which do not meet the conditions are screened out and deleted under the judgment conditions. And repeating the processes, continuously iterating, and finally estimating an optimal mathematical model parameter. The method has the advantages of good detection effect, convenience in estimating the optimal mathematical model parameters and high accuracy.
Further, the RANSAC algorithm yields a probability p that the result is useful, where p has the following expression (1):
p＝1-[1-ω n ] K …(1)；
where ω is the probability of the data in the sample point set, not known a priori, but can be given a robust value below the tree. n is the number of coordinate points required by one-time model fitting, K is the actual iteration number, and an expression (2) of K is as follows:
preprocessing by using RANSAC algorithm to obtain a point set, calculating a normal vector of a plane and distances d from all sampling points to a target plane by using a PCA method, wherein an expression (3) of d is as follows:
wherein x i ,y i ,z i Is the coordinate of the sampling point, N is the number of sampling points, x i ,y i ,z i The respective mean expressions (4) are as follows:
then, d is obtained i The standard deviation δ, the expression (5) of the standard deviation δ is as follows:
when d is i <When the distance is 2 delta, the point is reserved, and an optimal two-dimensional fitting equation is obtained through continuous iteration; after the information of the surrounding environment of the unmanned ship is obtained and processed, garbage on the water surface is identified by using a target detection algorithm yolov4, and then the target garbage is tracked by using a KCF algorithm.
The KCF algorithm is called Kernelized Correlation Filters, and has a very bright appearance in both tracking effect and tracking speed. The algorithm mainly uses a cyclic matrix to collect samples, and uses fast Fourier transform to perform accelerated calculation on the algorithm.
Specifically, through the expressions (1) - (5), the efficiency of identifying the water surface garbage can be improved, the unmanned boat can be conveniently controlled to track the target garbage in real time, the garbage identification precision is high, and the tracking effect is good.
In this embodiment, the step S3 specifically includes the following sub-steps:
s31, obtaining environment information obtained by the millimeter wave radar and the monocular camera, wherein the environment information comprises point cloud data of the millimeter radar and image data of the monocular camera.
And S32, carrying out preset time synchronization processing on the point cloud data and the image data.
And S33, respectively carrying out filtering and denoising treatment on the point cloud data and the image data.
And S34, carrying out target detection on the image data subjected to noise reduction.
And S35, performing projection transformation on the point cloud data and the image data, and performing data fusion positioning to obtain barrier motion information and corresponding coordinate values in a three-dimensional space around the unmanned ship.
By the method of the steps S31 to S35, path planning of the unmanned ship is not facilitated in the three-dimensional coordinate system, and barrier information in the three-dimensional space needs to be converted to a two-dimensional plane, so that the purposes of simplifying calculated amount and improving the real-time rate of path planning are achieved. By fusing data of the millimeter wave radar and picture information acquired by the camera, the surrounding environment can be fully sensed, accurate and high-precision target information and obstacle information are provided for the unmanned ship, and then the autonomous planning of the path is realized through the planning module. The problem of the single sensor acquisition information is not enough and leads to the unreasonable route planning is solved. The PCA method is used for replacing the least square method to fit the plane, and the data detected by the sensor are preprocessed through the Ransc method, so that the precision of the fitted two-dimensional plane can be improved, and the error is reduced.
In this embodiment, the step S4 specifically includes the following sub-steps:
and S41, establishing a preset unmanned ship operation model according to the autonomous decision.
Specifically, for convenience of analysis, the unmanned ship is regarded as a mass point, and the angular velocity of the heading angle is used for controlling the motion process of the unmanned ship, as shown in fig. 9.
The equation of motion for an unmanned boat can be expressed as:
wherein v is u Is the speed of the unmanned vehicle in a two-dimensional plane, α is the heading angle of the unmanned vehicle, and ω is the angular velocity of the unmanned vehicle.
In the process of water surface navigation, the course angle and the angular speed of the unmanned boat meet the following performance constraint conditions:
and S42, fusing a deep learning neural network into a DPG strategy according to the unmanned ship operation model to obtain a DDPG algorithm.
Specifically, the DDPG algorithm is a strategy learning method that merges a deep learning neural network into a DPG. In the algorithm, the unmanned ship learns the optimal path selection under different water surface conditions in a trial and error mode. The process of learning by the method is long in time consumption and low in efficiency. Therefore, the design optimizes the obstacle avoidance selection of the unmanned ship through a speed obstacle method so as to improve the learning efficiency.
Assume that the obstacle detected by the sensor is circular with a radius r o Radius of unmanned surface vehicle is r s Then the obstacle circle can be expanded to R = R s +r o Simplifying the unmanned boat into a mass point; as shown in fig. 10.
And S43, according to the boat-borne data acquired by the boat-borne sensors on the unmanned boat.
And S44, processing the data by adopting a speed obstacle method according to the ship-borne data.
And S45, acquiring a group of sample data according to the processing.
And S46, putting the sample data into an experience pool.
S47, selecting strategy action and making evaluation.
And S48, selecting a course angle according to the strategy action and the evaluation.
Specifically, the relative speed vuoi between the unmanned surface vehicle and the obstacle and the included angle alpha between the tangent line of the mass point corresponding to the unmanned surface vehicle and the expanded obstacle circle and the included angle alpha between the mass point and the center Oi of the obstacle can be obtained through the millimeter wave radar and the camera on the unmanned surface vehicle oi . The included angle between the relative velocity vector and the relative position of the unmanned boat and the barrier is alpha i 。
When alpha is i ≥α oi When the unmanned ship sails along the current sailing direction, the obstacle does not influence the unmanned ship; when alpha is i ≤α oi When the unmanned ship sails along the current sailing direction, the unmanned ship may collide with obstacles, and the sailing safety of the unmanned ship is affected. In the method of speed obstacle, according to α i And alpha oi The magnitude relation of the angle of the course angle is used for making action on whether the unmanned ship needs to execute obstacle avoidance action and the magnitude of the course angle required to be adjusted for obstacle avoidance.
The DDPG algorithm is a combination of the Actor-Critic algorithm and the DQN algorithm, is different from the DQN algorithm only in discrete space, and acts in continuous space, so that continuous adjustment quantity can be output through a neural network, the course angle of the unmanned ship can be properly modified, and the purpose of avoiding obstacles is achieved. The DDPG network structure consists of a real actor network, a target actor network, a real critic network and a target critic network. an operator network executes strategy actions, wherein the network weight parameter is theta, the input state is St, and the output action is at; the critic network gives a score Q of making an action, so that the Q value when an optimal selection is made is the maximum, wherein the network weight parameter is omega, the state st and the action at are input, and the output is an evaluation value Q.
The actor network update adopts a gradient descent method as shown in the following formula:
and m is the sampling number of the sample data. The critic network performs parameter updating through a mean square error loss function.
Gamma is the reward discount factor. As shown in fig. 11.
The state S is obtained by an on-board sensor, the state S is input into a real operator to obtain an action a, the action a is applied to the unmanned ship, the unmanned ship interacts with the environment and returns to the state S ' and the reward r ' at the next moment, and therefore a group of sample data (S, a, r, S ') is obtained and is placed into an experience pool. The values of actual Q (S, a) are obtained by inputting S and a in (S, a, r, S') to actual Critic, and let Q = Q (S, a). Then, S ' in (S, a, r, S ') is input to the target Actor, resulting in action a '. And inputting S ' and a ' together into the target Critic to obtain Q (S ', a ') so that the target Q value is Q ' = r + γ × Q (S ', a ') value. And then, the Q 'is regarded as a label, the real critic is updated to enable the output Q to be close to the label Q' as much as possible, the real Actor is updated, because the action output by the real Actor, the Q value is given in the real critic, the real Actor is updated to enable the Q value to be output to the maximum, and sampling is updated circularly so as to achieve the operation close to the optimal selection.
The frame of training the DDPG algorithm assisted by the velocity barrier method is shown in FIG. 12. When the DDPG algorithm carries out obstacle avoidance training, gaussian white noise is added when the target operator network outputs action, and the action of the determined value executed by the network is changed into random value action, so that the exploration capability of the DDPG algorithm is improved.
a t ’＝a t +EN…(10)；
In the formula: a is a t Outputting the action for the original operator network; EN is a random exploration factor which accords with Gaussian distribution; a is t ' is an output action with random exploration capability after the exploration factor is added. The exploratory noise employed is as follows:
EN～N(μ＝0，δ＝0.5)…(11)。
preferably, the autonomous obstacle avoidance method further comprises the following sub-steps:
and S5, making error selection by the DDPG algorithm.
And S6, returning the DDPG algorithm to the last state.
And S7, outputting the action value and improving the noise by the DDPG network.
And S8, controlling the unmanned ship to execute corresponding actions according to the action values, and obtaining sample data.
And S9, updating the DDPG network according to the sample data.
Specifically, in many experiments, the DDPG algorithm repeatedly triggers a training failure condition in an exploration failure area, and learning efficiency is poor. The design improves exploration randomness in a failure area so as to improve learning efficiency of the algorithm. The improvement method is shown in the following formula:
EN～0.5(N(-1，0.5)+N(1，0.5))…(12)；
compared with the original noise in the algorithm, the search amplitude is shifted from the interval between [ -0.5,0.5] to the interval between [ -1.5,0.5] and [0.5,1.5] to two sides in the failure area, so that the random search amplitude is increased, and the speed for finding the correct navigation direction is increased. And after jumping out of the failed area, repeatedly training the failed area for many times to accumulate a large amount of sample data and improve the accuracy of the course angle selection of the algorithm.
Specifically, by utilizing the multi-dimensional feature extraction capability of the DDPG algorithm, an effective collision avoidance strategy is learned from a large amount of empirical data, the stability and the accuracy of the collision avoidance strategy can still be maintained under an unknown water surface environment, meanwhile, a speed obstacle method is introduced to guide the training of the DDPG algorithm, and large noise is added in a failure area to improve the efficiency and the accuracy of algorithm training, so that the problems of low convergence speed and low sample data utilization rate of the DDPG algorithm training are effectively solved.
Example two
Referring to fig. 13, fig. 13 is a block diagram of an autonomous obstacle avoidance apparatus of the unmanned garbage collection boat according to the present invention. The embodiment of the invention also provides an autonomous obstacle avoidance device 200 of the garbage recycling unmanned ship, which is used for an autonomous obstacle avoidance system, wherein the autonomous obstacle avoidance system comprises a navigation positioning module, an environment sensing module, a path planning module and a driving module, and the autonomous obstacle avoidance device 200 comprises the following steps:
the acquisition unit 201 is used for acquiring satellite data through the navigation positioning module and transmitting an observed value and site coordinate information of the satellite data to the unmanned ship through a data chain;
the difference processing unit 202 is configured to perform real-time carrier phase difference processing on the satellite data and the received data link to obtain a positioning result;
the position obtaining unit 203 is used for the environment sensing module to obtain position information of the obstacle and the target point position of the unmanned ship by fusing information measured by the millimeter wave radar and the monocular camera;
and the driving unit 204 is used for making an autonomous decision through the path planning module and driving the unmanned ship to achieve automatic path planning and autonomous obstacle avoidance through the driving module.
Specifically, the acquisition unit 201 is used for acquiring satellite data through the navigation positioning module, and transmitting an observed value and site coordinate information of the satellite data to the unmanned ship through a data chain; the difference processing unit 202 is configured to perform real-time carrier phase difference processing on the satellite data and the received data link to obtain a positioning result; the position obtaining unit 203 is used for the environment sensing module to obtain position information of the obstacle and the target point position of the unmanned ship by fusing information measured by the millimeter wave radar and the monocular camera; the driving unit 204 is configured to perform an autonomous decision through the path planning module, and drive the unmanned vehicle through the driving module to achieve automatic path planning and autonomous obstacle avoidance. By fusing data of the millimeter wave radar and picture information acquired by a camera, the surrounding environment can be fully sensed, accurate and high-precision target information and obstacle information are provided for the unmanned ship, and then the autonomous planning of the path is realized through a planning module; the path planning module is used for conducting autonomous decision making, an effective collision avoidance strategy is learned from a large amount of experience data, the stability and the accuracy of the collision avoidance strategy can still be maintained under an unknown water surface environment, meanwhile, a speed obstacle method is introduced to conduct training of the ddpg algorithm, large noise is added in a failure area to improve the efficiency and the accuracy of algorithm training, and the path planning is reasonable, the obstacle avoidance is convenient and the accuracy is high.
EXAMPLE III
Referring to fig. 14, fig. 14 is a block diagram of a computer device according to the present invention. The embodiment of the present invention further provides a computer device, which includes a memory 301, a processor 302, and a computer program stored in the memory 301 and capable of running on the processor 302, where the processor 302 implements the steps in the autonomous obstacle avoidance method for a garbage collection unmanned ship according to the first embodiment when executing the computer program.
Example four
The embodiment of the invention also provides a computer-readable storage medium, where a computer program is stored on the computer-readable storage medium, and when the computer program is executed by a processor, the steps in the autonomous obstacle avoidance method for the unmanned garbage collection boat of the first embodiment are implemented.
The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention, and various modifications and changes may be made by those skilled in the art. Any tampering, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention should be included in the scope of the claims of the present invention.
Claims (10)
1. An autonomous obstacle avoidance method of a garbage recycling unmanned ship is used for an autonomous obstacle avoidance system, the autonomous obstacle avoidance system comprises a navigation positioning module, an environment sensing module, a path planning module and a driving module, and the autonomous obstacle avoidance method is characterized by comprising the following steps:
satellite data are collected through the navigation positioning module, and the observed value and the site coordinate information of the satellite data are transmitted to the unmanned ship through a data link;
carrying out real-time carrier phase difference processing on the satellite data and the received data link to obtain a positioning result;
the environment sensing module is used for obtaining position information of a barrier and a target point position of the unmanned ship by fusing information measured by the millimeter wave radar and the monocular camera;
and performing autonomous decision making through the path planning module, and driving the unmanned ship through the driving module to achieve automatic path planning and autonomous obstacle avoidance.
2. The autonomous obstacle avoidance method of a unmanned ship for garbage collection according to claim 1, wherein the obtaining of the positioning result by performing real-time carrier phase difference processing on the satellite data and the received data link specifically comprises the following substeps:
eliminating phase difference values in the data in an iterative process through a RANSAC algorithm;
carrying out plane fitting by utilizing a PCA method;
centimeter-level positioning results are obtained.
3. The autonomous obstacle avoidance method of the unmanned garbage collection boat as claimed in claim 2, wherein the RANSAC algorithm specifically includes the following substeps:
presetting a three-dimensional coordinate data set detected by an on-board sensor to be omega;
selecting a minimum sample subset meeting the conditions from the data set, and calculating a model parameter of the minimum subset as an initial parameter;
calculating a difference between the Ω and the initial parameter;
comparing the difference value with a preset threshold value to obtain a comparison result, screening out points which do not meet the conditions for judgment according to the comparison result, and deleting the points;
and repeating the processes, and continuously iterating to finally obtain a mathematical model parameter.
4. The autonomous obstacle avoidance method of the unmanned garbage collection boat as claimed in claim 3, wherein the RANSAC algorithm is used to obtain a result with a useful probability p, and the expression (1) of p is as follows:
p＝1-[1-ω n ] K …(1)；
wherein ω is the probability of data in the sample point set, n is the number of coordinate points required for model fitting once, K is the actual iteration number, and the expression (2) of K is as follows:
preprocessing by using RANSAC algorithm to obtain a point set, calculating a normal vector of a plane and distances d from all sampling points to a target plane by using a PCA method, wherein an expression (3) of d is as follows:
wherein x i ,y i ,z i Is the coordinate of the sampling point, N is the number of the sampling points, x i ,y i ,z i The respective mean expressions (4) are as follows:
then d is obtained i The standard deviation δ, the expression (5) of the standard deviation δ is as follows:
when d is i <When the distance is 2 delta, the point is reserved, and an optimal two-dimensional fitting equation is obtained through continuous iteration; after the information of the surrounding environment of the unmanned ship is obtained and processed, the garbage on the water surface is identified by using a target detection algorithm yolov4, and then the target garbage is tracked by using a KCF algorithm.
5. The autonomous obstacle avoidance method of the garbage collection unmanned ship according to claim 1, wherein the environment sensing module obtains the position information of the obstacle and the target point of the unmanned ship by fusing the information measured by the millimeter wave radar and the monocular camera, and specifically comprises the following substeps:
acquiring environment information acquired by a millimeter wave radar and a monocular camera, wherein the environment information comprises point cloud data of the millimeter radar and image data of the monocular camera;
performing preset time synchronization processing on the point cloud data and the image data;
filtering and denoising the point cloud data and the image data respectively;
performing target detection on the image data subjected to noise reduction;
and performing projection transformation on the point cloud data and the image data, and performing data fusion positioning to obtain the barrier motion information and the corresponding coordinate value in the three-dimensional space around the unmanned ship.
6. The autonomous obstacle avoidance method of the unmanned garbage collection boat according to claim 1, wherein the autonomous decision making by the path planning module, and the driving of the unmanned boat by the driving module to achieve the automatic path planning and autonomous obstacle avoidance specifically includes the following sub-steps:
establishing a preset unmanned ship operation model according to the autonomous decision;
fusing a deep learning neural network into a DPG strategy according to the unmanned ship operation model to obtain a DDPG algorithm;
according to the on-board data collected by the on-board sensor on the unmanned ship;
processing by adopting a speed obstacle method according to the ship-borne data;
obtaining a set of sample data according to the processing;
putting the sample data into an experience pool;
selecting a policy action and making an evaluation;
and selecting a course angle according to the strategy action and the evaluation.
7. The autonomous obstacle avoidance method of a unmanned ship for garbage collection according to claim 6, further comprising the substeps of:
the DDPG algorithm makes an error selection;
returning the DDPG algorithm to a last state;
the DDPG network outputs an action value and improves noise;
controlling the unmanned ship to execute corresponding actions according to the action values, and obtaining sample data;
and updating the DDPG network according to the sample data.
8. The utility model provides an autonomic obstacle-avoiding device of unmanned ship is retrieved to rubbish for autonomic obstacle-avoiding system, autonomic obstacle-avoiding system includes navigation orientation module, environmental perception module, route planning module and drive module, its characterized in that, autonomic obstacle-avoiding device includes following step:
the acquisition unit is used for acquiring satellite data through the navigation positioning module and transmitting an observed value and site coordinate information of the satellite data to the unmanned ship through a data link;
the differential processing unit is used for carrying out real-time carrier phase differential processing on the satellite data and the received data chain to obtain a positioning result;
the position obtaining unit is used for obtaining position information of the barrier and the unmanned ship target point location by the environment sensing module through fusing information measured by the millimeter wave radar and the monocular camera;
and the driving unit is used for carrying out autonomous decision making through the path planning module and driving the unmanned ship to achieve automatic path planning and autonomous obstacle avoidance through the driving module.
9. A computer arrangement comprising a memory, a processor and a computer program stored on the memory and executable on the processor, the processor implementing the steps in the autonomous obstacle avoidance method of a garbage collection unmanned craft according to any of claims 1 to 7 when executing the computer program.
10. A computer-readable storage medium, characterized in that the computer-readable storage medium has stored thereon a computer program which, when being executed by a processor, implements the steps in the autonomous obstacle avoidance method of a garbage collection unmanned boat according to any one of claims 1 to 7.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210831313.6A CN115373383A (en) | 2022-07-15 | 2022-07-15 | Autonomous obstacle avoidance method and device for garbage recovery unmanned boat and related equipment |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210831313.6A CN115373383A (en) | 2022-07-15 | 2022-07-15 | Autonomous obstacle avoidance method and device for garbage recovery unmanned boat and related equipment |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN115373383A true CN115373383A (en) | 2022-11-22 |
Family
ID=84061003
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202210831313.6A Pending CN115373383A (en) | 2022-07-15 | 2022-07-15 | Autonomous obstacle avoidance method and device for garbage recovery unmanned boat and related equipment |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN115373383A (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN116243720A (en) * | 2023-04-25 | 2023-06-09 | 广东工业大学 | AUV underwater object searching method and system based on 5G networking |
-
2022
- 2022-07-15 CN CN202210831313.6A patent/CN115373383A/en active Pending
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN116243720A (en) * | 2023-04-25 | 2023-06-09 | 广东工业大学 | AUV underwater object searching method and system based on 5G networking |
| CN116243720B (en) * | 2023-04-25 | 2023-08-22 | 广东工业大学 | AUV underwater object searching method and system based on 5G networking |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN111337941B (en) | Dynamic obstacle tracking method based on sparse laser radar data | |
| CN114384920B (en) | Dynamic obstacle avoidance method based on real-time construction of local grid map | |
| CN107065890B (en) | Intelligent obstacle avoidance method and system for unmanned vehicle | |
| Kim et al. | Traversability classification using unsupervised on-line visual learning for outdoor robot navigation | |
| CN109186610B (en) | Robust BSLAM method for AUV terrain matching navigation | |
| Jacobson et al. | Semi-supervised slam: Leveraging low-cost sensors on underground autonomous vehicles for position tracking | |
| CN114998276B (en) | Robot dynamic obstacle real-time detection method based on three-dimensional point cloud | |
| Mseddi et al. | YOLOv5 based visual localization for autonomous vehicles | |
| CN117589167A (en) | Unmanned aerial vehicle routing inspection route planning method based on three-dimensional point cloud model | |
| CN115373383A (en) | Autonomous obstacle avoidance method and device for garbage recovery unmanned boat and related equipment | |
| Zhou et al. | Terrain traversability mapping based on lidar and camera fusion | |
| Wang et al. | 3D-LIDAR based branch estimation and intersection location for autonomous vehicles | |
| CN109459046B (en) | Positioning and navigation method of suspension type underwater autonomous vehicle | |
| Skaldebø et al. | Dynamic positioning of an underwater vehicle using monocular vision-based object detection with machine learning | |
| Matsuo et al. | Outdoor visual localization with a hand-drawn line drawing map using fastslam with pso-based mapping | |
| CN116385493A (en) | Multi-moving-object detection and track prediction method in field environment | |
| CN113486871B (en) | Unmanned vehicle local autonomous control method, device and equipment based on depth map | |
| CN113721615B (en) | Navigation path planning method and system based on machine vision | |
| CN112817318B (en) | Multi-unmanned-boat collaborative search control method and system | |
| CN113554705A (en) | Robust positioning method for laser radar in changing scene | |
| CN113917930A (en) | Unmanned ship navigation state control method based on sensing data | |
| Yildiz et al. | CNN based sensor fusion method for real-time autonomous robotics systems | |
| Du et al. | Hierarchical path planning and obstacle avoidance control for unmanned surface vehicle | |
| CN112319468A (en) | Driverless lane keeping method for maintaining road shoulder distance | |
| Wang et al. | Detection and Tracking Dynamic Vehicles for Autonomous Driving Based on 2-D Point Scans |
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 |