Control method and system for automatic climbing of power transmission tower climbing robot

Abstract

The invention provides a control method and system for automatic climbing of a power transmission tower climbing robot. The control method comprises the steps that robot environment data and state information are obtained; performing point cloud processing on the environment data to obtain information of the position and the size of an obstacle; performing linear tracking on the environment data, and determining and adjusting a foot falling position; establishing a Markov decision model according to the obstacle information, the foot position and the robot state information; an obstacle crossing strategy is optimized through a depth deterministic strategy gradient method DDPG; and mapping the optimized obstacle crossing strategy to a tree framework in the behavior robot, and controlling the robot to act to complete an automatic climbing task. According to the method, a self-adaptive climbing strategy can be generated through environment data in combination with reinforcement learning, and efficient climbing of the robot in a complex environment is achieved.

Description

Technical Field

[0001] The present invention relates to the technical field of robot automatic control, and in particular to a control method and system for automatic climbing of a transmission tower climbing robot. Background Art

[0002] Transmission towers, as an integral part of the power system, undertake the important task of transmitting electricity. They are spread over a wide geographical area, ensuring a stable supply of electricity. However, the daily inspection and maintenance of these towers is a labor-intensive and dangerous task. Workers need to conduct inspections at high altitudes, facing adverse weather conditions and complex terrain environments, which not only tests their physical fitness, but also challenges their psychological endurance. Therefore, in order to ensure the safety of workers and improve work efficiency, it is particularly important to develop and apply advanced inspection technologies and equipment.

[0003] Transmission tower climbing robots can replace traditional manual labor to complete the inspection and maintenance of power towers, avoiding technicians from working in bad weather or complex tower truss environments. In addition, removable fall protection devices can be installed on the towers. Compared with the traditional safety rope climbing mode, the risk of technicians falling off the tower can be significantly reduced, improving the safety of tower maintenance operations. However, during the climbing process, the recognition of obstacles and the accuracy of the foothold have a significant impact on the stability and safety of the robot. Therefore, it is necessary to make further improvements in these aspects. Summary of the invention

[0004] In view of the defects in the prior art, the object of the present invention is to provide a control method and system for automatic climbing of a transmission tower climbing robot.

[0005] According to one aspect of the present invention, there is provided a control method for automatic climbing of a transmission tower climbing robot, comprising:

[0006] Get robot environment data and status information;

[0007] Performing point cloud processing on the environmental data to obtain information on the location and size of obstacles;

[0008] Perform linear tracking on the environmental data, and determine and adjust the foot position in combination with the status information;

[0009] Establishing a Markov decision model according to the obstacle information, the footfall position and the state information of the robot;

[0010] Using the Markov decision model as a framework, an optimized obstacle crossing strategy is obtained through a deep deterministic policy gradient method;

[0011] The optimized obstacle crossing strategy is mapped to the tree frame inside the robot, and the robot is controlled to complete the automatic climbing task.

[0012] Preferably, the obtaining of robot environment data and status information includes:

[0013] Pre-collect the 3D point cloud data of the tower;

[0014] Collect 3D point cloud data of towers, obstacles and backgrounds in real time;

[0015] Pre-collect the main image of the tower

[0016] Real-time collection of tower main material images;

[0017] Real-time measurement of the distance between the obstacle and the foot position;

[0018] The obtaining of robot status information includes:

[0019] The state information is obtained by communicating with the robot motion controller through a network; the state information refers to the position information, including the position and posture of each foot relative to the robot body.

[0020] Preferably, performing point cloud processing on the environmental data to obtain information on the location and size of obstacles includes:

[0021] Segment and process the pre-collected point cloud data and the real-time collected point cloud data;

[0022] For the segmented point cloud data, the bounding box method is used to calculate the geometric information of obstacles;

[0023] Combined with the distance data collected in real time, the location of the obstacle is determined.

[0024] Preferably, the linear tracking of the environmental data, combined with the state information, to determine and adjust the foot position, includes:

[0025] Determine the main material straight line of the tower according to the pre-collected main material image of the tower;

[0026] According to the real-time collected tower main material image, the LSD straight line detection algorithm is used to extract the edge straight line features of the tower main material;

[0027] Among the extracted straight line features of the main material edge of the tower, the straight line with the smallest distance from the pre-collected straight line or the straight line selected in the previous frame is selected to complete the tracking;

[0028] Based on the robot's state information and the selected straight line, the relative position between the moving foot and the main material of the tower, as well as the angle between the robot body and the main material of the tower are calculated;

[0029] Based on the relative position and the angle, the landing position of the sports foot is adjusted with the center of the main material of the tower and the angle being zero as a reference.

[0030] Preferably, the Markov decision model is only a stereotyped description of the problem, including the definition of state, action, and transition probability, specifically including:

[0031] State set: S = {s1, s2, ..., s n} represents the set of all possible states of the robot, where s i indicates the i-th state;

[0032] Action set: A = {a1, a2, ..., a m} represents the set of all possible actions of the robot, where a j represents the jth action;

[0033] Transition probability: P(s′|s,a) represents the probability of transitioning to state s′ after executing action a in state s;

[0034] Reward function: R(s,a,s′) represents the immediate reward obtained when action a is executed in state s and transferred to state s′.

[0035] Preferably, the Markov decision model is used as a framework to obtain an optimized obstacle crossing strategy through a deep deterministic policy gradient method, including:

[0036] Constructing an environment network, an actor network, and a critic network; wherein the environment network is a tower model, including the tower and obstacles on the tower main material; the position and size of the obstacles are randomly generated during the learning process; the actor network and the critic network both use a dual network of a real-time network and a target network, and adopt policy gradient optimization;

[0037] The environment network provides the robot state s t , the Critic network generates and evaluates the current action a t Whether the obstacle can be crossed; when it is judged that it can be executed, the Actor network executes the current action a t , the environment network gets reward r t and generate state s t+1 ;

[0038] The experience data s generated by the interaction between the Actor network and the environment network t ,a t ,r t ,s t+1 The data are stored in the experience pool R, and then a batch of data samples are extracted from the experience pool R to the Critic network for training and optimization.

[0039] Preferably, the tree framework comprises:

[0040] Execute the environment perception node to obtain environmental obstacle information and convert it into obstacle queue data;

[0041] Execute the autonomous decision node, obtain the optimized obstacle crossing strategy, and generate a robot action sequence;

[0042] Execute the action sequence node and control each joint of the robot to complete the action process according to the action sequence.

[0043] According to a second aspect of the present invention, there is provided a control system for automatic climbing of a transmission tower climbing robot, comprising:

[0044] Data acquisition module: obtain robot environment data and status information;

[0045] Perception module: performing point cloud processing on the environmental data to obtain information on the location and size of obstacles; performing linear tracking on the environmental data, and determining and adjusting the landing position in combination with the state information;

[0046] Autonomous decision-making module: establishing a Markov decision model according to the obstacle information, the landing position and the state information of the robot; using the Markov decision model as a framework, obtaining an optimized obstacle crossing strategy through a deep deterministic policy gradient method;

[0047] Execution module: maps the optimized obstacle crossing strategy to the tree frame inside the robot, and controls the robot to complete the automatic climbing task.

[0048] Preferably, the data acquisition module includes:

[0049] Binocular camera, used to collect 3D point cloud data of tower obstacles and calculate the size and position of obstacles;

[0050] Industrial camera, used to obtain the image of the main material of the tower, extract the straight line features of the edge of the main material, and display the environmental image in real time;

[0051] Laser distance measuring equipment is used to measure the distance between the obstacle and the landing position.

[0052] Preferably, the industrial camera is fixed next to the moving foot, and the industrial camera is parallel to the frame.

[0053] Compared with the prior art, the embodiments of the present invention have at least one of the following beneficial effects:

[0054] The control method and system for automatic climbing of a transmission tower climbing robot according to the embodiment of the present invention can generate an adaptive climbing strategy through environmental data combined with reinforcement learning, thereby achieving efficient climbing of the robot in complex environments.

[0055] The control method and system for automatic climbing of a transmission tower climbing robot according to an embodiment of the present invention have precise perception performance, and its point cloud processing and straight line tracking ensure the accuracy of environmental information acquisition, providing high-quality data support for decision-making.

[0056] The control method and system for automatic climbing of a transmission tower climbing robot according to an embodiment of the present invention have intelligent decision-making performance, and the obstacle crossing strategy generated by its reinforcement learning algorithm is adaptive and can achieve optimal action planning in different environments.

[0057] The control method and system for automatic climbing of a transmission tower climbing robot according to the embodiment of the present invention have reliability and safety performance. By accurately adjusting the landing position, the stability of the robot's climbing action is significantly improved and the safety risk is reduced. BRIEF DESCRIPTION OF THE DRAWINGS

[0058] Other features, objects and advantages of the present invention will become more apparent from the detailed description of non-limiting embodiments made with reference to the following drawings:

[0059] Figure 1 It is a flow chart of a control method for automatic climbing of a transmission tower climbing robot in one embodiment of the present invention;

[0060] Figure 2 is a structural diagram of a tower climbing robot in one embodiment of the present invention;

[0061] Figure 3 is a point cloud processing framework diagram in one embodiment of the present invention;

[0062] Figure 4 It is a framework diagram of the deep deterministic policy gradient method DDPG in one embodiment of the present invention;

[0063] Figure 5 is a schematic diagram of a behavior tree of an automatic climbing process in an embodiment of the present invention;

[0064] Figure 6 This is a framework diagram of a data acquisition module of a tower climbing robot in one embodiment of the present invention;

[0065] Figure 7 It is the installation position of the human sensor of the tower climbing monitor in one embodiment of the present invention;

[0066] Figure 2In the figure: 100 represents the first moving foot of the robot, 200 represents the robot body frame, and 300 represents the second moving foot of the robot. DETAILED DESCRIPTION

[0067] The present invention is described in detail below in conjunction with specific embodiments. The following embodiments will help those skilled in the art to further understand the present invention, but are not intended to limit the present invention in any form. It should be noted that, for those of ordinary skill in the art, several variations and improvements may be made without departing from the concept of the present invention. These all belong to the protection scope of the present invention.

[0068] In one embodiment of the present invention, a control method for automatic climbing of a transmission tower climbing robot is provided. Figure 1 As shown, the main process is as follows:

[0069] S100, obtaining robot environment data and status information;

[0070] S200, performing point cloud processing on the environmental data obtained by S100 to obtain information on the location and size of obstacles;

[0071] S300, linearly tracking the environmental data obtained in S100, and determining and adjusting the foot position in combination with the state information obtained in S100;

[0072] S400, establishing a Markov decision model according to the obstacle information obtained in S200, the foot position obtained in S300 and the robot state information in S100;

[0073] S500, based on the Markov decision model established in S00, obtains the optimized obstacle crossing strategy through the deep deterministic policy gradient method;

[0074] S600 maps the optimized obstacle crossing strategy obtained in S500 to the tree frame inside the robot and controls the robot to complete the automatic climbing task.

[0075] Through the control method in the above embodiment, the robot can automatically climb along the main material of the tower.

[0076] The robot used in the embodiment of the present invention is as follows Figure 2 As shown, the robot comprises a robot body 200 (also referred to as a frame), and a first moving foot 100 and a second moving foot 300 located on the robot body.

[0077] In one embodiment of the present invention, step S100 is implemented to obtain environmental data and status information of the robot.

[0078] The environmental data includes the three-dimensional point cloud data of the tower obstacle, which is used to subsequently obtain the size of the obstacle; it also includes the precise distance of the obstacle from the upper or lower foot of the tower, which is used to combine with the data of the binocular camera to obtain complete obstacle information, that is, the location and size of the obstacle.

[0079] The environmental data also includes the main material image of the tower, which is used to provide data for the subsequent automatic adjustment of the foot position of the sports foot. Of course, the tower environmental image can also be displayed to the operator in real time to assist in manual decision-making in some embodiments using manual mode.

[0080] In some specific embodiments, a binocular camera can be used to collect three-dimensional point cloud data of the tower obstacle and calculate the size and position of the obstacle; an industrial camera can be used to obtain the main material image of the tower, extract the straight line features of the main material edge, and display the environmental image in real time; a laser ranging device can be used to measure the distance between the obstacle and the landing position.

[0081] Based on the robot environment data collected in the above embodiment, in a preferred embodiment, step S200 is implemented, such as Figure 3 As shown, point cloud processing is performed on the environmental data to obtain information on the location and size of obstacles. Specifically, the following steps can be used:

[0082] S201, segmenting the pre-collected point cloud data and the real-time point cloud data;

[0083] The collected point cloud data includes the main material of the tower (the main structure of the transmission iron, which is in the shape of an L-shaped angle iron column. The robot climbs along this structure), obstacles and other background interferences. To calculate the size of the tower obstacle, it is necessary to first extract the corresponding part of the obstacle in the point cloud and exclude the main material point cloud and background point cloud of the tower. By analyzing the climbing posture of the robot on the tower, it can be seen that the part of the point cloud that needs to be retained is: the point cloud data on the side close to the robot on the plane where the main material surface of the tower is located. Therefore, it is necessary to obtain the equation of the plane where the main material surface of the tower is located in the binocular camera coordinate system, and use the plane equation as the boundary to discard the point cloud away from the robot. Since the relative position of the binocular camera and the robot is fixed, it can be considered that at least in a single climbing task, the relative position of the plane where the main material of the tower is located and the binocular camera remains basically unchanged. Then the plane equation of the main material surface of the tower can be obtained by pre-collecting data before the robot starts automatic climbing. After obtaining the plane coordinates, the point cloud can be segmented.

[0084] S202, calculating geometric information of the obstacle using a bounding box algorithm;

[0085] The segmented point cloud is preprocessed by denoising, and the obstacle size in the current depth image can be obtained by calculating the minimum bounding box of the point cloud.

[0086] It should be noted that during the lifting process of the robot frame (robot body), the depth image obtained by the binocular camera will change with the change of the frame position, and the calculated obstacle size will also change. The actual size of the obstacle can be obtained by analyzing the changes of these two.

[0087] S203, determine the location of the obstacle in combination with the laser ranging data.

[0088] In the above embodiment, steps S201-S203 perform point cloud processing on the collected environmental data, which can ensure the accuracy of environmental information acquisition and provide high-quality data support for subsequent control decisions.

[0089] Further, based on the robot environment data collected in the above embodiment, in a preferred embodiment, step S300 is implemented, such as Figure 3 As shown, the environmental data is tracked linearly, and the foot position is determined and adjusted in combination with the robot status information. Specifically, the following steps can be used:

[0090] S301, tracking the main material edge straight line in real time according to the pre-collected main material straight line parameters.

[0091] The main material identification of the transmission tower adopts the LSD line detection algorithm, which is an algorithm for detecting straight line segments in an image. It determines the edge points by calculating the image gradient, and attempts to construct and verify straight line segments based on the edge points, and finally returns a list of all straight line segments.

[0092] The images obtained by industrial cameras are easily affected by background interference and cannot identify the edge lines of the main material. Therefore, data pre-collection is also required. Specifically: during pre-collection, place a white board or white paper behind the main material to eliminate interference. Select the two longest straight lines in the image at this time, with an angle less than 15° and a closest distance greater than 10cm. Record the parameters of these two straight lines as the pre-collected straight line parameters.

[0093] When the robot starts to move, it needs to track the edge straight line of the main material of the tower according to the pre-collected straight line parameters, and find the straight line with the smallest difference between the straight line detected by the LSD straight line detection method in the N+1 frame image and the straight line found in the N frame. Of course, the first frame uses the pre-collected straight line parameters as the tracking standard.

[0094] S302, based on the robot's state information and the straight line selected in S301, calculate the relative position between the moving foot and the tower main material, as well as the angle between the robot body and the tower main material. Specifically, regarding the relative position, S301 detects the straight edge position of the main material, and S100 has obtained the robot's state information, which includes the position and posture of each foot relative to the robot body. Therefore, the relative position information of the moving foot and the main material can be obtained. Regarding the angle, Figure 7 As shown, there are two industrial cameras on the robot. In S301, each camera detects the edge position of the main material. The angle can be calculated based on the position deviation (d1, d2) detected by the two cameras and the installation distance L between the cameras.

[0095] θ = acos((d1-d2) / L).

[0096] S303, based on the included angle and relative position obtained in S302, the sports foot is adjusted with the center of the main material and the included angle being as close to zero as possible.

[0097] In a preferred embodiment of the present invention, step S400 is implemented to establish a Markov decision model (MDP) of the robot's obstacle-crossing process based on the obstacle information obtained by the environment perception module and the robot state information provided by the robot control system. This step abstracts the relationship between the robot's actions and state transitions during the climbing process into a model to provide a basis for strategy generation. The decision model is only a description of the problem, that is, it defines the current status and actions of the system, specifically:

[0098] State set: S = {s1, s2, ..., s n} represents the set of all possible states of the robot, where s i indicates the i-th state;

[0099] Action set: A = {a1, a2, ..., a m} represents the set of all possible actions of the robot, where a j represents the jth action;

[0100] Transition probability: P(s′|s,a) represents the probability of transitioning to state s′ after executing action a in state s;

[0101] Reward function: R(s,a,s′) represents the immediate reward obtained when action a is executed in state s and transferred to state s′.

[0102] Taking the Markov decision model (MDP) in the above embodiment as a framework, in a preferred embodiment of the present invention, step S500 is implemented, and the deep deterministic policy gradient algorithm (DDPG) is used for reinforcement learning optimization to iteratively optimize the obstacle crossing strategy. Through strategy evaluation, the optimal obstacle crossing strategy of the robot in a complex obstacle environment is obtained. Specifically, Figure 4 As shown, the deep deterministic policy gradient method (DDPG) in this embodiment takes the Actor (policy network)-Critic (value network) method as its basic framework, approximates the policy and action value functions through a deep learning network, and uses the stochastic gradient method to train the parameters in the policy network and value network models.

[0103] For both the policy network and the value network, a dual neural network model of the real-time network and the target network is used. At the same time, the experience replay mechanism is also used. The experience data generated by the interaction between the Actor (policy network) and the environment will be stored in the experience pool, and then a batch of data samples will be extracted for training, making the algorithm easier to converge.

[0104] In some specific embodiments, S500 may adopt the following specific process:

[0105] S501, build DDPG framework: build environment network, actor network, critic network connected in sequence; and experience pool between actor network and critic network;

[0106] S502, initialization: θ Q and θ μ Initialize the Critic network Q(s,a|θ Q ) and the Actor network μ(s|θ μ );θ Q′ =θ Q ,θ μ′ =θ μ , initialize the target network (the target network here refers to the entire DDPG network) weight parameters θ' and μ'; and initialize the experience pool R (random); initialize random noise to obtain the initial state s1 of the environment network;

[0107] S503, generate action: for the i-th state s t Use the Actor network to generate the i-th action, a t =μ(s t |θ μ )+N t , N t Indicates noise;

[0108] S504, execute action: use Actor network to execute action a t ;

[0109] S505, action feedback: get reward r from the environment network t and status t+1 ;

[0110] S506, update experience pool: (s t ,a t ,r t ,s t+1 ) is stored in R;

[0111] S507, randomly select N samples from R (s i ,a i ,r i ,s i+1 );

[0112] S508, update the Critic network by minimizing the loss function L:

[0113]

[0114] y i =r i +γθ'(s i+1 ,μ'(s i+1 |θ μ′ )|θ Q′ )

[0115] S509, update the Actor policy network using policy gradient:

[0116]

[0117] S510, update the target network: θ Q′ =τθ Q +(1-τ)θ Q′ ,θ μ′ =τθ μ +(1-τ)θ μ′ .

[0118] Loop S503-S510.

[0119] In the above-mentioned embodiment, the obstacle crossing strategy generated by the reinforcement learning algorithm is adaptive and can realize the optimal action planning in different environments.

[0120] In a preferred embodiment of the present invention, step S600 is implemented, and the strategy execution transmits the optimized strategy to the control software in the form of a list and maps it to the robot behavior tree structure. According to the node structure of the behavior tree, the robot is guided to perform specific actions to complete the climbing task across obstacles.

[0121] The climbing process is executed based on the behavior tree framework. Figure 5 As shown,

[0122] First, the environment perception node is executed to obtain environmental obstacle information and convert it into obstacle queue data.

[0123] Next, the autonomous decision node is executed to obtain the robot action sequence generated by the DDPG decision model.

[0124] Finally, the action sequence node is executed, and according to the action sequence, each joint of the robot is controlled to complete the action process.

[0125] Among them, the DDPG decision model is used to generate the robot action sequence, including raising the upper or lower foot to the specified position and lifting the frame to the bottom. The single-foot movement strategy corresponds to a command sequence: single-foot demagnetization, single-foot retraction, single-foot adjustment (visual acquisition of adjustment parameters, fixed foot rotation, mobile foot rotation), single-foot lifting to the specified position, single-foot extension, single-foot re-adjustment (visual acquisition of adjustment parameters, fixed foot rotation, mobile foot rotation) and single-foot power removal; the command sequence corresponding to the overall lifting is to lift the whole to the specified position.

[0126] Based on the same inventive concept, in another embodiment of the present invention, a control system for automatic climbing of a transmission tower climbing robot is provided, comprising:

[0127] Data acquisition module: obtain robot environment data and status information;

[0128] Perception module: Perform point cloud processing on environmental data to obtain information on the location and size of obstacles; perform linear tracking on environmental data, and determine and adjust the landing position in combination with status information;

[0129] Autonomous decision-making module: A Markov decision model is established based on obstacle information, foothold location, and robot status information. Using the Markov decision model as a framework, the optimized obstacle crossing strategy is obtained through a deep deterministic policy gradient method.

[0130] Execution module: maps the optimized obstacle crossing strategy to the tree frame inside the robot and controls the robot to complete the automatic climbing task.

[0131] The various modules / units in the above examples of the present invention may specifically refer to the implementation techniques of the corresponding steps of the control method for automatic climbing of the transmission tower climbing robot in the above embodiment, which will not be repeated here.

[0132] In a preferred embodiment, Figure 6As shown in the figure, the hardware used in the data acquisition module includes binocular cameras, industrial cameras and laser ranging equipment. Among them, the binocular camera is used to collect the three-dimensional point cloud data of the tower obstacle and calculate the size and position of the obstacle; the industrial camera is used to obtain the image of the main material of the tower, extract the straight line features of the main material edge, and display the environmental image in real time; the laser ranging equipment is used to measure the distance between the obstacle and the landing position.

[0133] In some specific embodiments, the range of the laser distance measurement is 1200mm, and the error is about ±2mm. In other specific embodiments, considering that the industrial camera of the tower main material image needs to be parallel to the tower main material and can capture the foot landing, the industrial camera can be fixed next to the moving foot and make the camera parallel to the frame, such as Figure 7 shown.

[0134] In some other specific embodiments, the autonomous decision-making module and the execution module are installed in the control software, that is, the control software receives the information obtained by the perception module and generates control instructions to guide the robot to complete the climbing task.

[0135] The above describes the specific embodiments of the present invention. It should be understood that the present invention is not limited to the above specific embodiments, and those skilled in the art can make various modifications or variations within the scope of the claims, which does not affect the essence of the present invention. The above preferred features can be used in any combination without conflicting with each other.

Claims

1. A control method for automatic climbing of a transmission tower climbing robot, characterized in that: include: Get robot environment data and status information; Performing point cloud processing on the environmental data to obtain information on the location and size of obstacles; Perform linear tracking on the environmental data, and determine and adjust the foot position in combination with the status information; Establishing a Markov decision model according to the obstacle information, the footfall position and the state information of the robot; Using the Markov decision model as a framework, an optimized obstacle crossing strategy is obtained through a deep deterministic policy gradient method; The optimized obstacle crossing strategy is mapped to the tree frame inside the robot, and the robot is controlled to complete the automatic climbing task.

2. The control method for automatic climbing of a transmission tower climbing robot according to claim 1, characterized in that: Get robot environment data, including: Pre-collect the 3D point cloud data of the tower; Collect 3D point cloud data of towers, obstacles and backgrounds in real time; Pre-collect the main material image of the tower; Real-time collection of tower main material images; Real-time measurement of the distance between the obstacle and the foot position; The obtaining of robot status information includes: The state information is obtained by communicating with the robot motion controller through a network, and the state information includes the position and posture of each foot relative to the robot body.

3. The control method for automatic climbing of a transmission tower climbing robot according to claim 2, characterized in that: Performing point cloud processing on the environmental data to obtain information on the location and size of obstacles, including: Segment the pre-collected 3D point cloud data of the tower and the real-time collected 3D point cloud data of the tower, obstacles and background; For the segmented point cloud data, the bounding box method is used to calculate the geometric information of the obstacle, that is, the position and posture; The location of the obstacle is determined by combining the real-time collected distance between the obstacle and the foot position.

4. The control method for automatic climbing of a transmission tower climbing robot according to claim 2, characterized in that: The linear tracking of the environmental data and combining the state information to determine and adjust the foot position include: Determine the main material straight line of the tower according to the pre-collected main material image of the tower; According to the real-time collected tower main material image, the LSD straight line detection algorithm is used to extract the edge straight line features of the tower main material; Among the extracted straight line features of the main material edge of the tower, the straight line with the smallest distance from the pre-collected straight line or the straight line selected in the previous frame is selected to complete the tracking; Based on the robot's state information and the selected straight line, the relative position between the moving foot and the main material of the tower, as well as the angle between the robot body and the main material of the tower are calculated; Based on the relative position and the angle, the landing position of the sports foot is adjusted with the center of the main material of the tower and the zero-degree angle as a reference.

5. The control method for automatic climbing of a transmission tower climbing robot according to claim 1, characterized in that: The Markov decision model describes the robot climbing strategy problem in a stereotyped manner, including the definition of states, actions, and transition probabilities, specifically including: State set: S = {s1, s2, ..., s n } represents the set of all possible states of the robot, where s i indicates the i-th state; Action set: A = {a1, a2, ..., a m } represents the set of all possible actions of the robot, where a j represents the jth action; Transition probability: P(s′|s,a) represents the probability of transitioning to state s′ after executing action a in state s; Reward function: R(s,a,s′) represents the immediate reward obtained when action a is executed in state s and transferred to state s′.

6. The control method for automatic climbing of a transmission tower climbing robot according to claim 1, characterized in that: The Markov decision model is used as a framework to obtain an optimized obstacle crossing strategy through a deep deterministic policy gradient method, including: Constructing an environment network, an actor network, and a critic network; wherein the environment network is a tower model, including obstacles on the tower and the main material of the tower, and the position and size of the obstacles are randomly generated during the learning process; the actor network and the critic network both use a dual network of a real-time network and a target network, and adopt policy gradient optimization; The environment network provides the robot state s t , the Critic network generates and evaluates the current action a t Whether the obstacle can be crossed; when it is judged that it can be executed, the Actor network executes the current action a t , the environment network gets reward r t and generate state s t+1 ; The experience data s generated by the interaction between the Actor network and the environment network t ,a t ,r t ,s t+1 The data are stored in the experience pool R, and then a batch of data samples are extracted from the experience pool R to the Critic network for training and optimization.

7. The control method for automatic climbing of a transmission tower climbing robot according to claim 1, characterized in that: The tree framework comprises: Execute the environment perception node to obtain environmental obstacle information and convert it into obstacle queue data; Execute the autonomous decision node, obtain the optimized obstacle crossing strategy, and generate a robot action sequence; Execute the action sequence node and control each joint of the robot to complete the action process according to the action sequence.

8. A control system for automatic climbing of a transmission tower climbing robot, characterized in that: include: Data acquisition module: obtain robot environment data and status information; Perception module: performs point cloud processing on the environmental data to obtain information on the location and size of obstacles; Perform linear tracking on the environmental data, and determine and adjust the foot position in combination with the status information; Autonomous decision-making module: establishing a Markov decision model according to the obstacle information, the landing position and the state information of the robot; using the Markov decision model as a framework, obtaining an optimized obstacle crossing strategy through a deep deterministic policy gradient method; Execution module: maps the optimized obstacle crossing strategy to the tree frame inside the robot, and controls the robot to complete the automatic climbing task.

9. The control system for automatic climbing of a transmission tower climbing robot according to claim 8, characterized in that: include: The data acquisition module comprises: Binocular camera, used to collect 3D point cloud data of tower obstacles and calculate the size and position of obstacles; Industrial camera, used to obtain the image of the main material of the tower, extract the straight line features of the edge of the main material, and display the environmental image in real time; Laser distance measuring equipment is used to measure the distance between the obstacle and the landing position.

10. The control system for automatic climbing of a transmission tower climbing robot according to claim 9, characterized in that: The industrial camera is fixed beside the moving foot, and the industrial camera is parallel to the robot body.

Images