Unmanned cleaning machine automatic operation system

Abstract

The present invention discloses an automatic operation system for an unmanned tank cleaning machine, comprising: a space recognition module adopts a regular grid data model to process data returned by a radar, combines the three-dimensional coordinate system of the tank cleaning machine body with the three-dimensional material pile contour and the cabin bulkhead coordinate system scanned by a laser radar, and establishes a three-dimensional material contour based on a DEM in the unified coordinate system; an intelligent control module obtains an optimal working path based on a global path planning algorithm or a local trajectory planning algorithm for the recognition result of the space recognition module, and the intelligent control module controls the tank cleaning machine to automatically operate according to the optimal working path; the space recognition module of the present application can adopt an effective data processing algorithm to reduce interference to a minimum, reduce the influence of particles with a small size, depict the contours of the real cabin and the material pile behind the dust, rain or fog curtain, effectively filter interference items such as dust, and then the intelligent control module calculates the working path for automatic operation.

Description

Technical Field

[0001] The present invention relates to the technical field of tank cleaning machines, and in particular to an automatic operation system of an unmanned tank cleaning machine. Background Art

[0002] A tank cleaning machine is a special machine that collects the remaining cargo that cannot be grabbed by grab buckets at the bottom of the cabin and in the dead corners of the cabin to the bottom of the hatch for further unloading operations when dry bulk cargo is unloaded using tools such as grab unloaders. At present, the cleaning of the cabin bottom is still completed by manually driving a tank cleaning machine. The operator needs to go deep into the bottom of the cabin and control the tank cleaning machine to pile the materials around the cabin to the center of the hatch, and then the ship unloader grabs and unloads them. Since the cabin is a semi-enclosed structure with poor ventilation and air permeability, traditional manual operation poses a high risk to the personal safety of workers. In addition, there is no one in the cabin to assist the driver in cleaning operations, resulting in low efficiency in cabin cleaning. Summary of the invention

[0003] According to an embodiment of the present invention, an automatic operation system of an unmanned cargo clearing machine is provided, comprising:

[0004] The spatial recognition module uses a regular grid data model to process the data returned by the radar, combining the 3D coordinate system of the cleaning machine body with the 3D material pile contour and the cabin bulkhead coordinate system scanned by the laser radar, and establishing the 3D material contour based on DEM in the unified coordinate system;

[0005] The intelligent control module obtains the optimal working path based on the recognition result of the space recognition module and the global path planning algorithm or the local trajectory planning algorithm, and the intelligent control module controls the cleaning machine to automatically operate according to the optimal working path;

[0006] The remote control module can display three-dimensional contour graphics, so that the driver can control the tank cleaning machine based on the remote control module.

[0007] Furthermore, a whole machine control module is arranged on the chassis of the tank cleaning machine, and a proportional valve is controlled by a high-precision proportional solenoid valve control board to realize smooth electrification control of the bottom driving system of the tank cleaning machine;

[0008] The whole machine control module aggregates the whole machine's electrical feedback signals, realizes real-time monitoring of the whole machine's electrical and instrumentation systems, receives data access from the intelligent control module, and performs unmanned automatic operation functions.

[0009] Furthermore, the space recognition module includes: integrated modeling 3D laser radar, video return camera, RTK mobile station, inclinometer, 360-degree image surround system, wireless bridge, and 5G communication;

[0010] The integrated modeling 3D lidar generates 3D reconstruction of the cabin and material distribution, the video feedback camera is used to shoot operation videos and transmit the videos remotely, the RTK mobile station performs high-precision coordinate positioning, the inclinometer is used to measure the horizontal angle of the tank cleaning machine, the 360-degree image surround system is used to shoot 360-degree images of the tank cleaning machine, and the wireless bridge and 5G communication feed back the data information of the tank cleaning machine to the remote control module.

[0011] Furthermore, image recognition as lidar point cloud data recognition, including bulkhead recognition and stockpile recognition, all utilize deep learning network structures to train the features of various detection models and achieve high-accuracy environmental condition recognition.

[0012] Furthermore, the image recognition interprets the feature data contained in the image by establishing and simulating the nervous system of the human brain, thereby fully mining the deep-level features of the target to be detected in the image, and using this data feature to better characterize the classification of the target.

[0013] Further, the remote control module includes: a remote control room, a wireless bridge and 5G communication;

[0014] The wireless bridge and 5G communication are used to receive data information fed back by the space identification module. The remote control room is used to display the first-person real-time video of the tank cleaning machine, the 3D model of the cabin, the operation trajectory display and extension, and the 360-degree panoramic image, to achieve tactile vibration feedback, restore the on-site picture and sound environment, so that the operator can also feel the real environment of the operation site in the remote operation room.

[0015] Furthermore, the intelligent control module selects the optimal working path based on the indicators of completeness, optimality, real-time performance, adaptability to environmental changes, and meeting the kinematic and dynamic requirements of the unmanned cabin cleaning machine.

[0016] Furthermore, the completeness means that the path planning algorithm can obtain effective results within a limited time, and the optimality means finding a collision-free path from the starting point to the end point according to a certain optimal standard.

[0017] Furthermore, the global path planning algorithm includes: a graph search-based method, a sampling-based method, and a curve interpolation method;

[0018] The graph search-based method is divided into two steps: establishing an environment map and path search. The environment map can be expressed by a metric map representation or a topological map representation, and the metric map representation can be expressed by a geometric representation or a spatial decomposition method.

[0019] The sampling-based methods include RRT (Random Rapid Search Tree) algorithm, PRM (Probabilistic Road Map) algorithm and State Grid Planner algorithm;

[0020] The curve interpolation method includes a Dubins curve or Reeds-Shepp curve fitting algorithm, a Clothoid curve fitting algorithm, a Bezier curve fitting algorithm, and a polynomial or spline curve fitting algorithm.

[0021] Furthermore, the local trajectory planning algorithm includes an artificial potential field method and a dynamic window algorithm;

[0022] In the artificial potential field method, when the unmanned cabin cleaning machine is moving, the target point generates gravitational force on it, and the obstacle generates repulsive force on it, and the movement speed and direction of the unmanned cabin cleaning machine are determined by the resultant force.

[0023] The dynamic window algorithm collects multiple groups of speeds in the speed space, and then predicts the trajectory of each group of speeds within a period of time, and then evaluates each group of trajectories through an evaluation function, and selects the speed corresponding to the trajectory with the minimum cost and sends it to the unmanned cabin cleaning machine.

[0024] According to an automatic operation system of an unmanned cargo tank cleaning machine according to an embodiment of the present invention, the spatial recognition module of the present application combines the three-dimensional coordinate system of the cargo tank cleaning machine body with the three-dimensional material pile contour and the cabin bulkhead coordinate system scanned by the laser radar, and establishes a three-dimensional material contour based on the DEM. In the cabin, the dust concentration is high, the coal is very humid, and even in harsh environments such as rainy and foggy weather, dust particles, fog droplets, water droplets, etc. will cause great interference to the measurement results of the laser radar. The spatial recognition module can adopt an effective data processing algorithm to reduce the interference to a minimum, reduce the influence of ignoring tiny particles, depict the contours of the real cabin and the material pile behind the dust, rain or fog curtain, and effectively filter interference items such as dust.

[0025] The intelligent control module uses the global path planning algorithm or the local trajectory planning algorithm to calculate the working path based on the recognition results of the space recognition module to obtain the optimal working path. The intelligent control module controls the tank cleaning machine to automatically operate according to the optimal working path to achieve unmanned driving. The remote control module makes it easy for workers to observe the working status of the unmanned tank cleaning machine and can perform remote driving operations.

[0026] It is to be understood that both the foregoing general description and the following detailed description are exemplary, and are intended to provide further explanation of the technology as claimed. BRIEF DESCRIPTION OF THE DRAWINGS

[0027] Figure 1 Schematic diagram of an automatic operation system of an unmanned tank cleaning machine according to an embodiment of the present invention;

[0028] Figure 2 The figure is a schematic diagram of image recognition of an automatic operation system of an unmanned cargo clearing machine according to an embodiment of the present invention. DETAILED DESCRIPTION

[0029] The preferred embodiments of the present invention will be described in detail below in conjunction with the accompanying drawings to further illustrate the present invention.

[0030] First, combine Figure 1~Figure 2 An automatic operation system of an unmanned tank cleaning machine according to an embodiment of the present invention is described, which is used to control the tank cleaning machine to perform operations.

[0031] like Figure 1~Figure 2 As shown, an automatic operation system of an unmanned cargo clearing machine according to an embodiment of the present invention comprises:

[0032] The spatial recognition module uses a regular grid data model to process the data returned by the radar, combining the 3D coordinate system of the cleaning machine body with the 3D material pile contour and the cabin bulkhead coordinate system scanned by the laser radar, and establishing the 3D material contour based on DEM in the unified coordinate system;

[0033] The intelligent control module obtains the optimal working path based on the recognition result of the space recognition module and the global path planning algorithm or the local trajectory planning algorithm, and the intelligent control module controls the cleaning machine to automatically operate according to the optimal working path;

[0034] The remote control module can display three-dimensional contour graphics, so that the driver can control the tank cleaning machine based on the remote control module.

[0035] The spatial recognition module of the present application combines the three-dimensional coordinate system of the cleaning machine body with the three-dimensional material pile contour and the cabin bulkhead coordinate system scanned by the laser radar, and establishes a three-dimensional material contour based on DEM. In the cabin, the dust concentration is high, the coal is very humid, and even in harsh environments such as rainy and foggy weather, dust particles, fog droplets, water droplets, etc. will cause great interference to the measurement results of the laser radar. The spatial recognition module can adopt an effective data processing algorithm to reduce the interference to a minimum, reduce the influence of ignored tiny particles, depict the contours of the real cabin and material pile behind the dust, rain or fog curtain, and effectively filter interference items such as dust.

[0036] The intelligent control module uses the global path planning algorithm or the local trajectory planning algorithm to calculate the working path based on the recognition results of the space recognition module to obtain the optimal working path. The intelligent control module controls the tank cleaning machine to automatically operate according to the optimal working path to achieve unmanned driving. The remote control module makes it easy for workers to observe the working status of the unmanned tank cleaning machine and can perform remote driving operations.

[0037] In this embodiment, DEM is a digital elevation model, which is a digital simulation of ground terrain through limited terrain elevation data. It is a physical ground model that represents ground elevation in the form of a set of ordered numerical arrays. It is a branch of the digital terrain model, and various other terrain feature values ​​can be derived from it. In the scanning process of the modeling laser radar in the prior art, due to the ups and downs of the surface, the surface data obtained are not equally spaced, especially when the sampling points are dense when going uphill and sparse when going downhill, and the surface data obtained are all discrete data. In order to obtain a surface height image of equal grids, it is necessary to normalize the discrete data obtained by radar scanning. Due to the non-stationarity of the actual bulk surface, the accuracy of DEM mainly depends on the density and distribution of the original sampling points. For the same original data, different interpolation methods have no significant effect on the accuracy of DEM. Considering the characteristics of computer data storage and processing and the convenience and speed of terrain analysis, calculation and representation, the grid spacing of the regular grid DEM can be adjusted according to the on-site situation to improve the accuracy.

[0038] An automatic operation system for an unmanned tank cleaning machine includes a whole machine control module arranged on the chassis of the tank cleaning machine, which controls a proportional valve through a high-precision proportional solenoid valve control board to realize electrified smooth control of the bottom drive system of the tank cleaning machine;

[0039] The whole machine control module aggregates the whole machine's electrical feedback signals, realizes real-time monitoring of the whole machine's electrical and instrumentation systems, receives data access from the intelligent control module, and performs unmanned automatic operation functions.

[0040] The power unit was changed from "diesel engine + fuel tank" to "electric motor + battery", and each drive mechanism is still the original hydraulic transmission mode. In order to achieve smooth electric control of the underlying drive, a reliable whole machine control module must be designed, and the hydraulic circuit must be transformed to convert the operation of each hydraulic unit including valve group control into controlling the proportional valve through a high-precision proportional solenoid valve control board, so as to achieve smooth electrification control of the underlying drive system of the whole machine.

[0041] As the basic hardware and software platform for the underlying control of the whole machine and the monitoring of vehicle body data, the whole machine control module is equipped with a high-speed industrial bus interface, which aggregates the electrical feedback signals of the whole machine and integrates them into the whole machine control module to realize real-time monitoring of the electrical and instrumentation system of the whole machine, receive data access from the intelligent control module, realize autonomous route planning and unmanned automatic operation functions, receive data access from the hatch remote control to realize the hatch remote control function, and receive data access from the remote control module to realize the remote control function.

[0042] The space recognition module includes: integrated modeling 3D laser radar, video return camera, RTK mobile station, inclinometer, 360-degree image surround system, wireless bridge, and 5G communication;

[0043] The integrated modeling 3D lidar generates 3D reconstruction of the cabin and material distribution, the video feedback camera is used to shoot operation videos and transmit the videos remotely, the RTK mobile station performs high-precision coordinate positioning, the inclinometer is used to measure the horizontal angle of the tank cleaning machine, the 360-degree image surround system is used to shoot 360-degree images of the tank cleaning machine, and the wireless bridge and 5G communication feed back the data information of the tank cleaning machine to the remote control module.

[0044] like Figure 2 As shown, image recognition as lidar point cloud data recognition, including bulkhead recognition and stockpile recognition, all utilize deep learning network structure to train the features of each detection model and achieve high-accuracy environmental condition recognition.

[0045] The image recognition method explains the characteristic data contained in the image by establishing and simulating the nervous system of the human brain, thereby fully mining the deep-level characteristics of the target to be detected in the image, and using this data feature to better characterize the classification of the target.

[0046] The deep learning image recognition method interprets the feature data contained in the image by building and simulating the human brain's nervous system, and then fully explores the deep features of the target to be detected in the image, and uses this data feature to better characterize the classification of the target. This method effectively avoids the defects of traditional image recognition methods and improves the accuracy of target image recognition.

[0047] Deep learning is a complex machine learning algorithm that learns the inherent laws and representation levels of sample data. Deep learning combines underlying features to form more abstract high-level representation attribute categories or features to discover the distributed feature representation of data. Ultimately, machines can interpret and analyze data and recognize image data like humans.

[0048] Convolutional neural networks in deep learning use massive amounts of similar or similar data to train a neural network, and then allow the neural network to understand and judge the content of the image. Based on local connections between neurons and hierarchical image transformation, convolutional neural networks apply neurons with the same parameters to different positions in the previous layer of neural networks to obtain a translation-invariant neural network structure. Combined with error gradient design and training of convolutional neural networks, they have excellent performance in line segment extraction tasks.

[0049] The application of convolutional neural network in line segment extraction solves the problem of light detail jitter and improves the extraction accuracy. The dynamic range of the line segment is increased, the line segment image is clear, and the features are obvious. In addition, the problem of broken extracted line segments is well solved, and the fault tolerance of the system is improved.

[0050] In the fully automatic mode of the cabin cleaning machine, image recognition is very important as a necessary supplement to the recognition of lidar point cloud data. It includes the needs of autonomous driving applications in the cabin, such as bulkhead recognition and material pile recognition. The deep learning network structure is used to train the characteristics of each detection model, thereby realizing high-accuracy environmental condition recognition and other recognition functions that are difficult to achieve with traditional methods.

[0051] The remote control module includes: a remote control room, a wireless bridge and 5G communication;

[0052] The wireless bridge and 5G communication are used to receive data information fed back by the space identification module. The remote control room is used to display the first-person real-time video of the tank cleaning machine, the 3D model of the cabin, the operation trajectory display and extension, and the 360-degree panoramic image, to achieve tactile vibration feedback, restore the on-site picture and sound environment, so that the operator can also feel the real environment of the operation site in the remote operation room.

[0053] The key to the remote control module lies in the continuity of video data transmission, the reliability of control data transmission, and low-latency technology.

[0054] The intelligent control module selects the optimal working path based on the indicators of completeness, optimality, real-time performance, adaptability to environmental changes, and meeting the kinematic and dynamic requirements of the unmanned tank cleaning machine. The intelligent control module realizes autonomous planning of the working path and operating actions based on the principle of "divided area operation, nearby material collection, and two-point material supply" of the ship unloader grab.

[0055] The completeness means that the path planning algorithm can obtain effective results within a limited time, and the optimality means finding a collision-free path from the starting point to the end point according to a certain optimal standard.

[0056] A series of discrete points or curves connecting the starting point and the target point are defined as a path. Path planning in unmanned driving technology refers to global path planning in an established environmental map and real-time autonomous obstacle avoidance based on the surrounding environment information and obstacle information perceived by the sensor, and finally planning a safe path to the target. Usually the planned path must meet certain optimal criteria, such as the shortest time, the shortest path or the least energy consumption, and must meet the kinematic and dynamic constraints of the unmanned tank cleaning machine. Therefore, a good path planning algorithm needs to meet the following indicators: completeness, optimality, real-time, adaptability to environmental changes, and meeting the kinematic and dynamic constraints of the unmanned tank cleaning machine. Completeness means that the path planning algorithm can obtain effective results within a limited time; optimality means finding a collision-free path from the starting point to the end point according to a certain optimal standard.

[0057] The global path planning algorithm includes: a graph search-based method, a sampling-based method, and a curve interpolation method;

[0058] The graph search-based method is divided into two steps: establishing an environment map and path search. The environment map can be expressed by a metric map representation or a topological map representation, and the metric map representation can be expressed by a geometric representation or a spatial decomposition method.

[0059] The sampling-based methods include RRT (Random Rapid Search Tree) algorithm, PRM (Probabilistic Road Map) algorithm and State Grid Planner algorithm;

[0060] The curve interpolation method includes a Dubins curve or Reeds-Shepp curve fitting algorithm, a Clothoid curve fitting algorithm, a Bezier curve fitting algorithm, and a polynomial or spline curve fitting algorithm.

[0061] The Dubins curve fitting algorithm generates paths using straight line segments and circular arcs. The Dubins curve is a very simple path generation strategy. The Dubins curve requires the vehicle to move only forward, while the Reeds-Sheep curve allows the vehicle to move both forward and backward. The Clothoid curve is a curve whose curvature changes linearly with the arc length, which is suitable for designing vehicle trajectories.

[0062] The local trajectory planning algorithm includes an artificial potential field method and a dynamic window algorithm;

[0063] In the artificial potential field method, when the unmanned cabin cleaning machine is moving, the target point generates gravitational force on it, and the obstacle generates repulsive force on it, and the movement speed and direction of the unmanned cabin cleaning machine are determined by the resultant force.

[0064] For the artificial potential field method, the selection of gravitational potential field function and repulsive potential field function is crucial. The original algorithm's selection of potential field function makes the unmanned cabin cleaning robot prone to target unreachable problems and local minimum points during movement.

[0065] The dynamic window algorithm collects multiple groups of speeds in the speed space, and then predicts the trajectory of each group of speeds within a period of time, and then evaluates each group of trajectories through an evaluation function, and selects the speed corresponding to the trajectory with the minimum cost and sends it to the unmanned cabin cleaning machine.

[0066] The trajectory evaluation function mainly includes three aspects: the distance to the target, the speed of the robot, and the distance to the next obstacle. The evaluation function comprehensively considers the above three factors to find an optimal path.

[0067] Above, refer to Figure 1~Figure 2An automatic operation system for an unmanned cargo tank cleaning machine according to an embodiment of the present invention is described. The spatial recognition module of the present application combines the three-dimensional coordinate system of the cargo tank cleaning machine body with the three-dimensional material pile contour and the cabin bulkhead coordinate system scanned by the laser radar, and establishes a three-dimensional material contour based on the DEM. In the cabin, where the dust concentration is high and the coal is highly humid, or even in harsh environments such as rainy and foggy weather, dust particles, fog droplets, water droplets, etc. will cause great interference to the measurement results of the laser radar. The spatial recognition module can adopt an effective data processing algorithm to reduce the interference to a minimum, reduce the influence of ignored tiny particles, depict the contours of the real cabin and the material pile behind the dust, rain or fog curtain, and effectively filter interference items such as dust.

[0068] The intelligent control module uses the global path planning algorithm or the local trajectory planning algorithm to calculate the working path based on the recognition results of the space recognition module to obtain the optimal working path. The intelligent control module controls the tank cleaning machine to automatically operate according to the optimal working path to achieve unmanned driving. The remote control module makes it easy for workers to observe the working status of the unmanned tank cleaning machine and can perform remote driving operations.

[0069] It should be noted that, in this specification, the terms "include", "comprises" or any other variations thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements includes not only those elements, but also includes other elements not explicitly listed, or also includes elements inherent to such process, method, article or device. In the absence of further restrictions, the elements defined by the sentence "comprises..." do not exclude the existence of other identical elements in the process, method, article or device including the elements.

[0070] Although the content of the present invention has been described in detail through the above preferred embodiments, it should be appreciated that the above description should not be considered as a limitation of the present invention. After reading the above content, it will be apparent to those skilled in the art that various modifications and substitutions of the present invention will occur. Therefore, the protection scope of the present invention should be limited by the appended claims.

Claims

1. An automatic operation system for an unmanned tank cleaning machine, characterized in that: include: The spatial recognition module uses a regular grid data model to process the data returned by the radar, combining the 3D coordinate system of the cleaning machine body with the 3D material pile contour and the cabin bulkhead coordinate system scanned by the laser radar, and establishing the 3D material contour based on DEM in the unified coordinate system; The intelligent control module obtains the optimal working path based on the recognition result of the space recognition module and the global path planning algorithm or the local trajectory planning algorithm, and the intelligent control module controls the cleaning machine to automatically operate according to the optimal working path; The remote control module can display a three-dimensional contour graphic, so that the driver can control the tank cleaning machine based on the remote control module; The whole machine control module installed on the chassis of the tank cleaning machine controls the proportional valve through the high-precision proportional solenoid valve control board, realizing electrified and smooth control of the bottom drive system of the tank cleaning machine; The whole machine control module aggregates the whole machine's electrical feedback signals, realizes real-time monitoring of the whole machine's electrical and instrumentation systems, receives data access from the intelligent control module, and performs unmanned automatic operation functions.

2. The automatic operation system of an unmanned tank cleaning machine as claimed in claim 1, characterized in that: The space recognition module includes: integrated modeling 3D laser radar, video return camera, RTK mobile station, inclinometer, 360-degree image surround system, wireless bridge, and 5G communication; The integrated modeling 3D lidar generates 3D reconstruction of the cabin and material distribution, the video return camera is used to shoot operation videos and transmit the videos remotely, the RTK mobile station performs high-precision coordinate positioning, the inclinometer is used to measure the horizontal angle of the tank cleaning machine, the 360-degree image surround system is used to shoot 360-degree images of the tank cleaning machine, and the wireless bridge and 5G communication feed back the data information of the tank cleaning machine to the remote control module.

3. The automatic operation system of an unmanned tank cleaning machine as claimed in claim 2, characterized in that: Image recognition, as a form of LiDAR point cloud data recognition, includes bulkhead recognition and stockpile recognition. It uses a deep learning network structure to train the features of each detection model and achieve high-accuracy environmental condition recognition.

4. The automatic operation system of an unmanned tank cleaning machine as claimed in claim 3, characterized in that: The image recognition method explains the characteristic data contained in the image by establishing and simulating the nervous system of the human brain, thereby fully mining the deep-level characteristics of the target to be detected in the image, and using this data feature to better characterize the classification of the target.

5. The automatic operation system of an unmanned tank cleaning machine as claimed in claim 2, characterized in that: The remote control module includes: a remote control room, a wireless bridge and 5G communication; The wireless bridge and 5G communication are used to receive data information fed back by the space identification module. The remote control room is used to display the first-person real-time video of the tank cleaning machine, the 3D model of the cabin, the operation trajectory display and extension, and the 360-degree panoramic image, to achieve tactile vibration feedback, restore the on-site picture and sound environment, so that the operator can also feel the real environment of the operation site in the remote operation room.

6. The automatic operation system of an unmanned tank cleaning machine as claimed in claim 1, characterized in that: The intelligent control module selects the optimal working path based on the indicators of completeness, optimality, real-time performance, adaptability to environmental changes, and meeting the kinematic and dynamic requirements of the unmanned cabin cleaning machine.

7. The automatic operation system of an unmanned tank cleaning machine as claimed in claim 6, characterized in that: The completeness means that the path planning algorithm can obtain effective results within a limited time, and the optimality means finding a collision-free path from the starting point to the end point according to a certain optimal standard.

8. The automatic operation system of an unmanned tank cleaning machine as claimed in claim 1, characterized in that: The global path planning algorithm includes: a graph search-based method, a sampling-based method, and a curve interpolation method; The graph search-based method is divided into two steps: establishing an environment map and path search. The environment map can be expressed by a metric map representation or a topological map representation, and the metric map representation can be expressed by a geometric representation or a spatial decomposition method. The sampling-based methods include RRT (Random Rapid Search Tree) algorithm, PRM (Probabilistic Road Map) algorithm and State Grid Planner algorithm; The curve interpolation method includes a Dubins curve or Reeds-Shepp curve fitting algorithm, a Clothoid curve fitting algorithm, a Bezier curve fitting algorithm, and a polynomial or spline curve fitting algorithm.

9. The automatic operation system of an unmanned tank cleaning machine as claimed in claim 1, characterized in that: The local trajectory planning algorithm includes an artificial potential field method and a dynamic window algorithm; In the artificial potential field method, when the unmanned tank cleaning machine is moving, the target point generates gravitational force on it, and the obstacle generates repulsive force on it, and the speed and direction of the unmanned tank cleaning machine are determined by the resultant force; The dynamic window algorithm collects multiple groups of speeds in the speed space, and then predicts the trajectory of each group of speeds within a period of time, and then evaluates each group of trajectories through an evaluation function, and selects the speed corresponding to the trajectory with the minimum cost and sends it to the unmanned cabin cleaning machine.

Images