Control system and method of crawler-type obstacle crossing robot

Abstract

The invention relates to a control system and method of a crawler-type obstacle crossing robot, the system comprises a machine end and a handheld end, the machine end comprises a control module, a driving mechanism, an environment sensing module, an analysis processing module and a navigation positioning module; the control module is electrically connected with the driving mechanism, the environment sensing module, the analysis processing module and the navigation positioning module. The handheld end comprises a main control module, a digital image transmission module, a display screen, a vibration module and an operation key position. By integrating manual remote control, autonomous control and following modes, the obstacle crossing robot is endowed with diversified operation capacity, and the intelligent operation requirement of a complex scene is met; the application of the MRAN-AOD-Net model ensures that the environment sensing module accurately obtains obstacle information under extremely severe conditions, and improves the effectiveness of an obstacle crossing strategy and the stability of task execution; and in the aspect of path planning, it is ensured that a path is accurately reproduced during tracking operation, the operation difficulty is reduced, and the operation efficiency is improved.

Description

Technical Field

[0001] The present invention relates to the technical field of robot control, and particularly relates to a control system and method for a tracked obstacle-crossing robot. Background Art

[0002] At present, in complex terrain operation scenarios such as mountain exploration, ruins search and rescue, and swamp survey, tracked obstacle-crossing robots have become the core equipment for performing high-risk tasks due to their excellent terrain adaptability. However, there are still many deficiencies in the existing technologies, which restrict the improvement of their performance and the expansion of their applications: 1. Insufficient complex environment perception ability; in harsh conditions such as dim light, snow and fog, or sandstorms, the environment perception system of existing tracked obstacle-crossing robots is prone to large errors in the recognition and judgment of obstacles, resulting in deviations in the formulation of obstacle-crossing strategies, and easily causing situations such as collisions and jams of the robot, seriously delaying the task progress; 2. Poor stability of repeated operation routes; in tasks that require repeated execution of the same path, the control systems of existing tracked obstacle-crossing robots generally lack effective route memory and stable execution mechanisms, resulting in differences in the driving routes each time the task is executed. Moreover, the inconsistency of the routes not only increases the control difficulty of the operator, reduces the operation efficiency, but also may cause the robot to collide with obstacles due to path deviations, affecting the equipment safety and task completion; 3. The human-machine interaction system needs to be improved; the operation modes of existing obstacle-crossing robots are relatively single, and the manual control mode is difficult to meet the requirements of intelligent operation; at the same time, traditional hand-held remote controllers generally adopt simple circuit designs, with few circuit board layers and compact component layouts. High-frequency and low-frequency signals are extremely prone to crosstalk, resulting in poor signal transmission stability and delayed response of operation instructions, seriously affecting the control accuracy and system reliability, and it is difficult to meet the fine operation requirements in complex scenarios. Summary of the Invention

[0003] The present invention provides a control system and method for a tracked obstacle-crossing robot to solve the problems raised in the above background art.

[0004] To solve the above technical problems, the technical solutions adopted by the present invention are: A control system for a tracked obstacle-crossing robot includes a machine end and a hand-held end. The machine end includes a control module, a traveling mechanism, an environment perception module, an analysis and processing module, and a navigation and positioning module. The control module is electrically connected to the traveling mechanism, the environment perception module, the analysis and processing module, and the navigation and positioning module respectively; the hand-held end includes a main control module, a digital video transmission module, a display screen, a vibration module, and operation keys. The main control module is electrically connected to the digital video transmission module, the display screen, the vibration module, and the operation keys respectively, and the main control module communicates with the control module through CAN; The environmental perception module includes a laser sensor for obtaining distance data between the machine end and surrounding obstacles, and a vision sensor for obtaining image data of the surrounding obstacles of the machine end; The analysis and processing module is used to receive the obtained distance data and image data, determine the obstacle crossing mode through processing of the two data, and synchronously perform map modeling through the map construction unit in the analysis and processing module; The control module is used to drive the traveling mechanism to operate the machine end according to the determined obstacle crossing mode; The main control module is used to receive the map modeling data and send the determined moving path of the machine end to the control module according to the map modeling data; The digital video transmission module is used to receive the image data processed by the analysis and processing module, and after being processed by the main control module, it is displayed on the display screen in the form of numbers or graphics; The operation key positions operate and control the traveling mechanism according to the information data on the display screen; The vibration module receives signals from the analysis and processing module through the main control module, and provides tactile vibration for the handheld end according to the received signals.

[0005] Preferably, the traveling mechanism includes a motor driver, a swing arm motor, a chassis motor, a planetary reducer, and four crawler drive wheels. The four crawler drive wheels are grouped in pairs of two, symmetrically distributed on both sides of the machine end, and a group of crawler drive wheels includes a chassis crawler and a swing arm crawler. The swing arm motor is connected to the swing arm crawler through a planetary reducer, the chassis motor is connected to the chassis crawler through a planetary reducer, and the swing arm motor and the chassis motor are respectively connected to the motor driver.

[0006] Preferably, the analysis and processing module includes an image processing unit, a classification unit, an obstacle crossing decision unit, an attitude switching unit, and a map construction unit; The image processing unit is a network model based on the MRAN-AOD-Net architecture, and is used to perform haze and dust removal processing on the obtained image data to obtain clear image data; The classification unit is a deep learning model based on the convolutional neural network architecture and optimized through training, and is used to extract feature information from the distance data and image data to obtain a classification result according to the extracted obstacle feature information; The obstacle crossing decision unit is used to receive the classification result, and select the obstacle crossing mode through a decision algorithm according to the current position, attitude and environmental information of the machine end, so as to drive the traveling mechanism to operate; The attitude switching unit is used to control the traveling mechanism to perform attitude switching according to the classification result; The map construction unit is used to process the received distance data and fuse the clear image data obtained by the image processing unit to generate a two-dimensional or three-dimensional map containing obstacle information data.

[0007] Preferably, it further includes a following module. The following module includes three mutually communicatively connected LD150 radio frequency modules. Two of the LD150 radio frequency modules are mounted on the machine end, and one LD150 radio frequency module is mounted on the handheld end. And the main control module uses a trilateration algorithm to calculate the relative position between the machine end and the handheld end, and generates a following control instruction for controlling the following of the machine end.

[0008] Preferably, the handheld end further includes a housing and a circuit board disposed in the inner cavity of the housing. The display screen and the operation keys are respectively arranged on the housing. The circuit board is arranged in a 6-layer board structure, and the 6-layer board structure from the top layer to the bottom layer is the component side, the upper ground layer, the signal layer, the power layer, the lower ground layer, and the soldering side in sequence. The main control module, the digital video transmission module, and the vibration module are all electrically connected to the component side, and the display screen and the operation keys are electrically connected to the main control module through the component side.

[0009] The technical solution of the present application also provides a control method for a control system of a tracked obstacle-crossing robot, including a manual remote control mode, an autonomous control mode, and a following mode; The manual remote control mode includes the following steps: S101. The user uses the operation keys to control the machine end to move forward. When the laser sensor measures that the straight-line distance in front of the machine end is less than 1 m, the machine end stops working; S102. The vision sensor is used to continuously capture the obstacles around the machine end and generate image data; S103. The image processing unit performs image processing on the obstacle image data, and sends the processed image data to the main control module through the digital video transmission module, so that the image data is displayed on the display screen in digital or graphic form; S104. The user manipulates the operation keys according to the real-time image data on the display screen to generate an operation instruction; S105. The operation instruction is sent to the control module through the main control module, so that the traveling mechanism can drive the machine end to cross or avoid obstacles according to the instruction to execute the task; The autonomous control mode includes the following steps: S201. For the initial operation, the user starts the machine end to move forward through the operation keys, and makes the environment perception module work together to continuously collect the information data of the objects in the environment around the machine end; S202. Process the information data collected in real time through the analysis and processing module, prompting the machine end to cross or avoid obstacles according to the processing results, and simultaneously generating two-dimensional or three-dimensional map data containing obstacle information data; S203. Use the control module to send the two-dimensional or three-dimensional map data to the handheld end, and the handheld end performs path planning operations through the installed path planning software to determine the movement path data of the machine end in the terrain; Among them, the movement path data includes driving path data and crossing and climbing control data; S204. Send the determined movement path data to the machine end through the main control module; S205. Perform trace operation. After the user starts the machine end, the control module controls the driving mechanism according to the determined movement path data, enabling the driving mechanism to drive the machine end to perform tasks; The following steps are included in the following follow-up mode: S301. The user starts the machine end through the operation keys; S302. Use three LD150 radio frequency modules to collect the distance data between the machine end and the handheld end; S303. Use the trilateration algorithm to calculate the collected distance data and generate a follow-up control instruction; S304. The machine end performs tasks in real-time following the handheld end through the follow-up control instruction.

[0010] Preferably, step S202 specifically includes the following steps: S2021. Use the map construction unit to preprocess the point cloud data collected by the laser sensor through filtering and noise reduction, and finally perform plane fitting and clustering operations to identify different obstacles and terrain features, denoted as laser environment data; S2022. Use the image processing unit to perform image processing on the obstacle image data collected by the vision sensor to identify the type and shape information of the obstacles, denoted as visual environment data; S2023. Use the laser environment data and the visual environment data as the input data of the classification unit, and perform multi-layer convolution and pooling operations through a convolutional neural network to extract the feature information in both data to accurately identify the features of the obstacles; S2024. Compare the identified obstacle features with the obstacle feature set obtained through training in the classification unit to obtain the classification result of the obstacles; Among them, the classification result of the obstacle is any one of low obstacle, medium obstacle, high obstacle or low-lying area; S2025. The obstacle crossing decision unit and the attitude switching unit determine the obstacle crossing mode of the machine end according to the classification result of the obstacle, and complete the obstacle crossing or avoidance of the machine end; S2026. Transmit the visual environment data to the map construction unit for fusion with the laser environment data to generate a two - dimensional or three - dimensional map containing detailed information on the position, shape, and type of obstacles.

[0011] Preferably, the determination of the obstacle - crossing mode in step S2025 is specifically as follows: When the classification result of the obstacle is a low - level obstacle, the obstacle - crossing decision unit determines it as a climbing mode in combination with the current position, attitude, and environmental information of the machine end to complete the obstacle - crossing of the machine end; When the classification result of the obstacle is a medium - level obstacle, the attitude switching unit controls the driving mechanism to perform attitude switching of the machine end. After the attitude switching, the obstacle - crossing decision unit determines it as a climbing mode in combination with the current position, attitude, and environmental information of the machine end to complete the obstacle - crossing of the machine end; When the classification result of the obstacle is a high - level obstacle, the obstacle - crossing decision unit determines it as a bypassing mode in combination with the current position, attitude, and environmental information of the machine end, and sends a signal to the main control module through the control module to prompt the handheld end to vibrate according to the signal, enabling the user to control using the operation keys to complete the obstacle avoidance of the machine end; When the classification result of the obstacle is a low - lying area, the obstacle - crossing decision unit determines it as a spanning mode in combination with the current position, attitude, and environmental information of the machine end to complete the obstacle - crossing of the machine end.

[0012] Preferably, the process of the image processing unit performing image processing on the obstacle image data is as follows: 1) Perform image enhancement on the collected image data through the MRAN network to obtain an enhanced image; 2) Estimate the depth and relative haze and dust levels of the enhanced image through the k - value estimation module in the AOD - Net network, and calculate the de - fogged and de - dusted restored image using the image generation module in the AOD - Net network. The calculation formula is as follows: (1); Among them, F(x) is the image enhanced by the MRAN network, J(x) is the restored clean image, A is the global atmospheric light, and t(x) is the transmission matrix; 3) The AOD - Net unifies the two parameters t(x) and A into a formula, that is, K(x) in formula (2), and directly minimizes the pixel - domain reconstruction error; (2); 4) Input J(x) into the MRAN network again to obtain the finally processed image, denoted as J final (x).

[0013] Preferably, the process of the trilateral positioning algorithm in the following - mode is as follows: 1) Coordinate system definition Take the midpoint of the machine - side LD1 and LD2 as the origin O of the coordinate system; Take the direction from LD1 to LD2 as the X - axis, and the vertical direction as the Y - axis (the positive direction points to the front of the robotic dog); 2) Acquisition of ranging data The distance measured by the handheld - side LD3 to LD1 is denoted as d1, and the distance measured by the handheld - side LD3 to LD2 is denoted as d2; 3) Position calculation Establish a hyperbola equation system: (3); Solve it by the difference method to obtain the position (x, y) of the handheld - side relative to the machine - side: (4); 4) 180° following detection Trigger following only when y < 0 to ensure that the following direction is controllable; Among them, LD1 and LD2 are two LD150 radio - frequency modules on the machine - side, LD3 is the LD150 radio - frequency module on the handheld - side, and L is the known fixed distance between LD1 and LD2.

[0014] By adopting the above - mentioned technical solutions, the beneficial effects obtained by the present invention are as follows: In the present invention, by integrating the manual remote - control mode, the autonomous control mode and the following mode, the tracked obstacle - climbing robot is equipped with diversified intelligent operation modes, which can fully meet the intelligent operation requirements of the tracked obstacle - climbing robot in complex scenarios, and significantly improve the flexibility, autonomy and reliability of its task execution.

[0015] In the present invention, by introducing the MRAN - AOD - Net model, it can ensure that the environmental perception module can quickly and accurately obtain the data information of obstacles under extremely harsh conditions such as dim light, snow - fog or sand - dust storms. And based on the accurate data information obtained, an efficient obstacle - climbing strategy can be formulated to effectively avoid the risk of getting stuck caused by environmental cognitive deviation, thus ensuring the stable and efficient execution of tasks in complex scenarios.

[0016] In the present invention, by the environmental perception module collecting information, the analysis and processing module generating map data, and using the path - planning software on the handheld - side to determine the mobile - path data, when performing trace - following operations in a specific and same environment, the control module can control the driving mechanism according to the mobile - path data, ensuring that the machine - side can accurately reproduce the operation route. This not only reduces the control difficulty of the operator, but also greatly improves the operation efficiency.

[0017] In the present invention, through the innovative design of a six-layer board structure, the power lines and signal lines can be physically isolated, reducing the interference of power noise on signal transmission at the source. Moreover, the independent wiring of high-frequency and low-frequency lines within the signal layer can greatly reduce the problem of signal crosstalk. At the same time, based on the scientific layout of component functions and signal flow directions, the signal transmission path is effectively shortened, significantly improving the signal transmission efficiency and stability.

[0018] In the present invention, the digital video transmission module adopts 5.8 GHz band wireless digital video transmission technology and combines intelligent image compression and decompression algorithms to achieve high-quality and low-latency image transmission, ensuring stable video transmission in complex environments, and providing a reliable guarantee for the stable operation and efficient operation of the tracked obstacle-crossing robot in diverse complex environments.

[0019] In summary, the present invention integrates manual remote control, autonomous control, and following modes, endowing the obstacle-crossing robot with diverse operation capabilities to meet the intelligent operation requirements in complex scenarios. The application of the MRAN-AOD-Net model ensures that the environmental perception module accurately obtains obstacle information under extremely harsh conditions, improving the effectiveness of obstacle-crossing strategies and the stability of task execution.

[0020] In terms of path planning, the environmental perception and path planning software cooperate to ensure accurate reproduction of the route during tracking operations, reducing the operation difficulty and improving the operation efficiency. At the same time, the hardware design of the six-layer board structure effectively isolates the power and signal lines, optimizes the high-frequency and low-frequency wiring, and improves the signal transmission stability. The digital video transmission module can effectively ensure stable video transmission in complex environments.

[0021] Therefore, the present invention adopts the integration of multiple technologies to provide comprehensive guarantee for the stable and efficient operation of the tracked obstacle-crossing robot in complex environments. BRIEF DESCRIPTION OF THE DRAWINGS

[0022] Figure 1 It is a schematic diagram of the first perspective structure of the machine end of the present invention.

[0023] Figure 2 It is a schematic diagram of the second perspective structure of the handheld end of the present invention.

[0024] Figure 3 It is a schematic diagram of the first perspective structure of the handheld end of the present invention.

[0025] Figure 4 It is a schematic diagram of the second perspective structure of the handheld end of the present invention.

[0026] Figure 5 It is a schematic diagram of the machine end of the present invention crossing a small obstacle.

[0027] Figure 6 It is a schematic diagram of the machine end of the present invention crossing a medium-sized obstacle.

[0028] Figure 7 Schematic diagram of the 6-layer board structure of the handheld circuit board of the present invention.

[0029] Figure 8 Schematic diagram of the modules of the machine end and the handheld end of the present invention.

[0030] Figure 9 Schematic diagram of the mode selection of the present invention.

[0031] Figure 10 Schematic diagram of the follow-up mode process of the present invention.

[0032] Figure 11 Schematic diagram of the manual remote control mode process of the present invention.

[0033] Figure 12 Schematic diagram of the autonomous control mode process of the present invention. Detailed implementation manners

[0034] In order to more clearly understand the above objects, features and advantages of the present invention, the present invention will be further described below with reference to the accompanying drawings and embodiments. It should be noted that, without conflict, the embodiments of the present application and the features in the embodiments can be combined with each other.

[0035] In the following description, many specific details are set forth in order to fully understand the present invention. However, the present invention can also be implemented in other ways different from those described herein. Therefore, the present invention is not limited by the specific embodiments disclosed in the following specification. Embodiment 1

[0036] As Figures 1-12 shown, the present invention provides a control system for a tracked obstacle-crossing robot, including a machine end and a handheld end. The machine end includes a frame 1, a control module, a traveling mechanism 2, an environment perception module, an analysis and processing module, and a navigation and positioning module. The traveling mechanism 2 and the environment perception module are respectively installed on the frame 1. The control module, the analysis and processing module, and the navigation and positioning module are respectively arranged in the inner cavity of the frame 1, and the control module is electrically connected to the traveling mechanism 2, the environment perception module, the analysis and processing module, and the navigation and positioning module respectively. The machine end also includes a switch button and an emergency stop button arranged on the frame 1. The switch button is used to start the machine end, and the emergency stop button is used to quickly stop the operation of the machine end. When the machine end is folded up, its size is limited within a space range of ≤540mm×460mm×520mm, and its overall weight (including the load) ≤35kg, which can improve its convenience during transportation and deployment and can adapt to complex and changeable working environments.

[0037] The traveling mechanism 2 includes a motor driver, a swing arm motor, a chassis motor, a planetary reducer, and four crawler drive wheels. The swing arm motor and the chassis motor respectively adopt self-hollow cup brushless motors with high efficiency and energy saving, which can effectively realize the efficient utilization of energy, conform to the concept of sustainable development. The four crawler drive wheels are grouped in pairs of two, symmetrically distributed on both sides of the machine end, and a group of crawler drive wheels includes a chassis crawler 21 and a swing arm crawler 22. The swing arm motor is connected to the swing arm crawler 22 through the planetary reducer, the chassis motor is connected to the chassis crawler 21 through the planetary reducer, and the swing arm motor and the chassis motor are respectively connected to the motor driver.

[0038] The environmental perception module includes a laser sensor 3 for obtaining the distance data between the machine end and surrounding obstacles and a vision sensor 4 for obtaining the image data of surrounding obstacles of the machine end. Among them, the vision sensor 4 can adopt a camera with an infrared fill light function to obtain clear visual information at night or in low light environments, ensuring real-time and comprehensive perception of the surrounding environment information and obstacle information, effectively improving the safety and reliability during the operation, and expanding its application in different environments.

[0039] The analysis and processing module is used to receive the obtained distance data and image data, determine the obstacle crossing mode through the processing of the two data, and synchronously perform map modeling through the map construction unit in the analysis and processing module.

[0040] The control module is used to drive the traveling mechanism 2 to operate the machine end according to the determined obstacle crossing mode.

[0041] The handheld terminal includes a housing 5, a circuit board, a main control module, a digital video transmission module, a display screen 6, a vibration module, and operation keys 7. The display screen 6 and the operation keys 7 are respectively arranged on the housing. The circuit board is arranged in a 6-layer board structure, and the 6-layer board structure from the top layer to the bottom layer is the component side (i.e., the top layer), the upper ground layer (i.e., ground layer 0), the signal layer, the power layer, the lower ground layer (i.e., ground layer 1), and the soldering side (i.e., the bottom layer). The main control module, the digital video transmission module, and the vibration module are all electrically connected to the component side, and the display screen 6 and the operation keys 7 are electrically connected to the main control module through the component side.

[0042] Combined with Figure 7As shown, inside the handheld terminal housing 5, the six-layer structure of the circuit board 6 from top to bottom is the top layer, ground layer 0, signal layer, power layer, ground layer 1, and bottom layer. Thus, the inner cavity of the housing 5 uses a six-layer board structure to separate the power lines and signal lines. When the power lines and signal lines are separated, it ensures to reduce the interference of power noise on signal transmission. At the same time, the high-frequency lines and low-frequency lines in the signal layer are separately routed, and ground protection lines are laid around the high-frequency lines and low-frequency lines to reduce crosstalk between signals. Also, the positions of the electronic components of the handheld terminal on the component side (i.e., the top layer) are reasonably arranged, grouped and laid out according to the functions of the components and the signal flow direction, shortening the signal transmission path and improving the signal transmission efficiency.

[0043] Furthermore, the main control module is a module equipped with the Linux system. Thus, relying on the open-source characteristics of the Linux system, users can customize the functions of the handheld remote control according to their own needs, thereby effectively improving the versatility and universality of the remote control, and being able to fully meet the diverse application needs of different industries and different user groups. It should be noted that the Linux system, as an open-source and widely recognized mature operating system in this field, provides a good foundation for various development scenarios with its high flexibility and scalability. Based on this, those skilled in the art can deeply customize the handheld terminal relying on their solid professional knowledge reserves and rich development experience to meet diverse functional requirements, and this article will not elaborate on this in detail.

[0044] The main control module is used to receive the map modeling data and send the determined moving path of the machine end to the control module according to the map modeling data.

[0045] The digital video transmission module is used to receive the image data processed by the analysis and processing module, and after being processed by the main control module, it is displayed on the display screen 6 in the form of numbers or graphics. Among them, the image data processed by the analysis and processing module can be transmitted to the main control module of the handheld terminal through the digital video transmission module, and the digital video transmission module is connected to the handheld terminal through a high-speed data interface to ensure that the image data can be quickly and stably transmitted to the display screen 6 of the handheld terminal.

[0046] It should be noted that the digital video transmission module adopts wireless transmission technology, such as digital video transmission technology in the 5.8GHz frequency band. Because of the characteristics of high transmission rate and strong anti-interference ability of the digital video transmission technology, it can achieve high-quality and low-latency image transmission, enabling users to obtain the surrounding environment information of the controlled device in real time and clearly, providing strong support for remote operation in manual mode and greatly improving the user experience. Specifically, intelligent image compression and decompression algorithms are developed at the software level of the handheld remote control. At the sending end, the processed image data is compressed in real time to reduce the amount of data transmission; at the receiving end, the image is restored through an efficient decompression algorithm to achieve stable video transmission in complex environments.

[0047] The operation key positions 7 operate and control the traveling mechanism 2 according to the information data on the display screen 6.

[0048] Combined Figure 3 As shown, as a further aspect, the operation key positions 7 include a double rocker, a double lever provided on the housing 5, and at least three buttons respectively distributed on the left and right sides of the housing 5. The handheld end controls the forward, backward, left and right traveling actions of the machine end through the double lever, and controls the swing arm, steering and self-rotation actions of the machine end through the double rocker, enabling the machine end to cross, climb and adjust its posture, and the posture adjustment of the machine end is that the front track swing arm flips backward and the rear track swing arm flips forward.

[0049] The buttons include a power-on button and a power-off button. By long-pressing the power-on button, the screen lights up to complete the power-on of the handheld end; conversely, by long-pressing the power-off button, the screen turns off to complete the power-off of the handheld end. The buttons also include some specific buttons, through which the autonomous control mode of the machine end and the operation mode of the handheld end can be flexibly switched.

[0050] It should be noted that when switching to the autonomous control mode of the machine end, it can be controlled by the control module in the frame 1 to achieve automatic obstacle crossing and traveling of the machine end; while when switching to the operation mode of the handheld end, the operation key positions 7 of the handheld end can generate command signals to control the machine end to achieve traveling, obstacle crossing or obstacle avoidance of the machine end. Specifically, after various operation instructions are input through the operation key positions 7, the operation instructions are processed by the main control module and sent to the control module of the machine end to complete the control of the machine end through the operation key positions 7.

[0051] The vibration module receives and analyzes the signal of the processing module through the main control module, and provides tactile vibration for the handheld end according to the received signal.

[0052] Combined Figure 4 As shown, as a further aspect, the housing 5 of the handheld end is provided with an HDMI output interface, a wired communication interface, a microphone and two speakers, which can meet the information interaction needs in diverse environments.

[0053] Further, communication modules are respectively provided on the machine side and the handheld side. The communication module includes a wired communication unit and a wireless communication unit. In this embodiment, the main control module is communicatively connected to the control module through CAN.

[0054] Furthermore, new type solid-state lithium batteries are respectively provided inside the frame 1 of the machine side and inside the housing 5 of the handheld side. The lithium batteries respectively supply power to the corresponding electrical devices. The voltage of the power supply battery on the machine side is 24V or 36V, and the voltage of the power supply battery on the handheld side is 12V.

[0055] By virtue of the advantages of high energy density, high charge and discharge efficiency, and good safety of the solid-state lithium battery, it can provide a battery life of ≥2.5h for the machine side and the handheld side respectively, effectively guaranteeing the demand for long-time continuous operation.

[0056] Furthermore, interfaces for charging the lithium battery are respectively provided on the frame 1 of the machine side and on the housing 5 of the handheld side to ensure regular or irregular charging of the machine side or the handheld side and ensure the normal operation of the machine side and the handheld side. It is worth emphasizing that the display screen 6 on the handheld side can be used to display the battery level value of the machine side and the battery level value of the handheld side body in real time, so that the user can always master the battery reserves of the machine side and the handheld side and reasonably plan the task duration or arrange charging. At the same time, the display screen 6 on the handheld side can also intuitively display the state parameters such as the rotation speed and working temperature of the swing arm motor and the chassis motor. Once the motor shows conditions such as overheating or abnormal rotation speed, the user can detect and handle it in time to prevent the deterioration of equipment failures. It should be noted that such a display technology has been publicly known in this field. Although it is not described in detail in this article, those skilled in the art can implement this technology in combination with the existing publicly known technology. Specifically, the control module on the machine side collects the lithium battery power data through an ADC (analog-to-digital converter) according to a set period, converts it into a digital value, processes it into a battery percentage data, and sends it to the handheld side after encapsulation according to a custom communication protocol. The handheld side receives and parses it and then displays the power information in real time through the display screen 6.

[0057] Further, a control system of a tracked obstacle-crossing robot further includes a following module. The following module includes three mutually communicatively connected LD150 radio frequency modules. Two of the LD150 radio frequency modules are mounted on the machine side, and one LD150 radio frequency module is mounted on the handheld side. The main control module uses a trilateration algorithm to calculate the relative position between the machine side and the handheld side and generates a following control instruction for controlling the following of the machine side.

[0058] Further, the analysis and processing module includes an image processing unit, a classification unit, an obstacle-crossing decision-making unit, an attitude switching unit, and a map construction unit; The image processing unit is a network model based on the MRAN-AOD-Net architecture, which is used to perform haze and dust removal processing on the acquired image data to obtain clear image data; Before inputting the image into the AOD-Net network, the image f(x) is first sent to the MRAN to obtain the enhanced image F(x); the MRAN uses convolutional kernels of multiple different scales to extract multi-scale features of the image, and its features contain information at different levels of detail of the image; at the same time, the introduction of residual connections enables the network to better learn the change information of the image and avoid losing important details during the feature extraction process; while the attention mechanism makes the network pay more attention to the regions important for target recognition in the image, such as the target object itself, to suppress the interference of unimportant regions such as the background; The AOD-Net network consists of two parts: a k-value estimation module and an image generation module. The k-value estimation module consists of five convolutional layers and is used to estimate the depth and relative haze and dust levels. It forms multi-scale features by fusing filters of different sizes to improve the haze and dust removal effect.

[0059] Therefore, input F(x) into the AOD-Net network. First, estimate the relevant parameters through the k-value estimation module, and then use the image generation module to calculate the haze and dust-removed image J(x), that is, the restored clean image, according to the relevant formula of the atmospheric scattering model. Finally, input J(x) into the MRAN network again to obtain the finally processed image J final (x); The classification unit is a deep learning model based on the convolutional neural network architecture and trained and optimized, which is used to extract feature information from the distance data and image data to obtain a classification result according to the extracted obstacle feature information; The obstacle crossing decision unit is used to receive the classification result and select an obstacle crossing mode through a decision algorithm based on the current position, attitude and environmental information of the machine end to drive the running mechanism 2 to operate; The attitude switching unit is used to control the attitude switching of the running mechanism 2 according to the classification result; The map construction unit is used to process the received distance data and fuse the clear image data obtained by the image processing unit to generate a two-dimensional or three-dimensional map containing obstacle information data. Embodiment 2

[0060] Combined with Figure 9 and Figure 11 As shown, the technical solution of the present application also provides a control method for the control system of a tracked obstacle crossing robot, including a manual remote control mode, and the manual remote control mode includes the following steps: S101. The user controls the machine end to move forward using the operation keys 7. When the straight-line distance measured by the laser sensor 3 in front of the machine end is less than 1 m, the machine end stops working; Among them, the control module sends a read instruction to the laser sensor 3 via the IIC bus in real time to obtain the ranging data. After verification and processing, it is encapsulated and sent to the handheld terminal for display according to the protocol through "protocol reception and transmission". That is, when it is detected that the ranging data is less than 1 m, there is an obstacle; S102. The vision sensor 4 is used to capture the obstacles around the machine end in real time and generate image data; S103. The image processing unit performs image processing on the obstacle image data, and sends the processed image data to the main control module via the digital image transmission module, so that the image data is displayed on the display screen 6 in the form of numbers or graphics; Specifically, after the vision sensor 4 of the machine end obtains the image data, the MRAN-AOD-Net network model is used to perform haze and dust removal processing on the obtained image data to obtain clear image data, that is, the restored clean image; then the clear image data is encoded and sent to the handheld terminal, and after receiving and decoding at the handheld terminal, the image data is displayed on the display screen 6 to facilitate the user to observe the obstacles around the machine end; S104. The user manipulates the operation keys 7 based on the real-time image data on the display screen 6 to generate an operation instruction; S105. The operation instruction is sent to the control module via the main control module, so that the traveling mechanism 2 can drive the machine end to cross or avoid obstacles according to the instruction to perform tasks; Specifically, when it is judged and confirmed at the handheld terminal that the front is a gentle slope (slope 0° - 15°), first send an enable instruction through the operation keys 7 to start the chassis motor, and then send a forward command to make the chassis motor control the corresponding crawler drive wheel (chassis crawler 21) to climb the slope smoothly at a moderate speed; when the chassis crawler 21 travels to the middle of the slope, if it is found that the climbing power is slightly insufficient, a differential driving instruction can be sent appropriately to fine-tune the speed difference between the left and right chassis motors and increase the climbing ability to ensure that the machine end can pass smoothly when the obstacle is a gentle slope. It should be noted that when there are small protrusions or unstable areas on the slope surface, when the chassis crawler 21 is about to approach the aforementioned area (distance about 0.3 - 0.5 m), a control instruction for the swing arm motor is sent through the operation keys 7 to make the swing arm motor drive the swing arm crawler 22 to lift and extend forward, push away or remove small obstacles affecting the travel, and finally the chassis crawler 21 continues to move forward.

[0061] When the handheld terminal determines that there is a steep slope ahead (slope 15°-30°), first send an enable command through operation key position 7 to start the chassis motor, and then send a differential driving command to adjust the speed difference of the left and right chassis motors according to the inclination direction of the slope, so that the chassis track 21 can slowly climb the slope in a zigzag route with a speed controlled at 0.1-0.2 m / s; when the chassis track 21 travels to a steeper area of ​​the steep slope (generally 2-3 meters from the top of the slope), adjust the posture of the machine end through differential driving to ensure sufficient stability, and control the swing arm motor through operation key position 7 so that the swing arm motor controls the swing arm track 22 to extend and fit the edge of the top of the slope or a stable point, and then cooperate with the differential driving and forward commands of the chassis motor to help the machine end climb to the top of the slope.

[0062] Combination Figure 5 As shown, when the handheld terminal determines that there is a small obstacle ahead, an enable command is first sent through the operation key 7 to start the chassis motor, and then a forward command is sent to make the chassis track 21 approach the obstacle at a relatively low speed (about 0.1-0.2 m / s). When the chassis track 21 is about 0.3 m away from the obstacle, the swing arm motor is operated to control the swing arm motor so that when the swing arm motor is running, the swing arm track 22 slowly extends and approaches the obstacle, providing certain assistance and support for the chassis track 21 in the future; then, instructions are continued to be sent to drive the chassis track 21 to move forward. During this period, with the synchronous operation and auxiliary effect of the swing arm track 22, the machine end can smoothly cross the small obstacle.

[0063] Combination Figure 6 As shown, when the handheld terminal determines that there is a medium-sized obstacle ahead, an enable command is first sent through operation key 7 to start the chassis motor, and then a differential driving command is sent to the left and right chassis motors to make the chassis green space move around the obstacle at a slower speed (0.05-0.1 m / s). During this period, the environmental perception module is used to update data in real time to determine the appropriate climbing angle and position.

[0064] After it is determined, the swing arm motor is started by sending an instruction through the operation key 7. The swing arm track 22 extends and moves downward, so that the swing arm track 22 contacts the ground in front of the bottom of the obstacle, forming a stable support. When the swing arm track 22 is in close contact with the ground, a stable supporting force is provided; at this time, the swing arm track 22 is below, supporting the front part of the entire machine end. After the swing arm track 22 is in place for support, the chassis track 21 is controlled to operate by sending an instruction through the operation key 7. Due to the support provided by the swing arm track 22, the chassis track 21 can gradually be raised to a position close to the top of the obstacle, and then the chassis track 21 continues to rotate. Using the supporting force of the swing arm track 22 and the friction between its own track and the ground surface, it slowly crawls forward to gradually cross the obstacle. It should be noted that during this obstacle-crossing process, the swing arm track 22 always maintains stable support to ensure the balance of the chassis track 21 when crossing; and as the chassis track 21 continuously rotates, the machine end gradually crosses the top of the obstacle, and the swing arm track 22 continues to assist in adjusting the posture of the machine end to ensure that the chassis track 21 can successfully cross the obstacle. Finally, after the machine end successfully crosses the obstacle, the swing arm track 22 retracts to a suitable position, and the chassis track 21 returns to the normal driving state to continue the subsequent tasks.

[0065] When it is judged and confirmed at the handheld end that the obstacle in front cannot be crossed or cannot be crossed during the attempt to cross (such as serious slipping of the chassis track 21, the swing arm track 22 cannot provide effective assistance, serious imbalance of the robot posture, etc.), through the real-time image data on the display screen 6, the operation key 7 can control the chassis track 21 to rotate left or right to ensure that the machine end moves around the obstacle to find other feasible paths, that is, to realize bypassing obstacle avoidance work. Embodiment III

[0066] Combined with Figure 9 and Figure 12 As shown, the technical solution of the present application also provides a control method for the control system of a tracked obstacle-crossing robot, including an autonomous control mode, and the autonomous control mode includes the following steps: S201. For the initial operation, the user starts the machine end to move forward by operating the operation key 7, and makes the environmental perception module work together to collect the information data of the objects in the surrounding environment of the machine end in real time; Specifically, through a specific key on the handheld end, the manual remote control mode can be switched to the autonomous control mode of the machine end; S202. The information data collected in real time is processed by the analysis and processing module, prompting the machine end to cross or avoid obstacles according to the processing result, and at the same time generating two-dimensional or three-dimensional map data including obstacle information data; Among them, step S202 includes the following steps: S2021. Use the map construction unit to preprocess the point cloud data collected by the laser sensor 3. The point cloud data is converted from the distance information obtained by the laser sensor 3. The point cloud data is preprocessed by filtering and noise reduction to remove abnormal data points. Finally, the point cloud data is subjected to plane fitting and clustering operations to identify different obstacles and terrain features, which is denoted as laser environment data; S2022. Use the image processing unit to perform image processing on the obstacle image data collected by the vision sensor 4 to identify the type and shape information of the obstacles, which is denoted as vision environment data; S2023. Use the laser environment data and the vision environment data as the input data of the classification unit, and perform multi-layer convolution and pooling operations through a convolutional neural network (CNN) based on deep learning to extract the feature information in both data to accurately identify the features of the obstacles; S2024. Compare the identified obstacle features with the obstacle feature set obtained through training in the classification unit to obtain the classification result of the obstacles; Among them, the classification result of the obstacle is any one of low obstacle, medium obstacle, high obstacle or low-lying area; It should be noted that the classification unit is a deep learning model based on the convolutional neural network architecture and optimized through training. When the model is trained, it is trained with a large number of sample data of different types of obstacles to learn the feature patterns of different obstacles, so as to ensure that it can accurately identify different terrain features such as low obstacles, medium obstacles, high obstacles, and low-lying areas; S2025. The obstacle crossing decision unit and the attitude switching unit determine the obstacle crossing mode of the machine end according to the classification result of the obstacle, and complete the obstacle crossing or obstacle avoidance of the machine end; Among them, the determination of the obstacle crossing mode in step S2025 is specifically as follows: When the classification result of the obstacle is a low obstacle, the obstacle crossing decision unit determines it as a climbing mode in combination with the current position, attitude and environmental information of the machine end, and completes the obstacle crossing of the machine end; When the classification result of the obstacle is a medium obstacle, the attitude switching unit controls the driving mechanism 2 to perform attitude switching of the machine end. After the attitude switching, the obstacle crossing decision unit determines it as a climbing mode in combination with the current position, attitude and environmental information of the machine end, and completes the obstacle crossing of the machine end; Here, the specific implementation instructions of the attitude switching unit are basically the same as the control instructions when facing large obstacles in the manual remote control mode. However, the attitude switching unit in this implementation has multiple sets of data, and it can cooperate with the environment perception module to monitor and adjust the attitude of the machine end in real time to ensure the smooth progress of the form switching and the balance of the machine end; When the obstacle classification result is a high obstacle, the obstacle crossing decision unit determines the bypass mode based on the current position, posture and environmental information of the machine end. When the bypass mode is determined, user intervention is required, and a signal is sent to the main control module through the control module, prompting the vibration module of the handheld end to vibrate according to the signal, reminding the user to rely on the image data on the display screen 6 and use the operation key position 7 to control the machine end to avoid obstacles and find a suitable path or climb over; When the obstacle classification result is a low-lying area, the obstacle crossing decision unit determines the crossing mode based on the current position, posture and environmental information of the machine end, and completes the obstacle crossing of the machine end; When the obstacle crossing decision unit decides on the obstacle crossing mode, the current position, posture and environmental information of the machine can be obtained through the real-time environmental perception module to cooperate with the decision algorithm in the obstacle crossing decision unit to select the best obstacle crossing solution. It is worth mentioning that the decision algorithm adopts a rule-based reasoning algorithm, combined with fuzzy logic control, and comprehensively considers multiple factors to ensure that the selected obstacle crossing strategy is both safe and efficient; S2026, transmitting the visual environment data to a map construction unit for fusion with the laser environment data to generate a two-dimensional or three-dimensional map containing detailed information on the location, shape and type of obstacles; S203, using the control module to send the two-dimensional or three-dimensional map data to the handheld terminal, and the handheld terminal performs a path planning operation through the installed path planning software to determine the moving path data of the machine terminal in the terrain. It should be noted that the path planning software has a path planning algorithm and can independently set the moving path data; Among them, the moving path data includes driving path data and crossing and climbing control data; it is worth noting that when the obstacle crossing mode is the detour mode, due to human intervention, relying on manual operation of the control keys, it is possible to find a suitable moving path or climbing path for the machine end, prompting the path planning software to record it, so as to ensure that when the user specifies the starting point and the end point on the handheld end, the path planning software can independently determine the moving path data through the path planning algorithm; S204, sending the determined moving path data to the machine end through the main control module; S205, tracking operation, after the user starts the machine end, the control module controls the driving mechanism 2 according to the determined moving path data, so that the driving mechanism 2 can drive the machine end to perform the task.

[0067] It should be noted that when the autonomous control mode is applied, it mainly focuses on executing the same path in a specific environment, that is, manual intervention is required when facing the detour mode during the initial operation, while no manual intervention is needed when following the saved mobile path data for the tracking operation. However, when new obstacles appear on the same path, the obstacle-crossing mode can be re-decided through the steps of the initial operation. That is, if the machine directly crosses the obstacle successfully, it will drive normally on the originally set mobile path data; while when deciding on the detour mode, manual intervention is required to drive the machine into the originally set mobile path data and ensure that the path planning software re-determines the mobile path data after the operation.

[0068] It should be emphasized that when the autonomous control mode is applied, it includes but is not limited to the execution of specific environment tasks and can also be used in scenarios of unknown environments, that is, the initial operation of the autonomous control mode is carried out in each unknown environment.

[0069] In addition, the path planning software includes but is not limited to path planning algorithms and can also set mobile path data for users. After the user manually sets it, the path planning software also provides a path checking function to automatically check whether the path is reasonable, such as whether the path passes through obstacles and whether the obstacle-crossing mode matches the obstacle type and the capabilities of the machine end. If there are any unreasonable points, the software will prompt the user to guide the user to make adjustments.

[0070] Furthermore, the process of the image processing unit performing image processing on the obstacle image data is as follows: 1) Image enhancement is performed on the collected image data through the MRAN network to obtain an enhanced image; 2) The depth and relative haze and dust levels of the enhanced image are estimated through the k-value estimation module in the AOD-Net network, and the restored image without haze and dust, that is, the restored clean image, is calculated using the image generation module in the AOD-Net network. The specific calculation formula is as follows: (1); Among them, F(x) is the image enhanced by the MRAN network, J(x) is the restored clean image, A is the global atmospheric light, and t(x) is the transmission matrix; 3) The AOD-Net unifies the two parameters t(x) and A into a formula, that is, K(x) in formula (2), and directly minimizes the pixel-domain reconstruction error; (2); 4) J(x) is input into the MRAN network again to obtain the finally processed image, denoted as J final (x).

[0071] In summary, through the introduction of the MRAN-AOD-Net model, the tracked obstacle-crossing robot of the present invention can ensure that the environmental perception module can accurately capture and analyze environmental information under extremely harsh conditions such as dim light, snow and fog, or sandstorms. With advanced deep learning algorithms and adaptive optimization mechanisms, the model can accurately obtain key data such as the size, position, and shape of obstacles. Based on the accurate data obtained, the machine end can formulate efficient obstacle-crossing strategies, effectively avoiding the risk of getting stuck caused by environmental cognitive biases, thus ensuring the stable and efficient execution of tasks in complex scenarios. Embodiment 4

[0072] Combined with Figure 9 and Figure 10 As shown, the technical solution of the present application also provides a control method for the control system of a tracked obstacle-crossing robot, including a following mode, and the following mode includes the following steps: S301. The user starts the machine end through the operation key 7; Specifically, through a specific button on the handheld terminal, the manual remote control mode can be switched to the following mode of the machine end, which can be used for movement on normal roads or slopes; S302. Three LD150 radio frequency modules are used to collect the distance data between the machine end and the handheld terminal; S303. Use the trilateration algorithm to calculate the collected distance data and generate a following control instruction; S304. The machine end executes tasks by following the handheld terminal in real time through the following control instruction.

[0073] Furthermore, the process of the trilateration algorithm in the following mode is as follows: 1) Coordinate system definition Take the midpoint of LD1 and LD2 of the machine end as the origin O of the coordinate system; Take the direction of LD1→LD2 as the X-axis, and the vertical direction as the Y-axis (the positive direction points to the front of the robot dog); 2) Acquisition of ranging data Take the distance measured by LD3 of the handheld terminal to LD1 as d1, and the distance measured by LD3 of the handheld terminal to LD2 as d2; 3) Position calculation Establish a hyperbola equation system: (3); Solve by the difference method to obtain the position (x, y) of the handheld terminal relative to the machine end: (4); 4) 180° following detection Follow is triggered only when y < 0 (the handheld terminal is behind the machine end) to ensure that the following direction is controllable; Among them, LD1 and LD2 are two LD150 radio frequency modules at the machine end, LD3 is the LD150 radio frequency module at the handheld end, and L is the known fixed distance between LD1 and LD2.

[0074] Dynamic adjustment: Distance control: Maintain Within the set range (1 - 2 m).

[0075] Angle control: Adjust the movement direction of the machine end so that θ = arctan(x / |y|) approaches 0 (directly behind).

[0076] In summary, the present invention innovatively integrates diverse intelligent operation modes, including manual remote control mode, autonomous control mode, and following mode. Among them, the manual remote control mode can achieve precise and flexible manual intervention, the autonomous control mode can achieve intelligent path planning and decision-making, and the following mode can achieve reliable following ability of the machine end. Therefore, the synergy and complementarity of the three modes can fully meet the intelligent operation requirements of the tracked obstacle-crossing robot in complex scenarios, and significantly improve the flexibility, autonomy, and reliability of its task execution.

[0077] Although the embodiments of the present invention have been shown and described, those of ordinary skill in the art can understand that various changes, modifications, substitutions, and variations can be made to these embodiments without departing from the principles and purposes of the present invention. The scope of the present invention is defined by the claims and their equivalents.

Claims

1. A control system for a tracked obstacle-crossing robot, characterized in that: The invention comprises a machine end and a handheld end, wherein the machine end comprises a control module, a driving mechanism, an environment perception module, an analysis and processing module, and a navigation and positioning module, wherein the control module is electrically connected to the driving mechanism, the environment perception module, the analysis and processing module, and the navigation and positioning module respectively; the handheld end comprises a main control module, a digital image transmission module, a display screen, a vibration module, and an operation key position, wherein the main control module is electrically connected to the digital image transmission module, the display screen, the vibration module, and the operation key position respectively, and the main control module is connected to the control module through CAN communication; The environment perception module includes a laser sensor for acquiring distance data between the machine end and surrounding obstacles and a visual sensor for acquiring image data of obstacles around the machine end; The analysis and processing module is used to receive the acquired distance data and image data, determine the obstacle crossing mode after processing the two data, and synchronously perform map modeling through the map building unit in the analysis and processing module; The control module is used to drive the travel mechanism to operate the machine end according to the obstacle crossing mode determined; The main control module is used to receive the map modeling data and send the determined machine end moving path to the control module according to the map modeling data; The digital image transmission module is used to receive the image data processed by the analysis and processing module, and display it on the display screen in the form of numbers or graphics after being processed by the main control module; The operating keys are used to operate and control the travel mechanism according to the information data on the display screen; The vibration module receives the analysis and processing module signal through the main control module, and provides tactile vibration for the handheld terminal according to the received signal.

2. The control system of the crawler obstacle surmounting robot according to claim 1, characterized in that: The traveling mechanism includes a motor driver, a swing arm motor, a chassis motor, a planetary reducer and four track drive wheels. The four track drive wheels are grouped in pairs and symmetrically distributed on both sides of the machine end. A group of track drive wheels includes chassis tracks and swing arm tracks. The swing arm motor is connected to the swing arm tracks through a planetary reducer, the chassis motor is connected to the chassis tracks through a planetary reducer, and the swing arm motor and the chassis motor are respectively connected to the motor driver.

3. The control system of the crawler obstacle surmounting robot according to claim 1, characterized in that: The analysis and processing module includes an image processing unit, a classification unit, an obstacle crossing decision unit, a posture switching unit and a map construction unit; The image processing unit is a network model based on the MRAN-AOD-Net architecture, which is used to perform defogging and dust removal processing on the acquired image data to obtain clear image data; The classification unit is a deep learning model based on a convolutional neural network architecture and has been trained and optimized, and is used to extract feature information from distance data and image data to obtain a classification result based on the extracted obstacle feature information; The obstacle crossing decision unit is used to receive the classification result, and select the obstacle crossing mode through the decision algorithm according to the current position, posture and environmental information of the machine end, so as to drive the driving mechanism to operate; The posture switching unit is used to control the travel mechanism to switch posture according to the classification result; The map construction unit is used to process the received distance data and fuse the clear image data obtained by the image processing unit to generate a two-dimensional or three-dimensional map containing obstacle information data.

4. The control system of the crawler obstacle surmounting robot according to claim 1, characterized in that: It also includes a following module, which includes three LD150 radio frequency modules that are communicatively connected to each other, two of which are mounted on the machine end, and one of which is mounted on the handheld end. The main control module uses a three-sided positioning algorithm to calculate the relative position of the machine end and the handheld end, and generates a following control instruction for controlling the following of the machine end.

5. The control system of the crawler obstacle surmounting robot according to claim 1, characterized in that: The handheld terminal also includes a shell and a circuit board arranged in the inner cavity of the shell. The display screen and operation keys are respectively arranged on the shell. The circuit board adopts a 6-layer board structure, and the 6-layer board structure is sequentially a component surface, an upper ground layer, a signal layer, a power supply layer, a lower ground layer, and a welding surface from the top layer to the bottom layer. The main control module, the digital image transmission module and the vibration module are all electrically connected to the component surface, and the display screen and the operation keys are electrically connected to the main control module through the component surface.

6. A control method using the control system of the crawler obstacle surmounting robot according to any one of claims 1 to 5, characterized in that: Including manual remote control mode, autonomous control mode and follow mode; The manual remote control mode includes the following steps: S101, the user uses the operation key to control the machine end to move forward. When the laser sensor measures the straight-line distance in front of the machine end to be less than 1m, the machine end stops working; S102, using a visual sensor to capture obstacles around the machine in real time and generate image data; S103, performing image processing on the obstacle image data through the image processing unit, and sending the processed image data to the main control module through the digital image transmission module, so that the image data is displayed on the display screen in the form of numbers or graphics; S104, the user manipulates the operation keys according to the real-time image data on the display screen to generate an operation instruction; S105, the operation instruction is sent to the control module via the main control module, so that the driving mechanism can drive the machine end to overcome or avoid obstacles according to the instruction to perform the task; The autonomous control mode includes the following steps: S201, the first operation, the user starts the machine end to move forward by operating the key position, and enables the environment perception module to work together to collect information data of objects in the environment around the machine end in real time; S202, processing the information data collected in real time through the analysis and processing module, prompting the machine end to overcome or avoid obstacles according to the processing results, and simultaneously generating two-dimensional or three-dimensional map data containing obstacle information data; S203, using the control module to send the two-dimensional or three-dimensional map data to the handheld terminal, and the handheld terminal performs a path planning operation through the installed path planning software to determine the movement path data of the machine terminal in the terrain; The moving path data includes driving path data and crossing and climbing control data; S204, sending the determined moving path data to the machine end through the main control module; S205, tracking operation, after the user starts the machine end, the control module controls the driving mechanism according to the determined moving path data, so that the driving mechanism can drive the machine end to perform the task; The following mode includes the following steps: S301, the user starts the machine end by operating the key position; S302, using three LD150 radio frequency modules to collect distance data between the machine end and the handheld end; S303, using a three-sided positioning algorithm to calculate the collected distance data and generate a following control instruction; S304, the machine terminal follows the handheld terminal to perform tasks in real time by following the control instructions.

7. The control method of the control system of the crawler obstacle surmounting robot according to claim 6, characterized in that: Step S202 specifically includes the following steps: S2021. Preprocess the point cloud data collected by the laser sensor by filtering and noise reduction using a map construction unit, and finally perform plane fitting and clustering operations to identify different obstacles and terrain features, which are recorded as laser environment data; S2022, using an image processing unit to process the obstacle image data collected by the visual sensor to identify the type and shape information of the obstacle, and record it as visual environment data; S2023. Use the laser environment data and the visual environment data as input data of the classification unit, and perform multi-layer convolution and pooling operations through a convolutional neural network to extract feature information from the two data to accurately identify the features of obstacles; S2024. Compare the identified obstacle features with the obstacle feature set trained in the classification unit to obtain a classification result of the obstacle; The classification result of the obstacle is any one of low obstacle, medium obstacle, high obstacle or low-lying place; S2025, the obstacle crossing decision unit and the posture switching unit determine the obstacle crossing mode of the machine end according to the classification result of the obstacle, and complete the obstacle crossing or obstacle avoidance of the machine end; S2026. Transmit the visual environment data to the map construction unit for fusion with the laser environment data to generate a two-dimensional or three-dimensional map containing detailed information on the location, shape and type of obstacles.

8. The control method of the control system of the crawler obstacle surmounting robot according to claim 7, characterized in that: The determination of the obstacle crossing mode in step S2025 is specifically as follows: When the obstacle classification result is a low obstacle, the obstacle crossing decision unit determines the climbing mode based on the current position, posture and environmental information of the machine end, and completes the obstacle crossing of the machine end; When the obstacle classification result is a medium obstacle, the posture switching unit controls the driving mechanism to switch the posture of the machine end. After the posture switching, the obstacle crossing decision unit determines the climbing mode based on the current position, posture and environmental information of the machine end, and completes the obstacle crossing of the machine end. When the obstacle classification result is a high obstacle, the obstacle decision unit determines the bypass mode based on the current position, posture and environmental information of the machine end, and sends a signal to the main control module through the control module, prompting the handheld end to vibrate according to the signal, so that the user can use the operation key control to complete the obstacle avoidance of the machine end; When the obstacle is classified as a low-lying area, the obstacle crossing decision unit determines the crossing mode based on the current position, posture and environmental information of the machine end, and completes the obstacle crossing of the machine end.

9. The control method of the control system of the crawler obstacle surmounting robot according to claim 7, characterized in that: The process of the image processing unit performing image processing on the obstacle image data is as follows: 1) Performing image enhancement on the collected image data through the MRAN network to obtain an enhanced image; 2) The depth and relative haze and dust levels of the enhanced image are estimated by the k-value estimation module in the AOD-Net network, and the restored image with dehazing and dust removal is calculated by the image generation module in the AOD-Net network. The calculation formula is as follows: （1）； Among them, F(x) is the image enhanced by the MRAN network, J(x) is the restored clean image, A is the global atmospheric light, and t(x) is the transmission matrix; 3) AOD-Net unifies the two parameters t(x) and A into one formula, namely K(x) in equation (2), and directly minimizes the pixel domain reconstruction error; （2）； 4) Input J(x) into the MRAN network again to obtain the final processed image, denoted as J final (x).

10. The control method of the control system of the crawler obstacle surmounting robot according to claim 6, characterized in that: The process of the three-side positioning algorithm in the following mode is as follows: 1) Coordinate system definition The midpoint of the machine ends LD1 and LD2 is taken as the origin O of the coordinate system; The direction from LD1 to LD2 is the X-axis, and the vertical direction is the Y-axis; 2) Distance measurement data acquisition The distance from handheld terminal LD3 to LD1 is recorded as d1, and the distance from handheld terminal LD3 to LD2 is recorded as d2; 3) Position calculation Set up the hyperbolic equation system: （3）； By using the difference method to solve, we can get the position (x, y) of the handheld terminal relative to the machine terminal: （4）； 4) 180° follow-up detection The following is triggered only when y<0 to ensure that the following direction is controllable; Among them, LD1 and LD2 are two LD150 radio frequency modules on the machine end, LD3 is the LD150 radio frequency module on the handheld end, and L is the known fixed distance between LD1 and LD2.

Images