Unmanned ship track and speed intelligent control method based on Beidou positioning system and deep learning technology

Abstract

The invention relates to an unmanned ship track and speed intelligent control method based on a Beidou positioning system and a deep learning technology, and the method comprises the steps: building an initial environment space model for known information, dividing a planned track region into grids with the same size, simulating an obstacle through a grid method, marking an original obstacle as an infeasible region, and carrying out the calculation of the initial environment space model. The remaining grids are areas capable of track under normal conditions; a Beidou satellite positioning system (BDS) collects accurate position information, navigation speed information and the like of the unmanned ship in real time and returns the information in real time; and based on the returned information, taking the accurate position of the unmanned ship as a starting point, carrying out space search, namely emitting rays from the starting point to a target point, if the rays are blocked by an obstacle, indicating that the obstacle exists between the starting point and the target point, and then carrying out 360-degree scanning clockwise by taking the starting point as a center to obtain surrounding obstacle information, the environment space model is updated and optimized in real time; on the basis of the position of the unmanned ship and surrounding obstacle information, optimization is carried out by determining a cost function to obtain child nodes, and then the child nodes are obtained through continuous scanning and optimization in the sailing process until the unmanned ship reaches a target point; in the whole navigation process of the unmanned ship from a starting point to a target point through N child nodes, the unmanned ship updates the set navigation speed in real time according to position information and environment perception returned by a Beidou satellite positioning system (BDS); based on the real-time accurate position of the unmanned ship, the optimized track line, the updated navigation speed and the optimized environment space model, a Matlab 2018 platform is used to construct a deep learning intelligent optimization model for unmanned ship track and speed control, and model solution is carried out to realize optimal solution of control variables. The system mainly has the advantages of high real-time information precision, high sensitivity and intelligence.

Description

Technical Field

[0001] The present invention belongs to the field of intelligent control methods for unmanned ships, and particularly relates to an intelligent control method for the track and speed of an unmanned ship based on the Beidou positioning system and deep learning technology. Background Art

[0002] An unmanned ship is a fully automatic surface robot that can sail on the water surface according to a preset task by means of precise satellite positioning and its own sensors, and has important research significance and value. If an unmanned ship is used to perform relatively dangerous tasks, since the degree of operator participation is low, the risk of personnel injury can be greatly reduced. For example, when using an unmanned ship to monitor highly polluted waters, personnel can be prevented from being exposed to harmful elements. However, sensitive, robust, and real-time trajectory planning and speed setting are important guarantees for improving the success rate of tasks. However, the current maturity of unmanned ship technology is not high enough, and there are still problems such as insufficient flexibility, low reliability, and lack of real-time performance. Therefore, in order to improve the above problems, this study proposes an intelligent control method for the track and speed of an unmanned ship based on the Beidou positioning system and deep learning technology. The application of the Beidou positioning system improves the accuracy and precision of the real-time information of the unmanned ship, and the use of deep learning technology can intelligently optimize the best track and instantaneous speed. Summary of the Invention

[0003] The purpose of the present invention is to improve the problems existing in the existing intelligent control methods for unmanned ships, and to propose an intelligent control method for the track and speed of an unmanned ship based on the Beidou positioning system and deep learning technology.

[0004] In order to achieve the above purpose, the technical solution adopted by the present invention is:

[0005] An intelligent control method for the track and speed of an unmanned ship based on the Beidou positioning system and deep learning technology includes the following steps:

[0006] Step 1, establish an initial environmental space model for known information, divide the planned track area into grids of the same size, use the grid method to simulate obstacles, mark the original obstacles as infeasible areas, and the remaining grids are navigable areas under normal conditions;

[0007] Step 2, the Beidou satellite positioning system (BDS) real-time collects accurate position information and navigation speed information of the unmanned ship, etc., and returns the information in real time;

[0008] Step 3, based on the returned information, start a space search with the accurate position of the unmanned ship as the starting point, that is, emit a ray from the starting point to the target point. If the ray is blocked by an obstacle, it means there is an obstacle between the starting point and the target point. Then, start a 360° scan clockwise with the starting point as the center to obtain surrounding obstacle information, and update and optimize the environmental space model in real time;

[0009] Step 4: Based on the position of the unmanned ship and the surrounding obstacle information, optimize by determining the cost function to obtain child nodes, and then continuously scan and optimize during the navigation process to obtain child nodes until the unmanned ship reaches the target point;

[0010] Step 5: During the entire navigation process of the unmanned ship from the starting point through N child nodes to the target point, the unmanned ship updates the set navigation speed in real time according to the position information returned by the Beidou Satellite Navigation System (BDS) and environmental perception.

[0011] Step 6: Based on the real-time accurate position of the unmanned ship, the optimized navigation track, the updated navigation speed, and the optimized environmental space model, use the Matlab 2018 platform to construct a deep learning intelligent optimization model for the track and speed control of the unmanned ship, and solve the model to achieve the optimal solution of the control variables.

[0012] In the above Step 1, establish the initial environmental space model: The determination of the grid granularity affects the accuracy of the track planning. If the value of the grid granularity is too small, it will cause too much information in the environmental space; if the value of the grid granularity is too large, when there are many obstacles, it will lead to the inability to find an effective track. Therefore, it is necessary to determine the grid granularity according to the density of obstacles in the environment. When calculating the area of the obstacle, for the obstacle area that is not rectangular, expand the edge of the obstacle to form a rectangular area, and calculate the expanded part as an obstacle. Determine and calculate the area of the obstacle by rectangularizing the obstacle, and then use the ratio of the total area of the obstacle to the total area of the grid space to determine the grid granularity;

[0013] In the above Step 2, the Beidou Satellite Navigation System (BDS) has high positioning accuracy in the Asian region and can maintain high positioning accuracy even in complex environments. Therefore, it is very suitable for the track positioning of the unmanned ship and can be accurate to the meter level;

[0014] In the above Step 3, obtain the first-generation child node set through a 360° scan of the starting point. That is, judge whether there are obstacles blocking from each first-generation child node to the target point. When it is judged that there are obstacles from all first-generation child nodes to the target point, calculate the cost value of the first-generation child nodes through the cost function. The cost function is obtained from the sum of the squares of the distances from the child nodes to the starting point and the distances to the target point. From this, judge the first-generation child node with the smallest cost value, that is, the first-generation optimal child node;

[0015] In the above step 4, after determining the first-generation optimal sub-node, a ray is emitted centered on this point. After being blocked by an obstacle or the map edge, a 360° clockwise scan is started to identify the next-generation sub-node set. By calculating the cost value of each second-generation sub-node, which is obtained by summing the square of the distance from the second-generation sub-node to the first-generation optimal point and the distance to the target point, the second-generation optimal sub-node is found. Thus, sub-nodes of each generation are scanned, and the optimal sub-node of each generation is determined. When a ray is emitted from a sub-node of a certain generation towards the target point and the ray is not blocked by an obstacle, it is determined that the target end point has been scanned, and the scan ends;

[0016] In the above step 6, based on the real-time accurate position of the unmanned ship, the optimized track line, the updated navigation speed, and the optimized environmental space model, network deep learning training is carried out, exploratory data analysis is carried out, and the Matlab 2018 platform is used to construct a deep learning intelligent optimization model for the track and speed control of the unmanned ship, and the model is solved to achieve the optimal solution of the control variables.

[0017] The main advantages of the present invention are as follows:

[0018] 1. High real-time information accuracy: The Beidou satellite positioning system (BDS) has high positioning accuracy in the Asian region and can maintain high positioning accuracy even in complex environments. Therefore, it is very suitable for the track positioning of unmanned ships and can be accurate to the meter level;

[0019] 2. High sensitivity and intelligence: Based on the real-time accurate position of the unmanned ship, the optimized track line, the updated navigation speed, and the optimized environmental space model, network deep learning training is carried out, exploratory data analysis is carried out, and the Matlab 2018 platform is used to construct a deep learning intelligent optimization model for the track and speed control of the unmanned ship, and the model is solved to achieve the optimal solution of the control variables. Brief Description of the Drawings

[0020] Figure 1 The technical roadmap of the method of the present invention. Detailed Embodiments

[0021] Embodiment 1

[0022] An unmanned ship navigation experiment was carried out in the lake area of a campus in Tianjin. First, an initial environmental space model was established based on the known information. The planned track area was divided into grids of the same size, and the grid method was used to simulate obstacles. The original obstacles were marked as infeasible areas, and the remaining grids were navigable areas under normal conditions. Secondly, the Beidou Satellite Navigation System (BDS) was used to collect the accurate position information and navigation speed information of the unmanned ship in real time and return the information in real time. Then, based on the returned information, starting from the accurate position of the unmanned ship, a space search was carried out. That is, a ray was emitted from the starting point to the target point. If the ray was blocked by an obstacle, it indicated that there was an obstacle between the starting point and the target point. Then, starting from the starting point, a 360° scan was carried out clockwise to obtain the surrounding obstacle information and update and optimize the environmental space model in real time. Next, based on the position of the unmanned ship and the surrounding obstacle information, the cost function was determined to optimize and obtain child nodes. Then, during the navigation process, child nodes were obtained through continuous scanning and optimization until the unmanned ship reached the target point. After that, during the entire navigation process of the unmanned ship from the starting point through N child nodes to the target point, the unmanned ship updated the set navigation speed in real time according to the position information and environmental perception returned by the Beidou Satellite Navigation System (BDS). Finally, based on the real-time accurate position of the unmanned ship, the optimized track line, the updated navigation speed, and the optimized environmental space model, using the Matlab 2018 platform, a deep learning intelligent optimization model for the track and speed control of the unmanned ship was constructed and the model was solved to achieve the optimal solution of the control variables. The results show that the unmanned ship can avoid obstacles and update and optimize the track path and navigation speed in real time, proving the effectiveness and reliability of the method.

Claims

1. An intelligent control method for the track and speed of an unmanned ship based on Beidou positioning system and deep learning technology. It is characterized in that The steps include: Step 1: Establish an initial environment space model based on known information, divide the planned track area into grids of the same size, use the grid method to simulate obstacles, mark the original obstacles as infeasible areas, and the remaining grids are the feasible track areas under normal conditions; Step 2: The Beidou Satellite Positioning System (BDS) collects the precise location information and navigation speed information of the unmanned ship in real time and returns the information in real time; Step 3: Based on the returned information, the precise position of the unmanned boat is used as the starting point to conduct a spatial search, that is, a ray is emitted from the starting point to the target point. If the ray is blocked by an obstacle, it means that there is an obstacle between the starting point and the target point. Then, a 360° clockwise scan is performed starting from the starting point to obtain the surrounding obstacle information and update the optimized environment space model in real time. Step 4: Based on the location of the unmanned ship and the surrounding obstacle information, the cost function is determined to optimize the child nodes, and then the child nodes are obtained by continuous scanning and optimization during the navigation process until the unmanned ship reaches the target point, thereby determining the actual navigation path of the unmanned ship; Step 5: During the entire navigation process of the unmanned ship from the starting point through N subnodes to the target point, the unmanned ship updates the set navigation speed in real time according to the location information and environmental perception returned by the Beidou Satellite Positioning System (BDS); Step 6: Based on the real-time precise position of the unmanned ship, the optimized track route, the updated navigation speed, and the optimized environmental space model, use the Matlab 2018 platform to build a deep learning intelligent optimization model for the unmanned ship track and speed control, and solve the model to achieve the optimal solution of the control variables.

2. The method according to claim 1, Features: In the step 1, the initial environment space model is established: the determination of the grid granularity affects the accuracy of the track planning. If the grid granularity is too small, the amount of information in the environment space will be too large; if the grid granularity is too large, when there are many obstacles, it will lead to the inability to find a valid track. Therefore, the grid granularity needs to be determined by the density of obstacles in the environment. When calculating the obstacle area, for obstacle areas that are not rectangular, the edge of the obstacle is expanded to form a rectangular area, and the expanded part is counted as an obstacle. The area of ​​the obstacle is determined and calculated by rectangularizing the obstacle, and then the grid granularity is determined by the ratio of the total area of ​​the obstacle to the total area of ​​the grid space.

3. The method according to claim 1, Features: In the step 2, the BeiDou satellite positioning system (BDS) has a high positioning accuracy in Asia and can maintain a high positioning accuracy even in complex environments. Therefore, it is very suitable for the track positioning of unmanned ships and can be accurate to the meter level.

4. The method according to claim 1, Features: In step 3, a 360° scan from the starting point is performed to obtain a set of first-generation child nodes. That is, it is determined whether there are obstacles blocking the first-generation child nodes from the target point. When it is determined that there are obstacles from all first-generation child nodes to the target point, the cost value of the first-generation child nodes is calculated by the cost function. The cost function is derived from the square of the distance from the child node to the starting point and the square of the distance to the target point. From this, the first-generation child node with the smallest cost value is determined, that is, the first-generation optimal child node.

5. The method according to claim 1, Features: In the step 4, after the optimal child node of the first generation is determined, a ray is emitted from this point as the center. After being blocked by obstacles or the edge of the map, a 360° clockwise scan is started to determine the next generation child node set. The generation value of each second generation child node is calculated, that is, the square of the distance from the second generation child node to the optimal point of the first generation and the square of the distance to the target point, thereby finding the optimal child node of the second generation. In this way, each generation of child nodes is scanned out, and the optimal child node of each generation is determined. When a child node of a certain generation emits a ray to the target point and the ray is not blocked by an obstacle, it is determined that the scan has reached the target end point and the scan is completed.

6. The method according to claim 1, Features: In step six, based on the real-time precise position of the unmanned ship, the optimized track route, the updated navigation speed and the optimized environmental space model, network deep learning training is carried out, data exploratory analysis is carried out, and a deep learning intelligent optimization model for the unmanned ship track and speed control is constructed using the Matlab 2018 platform, and the model is solved to achieve the optimal solution of the control variables.

Images