Obstacle crossing control method and device for climbing robot based on multi-source intelligent perception

Abstract

The application relates to the technical field of robot control, in particular to a climbing robot obstacle control method and equipment based on multi-source intelligent sensing, which comprises the following steps: acquiring real-time state data of a climbing robot, acquiring attitude deviation characteristics and variation trends of an actual attitude of the robot body relative to a reference attitude according to body attitude data of a current control period, so as to determine a target adjustment speed of coil current in a next control period; acquiring stress deviation characteristics reflecting unbalanced load states of each working electromagnetic wheel according to motor currents of each working electromagnetic wheel in the current control period, so as to determine target adjustment accelerations of coil currents of each working electromagnetic wheel in the next control period, generate S-shaped control trajectories corresponding to the coil currents of the working electromagnetic wheels, and control the actual coil currents of each working electromagnetic wheel in the next control period to be dynamically adjusted to theoretical coil currents along the S-shaped control trajectories. Therefore, the stability of the climbing robot obstacle control is improved.

Description

Technical Field

[0001] This application relates to the field of robot control technology, specifically to a method and device for controlling obstacle crossing of climbing robots based on multi-source intelligent perception. Background Technology

[0002] Maintenance work on high-voltage transmission lines is accompanied by extremely high safety risks. Electromagnetic adsorption climbing robots have been widely used in high-altitude work scenarios such as transmission towers due to their advantages such as strong adsorption force and flexible control.

[0003] Existing climbing robots typically rely on environmental perception during obstacle crossing control, using only the identification of obstacles to plan and trigger obstacle-crossing actions. However, in complex high-altitude environments, when a climbing robot performs an obstacle-crossing action (i.e., some electromagnetic wheels are de-energized and lifted), the number of support points for the entire robot decreases, causing a sudden change in the overall adsorption load. Coupled with external disturbances such as high-altitude winds, this can easily lead to deviations in the robot's operating posture and severe unevenness in the local load on the remaining working electromagnetic wheels.

[0004] Traditional control methods often employ fixed control parameters when adjusting the current in the electromagnetic wheel coils to maintain adhesion, failing to deeply integrate the climbing robot's real-time state perception. This results in the control process being unable to adaptively adjust the coil current response speed and acceleration based on the robot's real-time attitude deviations and the degree of local force deviations in each electromagnetic wheel. When facing sudden changes in obstacle crossing forces, fixed-parameter control methods are prone to problems such as lag in adhesion force adjustment, uneven adjustment processes, or insufficient compensation for local adhesion forces. This can lead to overall instability of the climbing robot or even falls from heights, severely reducing the stability and operational safety of obstacle crossing in complex scenarios. Summary of the Invention

[0005] To address the aforementioned technical problems, the purpose of this application is to provide a method and device for controlling obstacle crossing of a climbing robot based on multi-source intelligent perception. The specific technical solution adopted is as follows:

[0006] In a first aspect, embodiments of this application provide a method for controlling obstacle crossing of a climbing robot based on multi-source intelligent perception, the method comprising the following steps:

[0007] Acquire real-time status data of the climbing robot and determine the target basic adsorption force to maintain stable adsorption of the climbing robot;

[0008] When performing an obstacle-crossing maneuver that causes some electromagnetic wheels to lose power and lift, the target base attraction force is redistributed to the working electromagnetic wheels that are currently maintaining the attraction state, and the theoretical coil current of each working electromagnetic wheel is obtained.

[0009] Based on the body attitude data of the current control cycle in the real-time status data, obtain the attitude deviation characteristics and changing trend of the actual body attitude relative to the reference attitude, so as to determine the target adjustment speed of the coil current in the next control cycle.

[0010] Based on the motor current of each working electromagnetic wheel in the current control cycle from the real-time status data, the force deviation characteristics reflecting the unbalanced load state of each working electromagnetic wheel are obtained, so as to determine the target adjustment acceleration of the coil current of each working electromagnetic wheel in the next control cycle.

[0011] Using the target adjustment speed and target adjustment acceleration as control parameters, an S-shaped control trajectory is generated for the corresponding working electromagnetic wheel coil current, so as to control the actual coil current of each working electromagnetic wheel to be dynamically adjusted to the theoretical coil current along the S-shaped control trajectory in the next control cycle.

[0012] In one embodiment, obtaining the theoretical coil current of each working electromagnetic wheel includes:

[0013] The target basic adsorption force is evenly distributed to each working electromagnetic wheel that is currently maintaining the adsorption state, and the target distributed adsorption force that each working electromagnetic wheel needs to bear at the moment is determined.

[0014] Based on a preset electromagnetic force conversion model and combined with the structural parameters of the working electromagnetic wheel, the target distribution adsorption force is converted into the theoretical coil current of each working electromagnetic wheel.

[0015] In one embodiment, determining the attitude deviation characteristics includes:

[0016] The instantaneous attitude deviation of the computer body attitude data in a set axis is obtained by unifying the dimensions of the instantaneous attitude deviation in different axes and extracting statistical features from the unified instantaneous attitude deviation to obtain the attitude deviation features.

[0017] In one embodiment, determining the target adjustment speed of the coil current for the next control cycle includes:

[0018] Calculate the rate of change of the instantaneous attitude offset within the current control cycle, and extract the rate of change component characterizing the expansion of the attitude deviation as the trend of change;

[0019] The attitude deviation features and the changing trend are weighted and fused into an adaptive amplification coefficient, and the preset initial value of the adjustment speed is amplified to obtain the target adjustment speed.

[0020] In one embodiment, the process of obtaining the force deviation feature is as follows:

[0021] For each sampling moment within the current control cycle, calculate the statistical benchmark value of the motor current of all the working electromagnetic wheels, and determine the relative difference of the motor current of each working electromagnetic wheel relative to the statistical benchmark value;

[0022] The relative differences at each sampling time within the current control cycle are aggregated to obtain the force deviation characteristics of each working electromagnetic wheel.

[0023] In one embodiment, determining the target adjustment acceleration of the current in each working electromagnetic wheel coil of the next control cycle includes:

[0024] An acceleration gain coefficient is constructed using the force deviation characteristics, and the preset initial value of the adjustment acceleration is amplified by the acceleration gain coefficient to obtain the target adjustment acceleration of the current of each working electromagnetic wheel coil.

[0025] In one embodiment, determining the relative difference between the motor current of each working electromagnetic wheel and the statistical reference value includes:

[0026] The relative difference is determined based on the ratio of the motor current of each working electromagnetic wheel to the average motor current of all working electromagnetic wheels at each sampling time.

[0027] In one embodiment, the step of dynamically adjusting the actual coil current of each working electromagnetic wheel along an S-shaped control trajectory to the theoretical coil current in the next control cycle includes:

[0028] The actual coil current of the working electromagnetic wheel in the current control cycle is obtained. The actual coil current is used as the starting point of the adjustment, and the corresponding theoretical coil current is used as the ending point of the adjustment. Combined with the target adjustment speed, target adjustment acceleration, and control jerk, an S-shaped control trajectory with smooth transition of coil current is fitted and generated.

[0029] In one embodiment, the control jerk is determined based on the target adjustment speed, the target adjustment acceleration, and the duration of the control cycle.

[0030] Secondly, embodiments of this application also provide an obstacle-crossing control device for climbing robots based on multi-source intelligent sensing, including:

[0031] Theoretical current calculation module: acquires real-time status data of the climbing robot and determines the target basic adsorption force to maintain the stable adsorption of the climbing robot; when performing obstacle crossing action and causing some electromagnetic wheels to lose power and lift, the target basic adsorption force is redistributed to the working electromagnetic wheels that are currently maintaining the adsorption state, and the theoretical coil current of each working electromagnetic wheel is obtained.

[0032] Adjustment speed generation module: Based on the body attitude data of the current control cycle in the real-time status data, obtain the attitude deviation characteristics and changing trend of the actual body attitude relative to the reference attitude, so as to determine the target adjustment speed of the coil current in the next control cycle.

[0033] Adjustment acceleration generation module: Based on the motor current of each working electromagnetic wheel in the current control cycle from the real-time status data, obtain the force deviation characteristics reflecting the unbalanced load state of each working electromagnetic wheel, so as to determine the target adjustment acceleration of the coil current of each working electromagnetic wheel in the next control cycle;

[0034] S-shaped trajectory control module: Using the target adjustment speed and target adjustment acceleration as control parameters, it generates an S-shaped control trajectory for the corresponding working electromagnetic wheel coil current, so as to control the actual coil current of each working electromagnetic wheel to be dynamically adjusted to the theoretical coil current along the S-shaped control trajectory in the next control cycle.

[0035] This application has at least the following beneficial effects:

[0036] This application achieves this by dynamically redistributing the target basic adsorption force maintaining the overall machine's adsorption to the remaining electromagnetic wheels that maintain their working state during obstacle crossing when the local wheel system loses power and lifts. It accurately calculates the theoretical coil current under the current special stress state, ensuring that even with fewer core operating points, the remaining electromagnetic wheels can immediately provide sufficient total adsorption force compensation, thus mitigating the risk of slippage caused by obstacle crossing from a global force perspective. Secondly, by extracting the machine's attitude deviation characteristics and their changing trends during the current control cycle and integrating them into global control parameters, the target adjustment speed for the next control cycle is determined. When the machine experiences severe or continuously deteriorating pitch or roll instability, a higher current adjustment rate is automatically applied, significantly shortening the response delay of adsorption force compensation. This achieves feedforward control and rapid attitude stabilization, overcoming the limitations of fixed parameters. The control lag is addressed; furthermore, by utilizing the current sensitivity of each independent drive motor to reflect changes in mechanical load, the force deviation characteristics of each working electromagnetic wheel relative to the overall average are captured. Based on this, a target adjustment acceleration is set for each electromagnetic wheel. For specific wheels subjected to sudden and harsh loads, a higher acceleration is given to rapidly calibrate the adsorption state, significantly improving the local adsorption redundancy safety under harsh and variable conditions. Finally, the target adjustment speed and target adjustment acceleration are jointly incorporated into the control logic to fit the S-shaped control trajectory of the traction electromagnetic wheel coil current, ensuring extremely smooth current change slope and first / second-order continuity, eliminating the impact and current overshoot caused by disordered adjustment. This makes the entire high-voltage obstacle crossing process in complex field operations both highly stable and smooth, greatly reducing mechanical impact and mechanical vibration. Attached Figure Description

[0037] Figure 1A flowchart illustrating the steps of a climbing robot obstacle-crossing control method based on multi-source intelligent perception provided in one embodiment of this application; Figure 2 This is a structural diagram of a climbing robot; Figure 2 In the diagram, 101 is the body of the climbing robot, 102 is the robotic arm, 103 is the monitoring pan-tilt unit, and 104, 105, and 106 are the three wheel modules of the climbing robot. Figure 3 This is a schematic diagram of a climbing robot overcoming an obstacle in a single attempt. Figure 3 In the diagram, 31 represents a power transmission tower, and 32 represents an obstacle. Sub-figure ① shows the climbing robot moving to the obstacle; sub-figure ② shows the climbing robot entering obstacle-crossing mode; sub-figure ③ shows the climbing robot's wheel module 1 lifting to cross the obstacle; sub-figure ④ shows the climbing robot's wheel module 2 lifting to cross the obstacle; sub-figure ⑤ shows the climbing robot's wheel module 3 lifting to cross the obstacle; and sub-figure ⑥ shows the climbing robot completing the obstacle crossing. Detailed Implementation

[0038] The following description, in conjunction with the accompanying drawings, details the specific scheme of the obstacle-crossing control method and equipment for climbing robots based on multi-source intelligent perception provided in this application.

[0039] Please see Figure 1 The diagram illustrates a flowchart of a climbing robot obstacle-crossing control method based on multi-source intelligent perception according to an embodiment of this application. The method includes the following steps:

[0040] The climbing robot is placed on a power transmission tower (in this embodiment, a vertical steel pipe power transmission tower is used as an example) to perform climbing operations, and real-time status data of the climbing robot during its movement is acquired. Specifically, the real-time status data includes: obstacle information, robot posture data, coil current data and motor current data of each electromagnetic wheel, and the climbing speed of the climbing robot. The specific acquisition method is as follows:

[0041] The obstacle information is obtained by environmental modeling and target recognition through depth image information of the scene in front captured by the monitoring pan-tilt unit. Based on the depth value of the identified obstacle in the depth image, the actual distance between the obstacle and the climbing robot is directly calculated.

[0042] The body attitude data includes the attitude angles of the climbing robot, which are acquired in real time through the IMU unit inside the climbing robot;

[0043] The coil current data and motor current data of each electromagnetic wheel are obtained by monitoring the underlying electrical operation data of the climbing robot, and are used to reflect the current actual electromagnetic attraction force of each electromagnetic wheel and the changes in the physical load it bears.

[0044] The climbing speed is obtained by solving the chassis kinematics.

[0045] To ensure the timing consistency of multi-source data fusion analysis, the sampling frequency for the aircraft attitude data and climbing speed is set to 500Hz (i.e., a sampling period of 2ms). Since the PWM control frequencies of the electromagnetic coil and drive motor are typically high (2kHz and 1kHz respectively in this embodiment), this embodiment calculates the effective values ​​of the coil current and motor current within each sampling period of 2ms. After processing, the final output coil current data and motor current data are strictly consistent with the aircraft attitude data and climbing speed in terms of timing and data volume.

[0046] The structural diagram of the climbing robot in this embodiment is shown below. Figure 2 As shown, 101 is the climbing robot body, which internally houses a battery module (for powering the climbing robot), an industrial control computer (composed of an industrial embedded host, used to receive and process real-time status data during the climbing robot's operation, and output coil current adjustment commands for generating an S-shaped control trajectory based on the calculated target adjustment speed and target adjustment acceleration), and an electrical mounting plate (integrating the power and control systems required by the climbing robot, receiving commands from the industrial control computer to control the robotic arm, movement, and obstacle crossing of the climbing robot); 102 is the robotic arm. 103 is a monitoring pan-tilt unit used to acquire real-time scene information in front of the climbing robot; 104, 105, and 106 are the three wheel modules of the climbing robot (numbered 1, 2, and 3 respectively, each wheel module includes two independently controllable electromagnetic wheels), which are composed of linear guide rails, electric push rods, integrated joint motors, electromagnetic wheels, and wheel system brackets. They can be fixed to the power transmission tower by electromagnetic attraction for vertical movement and can also perform the up and down movement of the wheel module on the linear guide rails by electric push rods and motors to complete obstacle crossing.

[0047] Furthermore, based on the above data and operating environment parameters, the target foundation adhesion force for maintaining stable adhesion of the climbing robot is determined. First, the design parameters of the target power transmission tower are obtained, including the inclination angle between the tower structure and the horizontal ground. A dynamic friction model between the gear train module and the transmission tower surface was established using experimental calibration and contact mechanics theory to obtain the dynamic friction coefficient at the current climbing speed. .

[0048] Based on force analysis, the total frictional force required to keep the climbing robot from sliding down... Must meet: Where m is the total weight of the robot (including the load), and g is the acceleration due to gravity (taken as 9.8 m / s²). 2The actual frictional force between the robot and the tower is determined by both the adsorption force and the normal component of gravity. , where sin is the trigonometric sine function and cos is the trigonometric cosine function.

[0049] It should be noted that, in this embodiment, the climbing robot performs vertical climbing operations on the steel pipe pole, and the tilt angle... The angle is 90°, and the normal component of gravity is perpendicular to the tower surface. The value is 0. Therefore, the physical force model is simplified, and the theoretical critical adsorption force that can maintain stable robot adsorption must satisfy... Considering that in actual climbing operations, when the adsorption force equals a critical value, even minor disturbances can easily cause a fall, while excessive adsorption force increases movement resistance and affects climbing efficiency, an electromagnetic adsorption force coefficient for normal operation is introduced. The final target basic adsorption force is set as .in, The value range is (1, 1.5]. Implementers can set the safety margin according to the specific scenario. In this embodiment... The preferred value is 1.1.

[0050] The climbing robot overcomes obstacles by recognizing them and lifting them using its wheel system module. A schematic diagram of the climbing robot during a single obstacle-crossing maneuver is shown below. Figure 3 As shown. Among them, Figure 3 As shown in sub-figure ①, the climbing robot moves to the obstacle. Based on the obstacle distance recognition information obtained from environmental visual perception and the climbing robot's movement speed, it calculates the current distance between wheel module 1 and the obstacle. When the distance is less than the obstacle-crossing threshold, it... Figure 3 The climbing robot shown in sub-figure ② first lifts its body; the obstacle crossing threshold ranges from [3, 10] cm, and its size can be set by the implementer according to the implementation scenario. In this embodiment, the obstacle crossing threshold is 5 cm.

[0051] Figure 3 As shown in sub-diagram ③, the driving parameters of wheel modules 2 and 3 are controlled to increase their adsorption force. Then, the coil current of wheel module 1 is cut off to make it lose its electromagnetic adsorption ability and lift it over obstacles.

[0052] Figure 3 As shown in sub-figure ④, the distance between wheel module 2 and the obstacle is calculated based on the size parameters between the wheel modules of the climbing robot and the movement speed when crossing the obstacle. When the distance is less than the obstacle crossing threshold, the wheel module 1 is controlled to descend and restore its electromagnetic adsorption capacity. The drive parameters of wheel modules 1 and 3 are controlled to increase their adsorption force. The coil current of wheel module 2 is cut off to make it lose its electromagnetic adsorption capacity and lift it to cross the obstacle.

[0053] Figure 3 As shown in sub-figure ⑤, the distance between wheel module 3 and the obstacle is calculated based on the size parameters between the wheel modules of the climbing robot and the speed of movement when crossing the obstacle. When the distance is less than the obstacle crossing threshold, the wheel module 2 is controlled to descend and restore its electromagnetic adsorption capacity. The drive parameters of wheel modules 1 and 2 are controlled to increase their adsorption force. The coil current of wheel module 3 is cut off to make it lose its electromagnetic adsorption capacity and lift it to cross the obstacle.

[0054] final Figure 3 As shown in sub-diagram ⑥, the obstacle crossing is completed, the control wheel module 3 descends and restores its electromagnetic adsorption capability.

[0055] To address the problem that traditional obstacle crossing control methods typically employ fixed parameters and cannot cope with sudden changes in obstacle crossing forces leading to instability, this application proposes dynamic adaptive adjustment of control parameters based on S-curve control.

[0056] First, the control cycle of the climbing robot's drive parameters is set to 20ms, and the currently executing control cycle is recorded as the nth control cycle.

[0057] Then, a three-dimensional Cartesian coordinate system is constructed for the climbing robot to follow the movement. Its direction is defined as follows: when the robot is attached to the surface of the power transmission tower and climbing vertically, the positive X-axis is the direction of movement of the climbing robot (i.e., the opposite direction of gravity), the positive Y-axis is the left direction of the climbing robot, and the positive Z-axis is the direction perpendicular to the surface of the power transmission tower and outward.

[0058] In this coordinate system, the X-axis and Y-axis attitude angles in the robot's attitude data are defined as the angles of rotation of the climbing robot around the X-axis (representing left and right tilt) and around the Y-axis (representing up and down pitch), respectively. Based on this, the various operational variables of the climbing robot in the nth control cycle are obtained: the X-axis and Y-axis attitude angles in the robot's attitude data are expressed as... , The coil current data for each working electromagnetic wheel (corresponding to the left and right wheels of wheel train modules 1, 2, and 3 respectively) are expressed as follows: , , , , , The motor current data for each working electromagnetic wheel (corresponding to the left and right wheels of wheel train modules 1, 2, and 3 respectively) are expressed as follows: , , , , , .

[0059] To maintain stable climbing of the climbing robot, under normal conditions, all wheel modules in operation share the overall target base adhesion force equally. For example, when wheel modules 1, 2, and 3 are all in the adsorption working state, the adsorption force borne by each wheel module satisfies the following conditions: And the sum of the three equals ,in, The adsorption force borne by wheel module 1 The adsorption force borne by wheel module 2. The adsorption force borne by wheel module 3.

[0060] Furthermore, based on Maxwell's electromagnetic field theory and Ampere's circuital law, the conversion relationship between the electromagnetic attraction force F and the coil current I is derived as follows: ,in, Let be the air gap length of the electromagnet. The number of coil turns. The permeability of free space, This represents the contact area between a single electromagnetic wheel and the transmission tower. Based on this, the theoretical coil current required to maintain the attraction of each working electromagnetic wheel can be calculated. .

[0061] It should be noted that the theoretical coil current It is not a fixed value, but rather changes dynamically as the climbing robot performs its obstacle-crossing actions. During obstacle-crossing, the target wheel module (e.g., wheel module 1) cuts off the coil current for de-energization and lifting, leaving only wheel modules 2 and 3 in operation. At this point, the target's base adhesion force needs to be... The load is redistributed evenly to the working wheel train modules that are currently in an adsorption state, meaning that wheel train modules 2 and 3 each take on the task. Based on this, the theoretical coil current of each working electromagnetic wheel (i.e., the electromagnetic wheel that remains in the adsorption state) at the current moment is recalculated. Meanwhile, based on the aforementioned data sampling frequency (500Hz, i.e., sampling period of 2ms) and control period (20ms), the number of sampling data points included in a single control period can be determined to be T=10.

[0062] In this embodiment, the obstacle-crossing control of the remaining working wheel modules is described as an example when wheel module 1 loses power and lifts to cross an obstacle. To address the problem of body posture instability caused by force changes and external disturbances during obstacle crossing, this application first constructs an adaptive speed weight by analyzing the posture deviation characteristics and changing trends of the climbing robot, which is used to positively adjust the target adjustment speed of the S-shaped control trajectory in the next control cycle.

[0063] During obstacle crossing, the robot's overall support distribution constantly changes due to the alternating lifting and lowering of the wheel modules, which can easily cause deviations in the robot's motion posture angle. When the deviation increases, even if the overall target's basic adsorption force remains unchanged, the robot already shows a tendency to become unstable. If the coil current adjustment response is lagging and cannot quickly provide support and adsorption force, it will lead to a serious risk of falling.

[0064] Considering that the greatest impact on stability is typically the lateral roll around the X-axis and the vertical pitch around the Y-axis, attitude angles along the X and Y axes are used to quantitatively analyze attitude deviation characteristics. Specifically:

[0065] First, calculate the robot's attitude offset within the nth control cycle. Considering that the robot is climbing a vertical steel pipe in this embodiment, the standard attitude angles of its X and Y axes under the reference attitude are... and All are 0°. Therefore, the X-axis and Y-axis attitude angles at the last sampling moment of the nth control cycle (i.e., time T) are obtained. and Calculate the instantaneous attitude offset of its actual attitude relative to the reference attitude. , : , .

[0066] Furthermore, to quantify the relative severity of short-term attitude instability and to standardize the dimensions of instantaneous attitude deviations along different axes, the maximum values ​​of the instantaneous attitude deviations along the X and Y axes for all control cycles within the past second, including the nth control cycle, are selected. These values ​​are then used to evaluate the current instantaneous attitude deviation. and The maximum value is normalized to map it to the interval [0, 1]. Specifically, if the climbing robot has been running for less than 1 second, statistics are performed from the start of the climbing robot to the current nth control cycle; if the maximum value is 0, the corresponding normalized result is directly recorded as 0. The instantaneous attitude offset after normalization is then calculated. and Statistical feature extraction is performed to obtain the attitude deviation features of the actual body posture relative to the reference posture in the nth control cycle. In this embodiment, the attitude deviation characteristics of the nth control cycle ;in, Instantaneous attitude offset The result after normalization Instantaneous attitude offset The result after normalization.

[0067] It should be understood that attitude deviation characteristics It quantifies the degree of attitude deviation of the organism relative to recent historical levels within the current control cycle. The larger the value, the more severe the instability of the current body. At this time, it is necessary to increase the target adjustment speed of the coil current in the next control cycle so that the actual coil current of each electromagnetic wheel can respond faster and reach the theoretical coil current state, thereby quickly restoring the stability of the body.

[0068] Furthermore, the rate of change of the instantaneous attitude offset within the nth control cycle is calculated. Using the first-order least squares method, linear fitting is performed on the instantaneous attitude offsets along the X and Y axes at all sampling times within the nth control cycle. The slopes of the fitted lines are then used as the rates of change of the instantaneous attitude offsets along the X and Y axes, respectively. and A positive and large slope indicates that the attitude deviation of the aircraft is rapidly increasing; a negative slope indicates that the attitude deviation is decreasing.

[0069] To extract the rate of change component characterizing the increase in attitude deviation as the final trend, the calculated... and The determination is as follows: if the slope is less than 0, it indicates that the body attitude is recovering to the expected stable state. At this time, no additional adjustment acceleration should be triggered, so the rate of change component of the corresponding axis is directly assigned to 0; if the slope is greater than or equal to 0, the original value is retained.

[0070] Subsequently, in order to convert the rate of change with physical dimensions into a dimensionless relative severity, a preset maximum tolerable rate of change threshold (e.g., set to...) is used. The maximum permissible instability rollover angular velocity of the organism is statically mapped onto the processed rate of change components. Specifically, the current... and Dividing by the maximum tolerable rate of change threshold, if the calculated result is greater than 1, it is truncated to 1, thereby reducing the tolerance rate of change. and Map to the interval [0, 1]. After mapping, calculate the average of the two axes to obtain the attitude offset change trend in the nth control cycle. , In the formula, for The result after mapping processing for The result after mapping.

[0071] It reflects the relative speed at which the organism evolves towards instability and deterioration within the current control cycle. The larger the value, the greater the need to improve the control response speed; conversely, a value of 0 indicates that the organism is recovering from instability.

[0072] Based on this, the previously determined attitude deviation characteristics can be subsequently... and attitude offset change trend Weighted fusion is performed to construct an adaptive amplification coefficient, which is then amplified to the preset initial value of the adjustment speed, ultimately obtaining the target adjustment speed of the coil current in the (n+1)th control cycle. .

[0073] In this embodiment, the target adjustment speed is calculated as follows:

[0074]

[0075]

[0076] In the formula, The adaptive speed weights for the (n+1)th control cycle. As the first weight, It is the second weight, and Its size can be set by the implementer according to the sensitivity requirements of the implementation scenario to deviation and rate of change. In this embodiment, it is set as follows: ,Will Defined as the adaptive amplification factor. To adjust the initial speed value, the implementer can set it according to the weight and workload of the climbing robot in the implementation scenario. The greater the weight and workload of the climbing robot, the larger the coil current is required to maintain a larger attraction force, and the faster the coil current is adjusted. In this embodiment... A / s.

[0077] When the attitude deviation characteristics in the nth control cycle and attitude offset change trend A large value indicates that the climbing robot's posture deviation is significant and rapidly deteriorating. To prevent excessive posture deviation from causing severe instability or even a fall, the control system uses an adaptive amplification factor. Significantly increasing the target adjustment speed accelerates the response speed of the coil current to the target's attraction force, allowing the electromagnetic wheel to quickly provide sufficient attraction force to stabilize the body; conversely, when , When the value approaches 0, it indicates that the current motion posture of the organism is stable, and the adaptive velocity weight is adjusted accordingly. When the value approaches 0, the adaptive amplification factor approaches 1, and the control system does not need to additionally amplify the response speed; it directly adjusts the speed according to the preset initial value. The coil current of each electromagnetic wheel is adjusted smoothly.

[0078] Furthermore, based on the above processing, this application obtains the force deviation characteristics by analyzing the fluctuation characteristics of the motor current on each working electromagnetic wheel of the climbing robot, which is used to reflect the unbalanced state of the local load of each electromagnetic wheel during the obstacle crossing process of the climbing robot.

[0079] Since the drive motors provide power to the electromagnetic wheels, their motor current directly reflects the changes in the physical load borne by those wheels. When the climbing robot experiences uneven stress due to obstacle-crossing actions (reduced force points) or external wind loads, the load borne by different electromagnetic wheels will vary. To maintain stable forward movement, each drive motor will automatically adjust and maintain a different motor current. Therefore, when the motor current of a particular electromagnetic wheel continuously deviates from the overall average level, it indicates a persistent force imbalance (overload or suspension) on that wheel.

[0080] Therefore, taking the calculation of the control parameters of the left wheel of wheel module 2 in the next control cycle after wheel module 1 is de-energized and lifted as an example, the process of obtaining the target adjustment acceleration is explained:

[0081] First, for each sampling time t within the current nth control cycle, acquire the motor current of all electromagnetic wheels that remain in operation (i.e., the four electromagnetic wheels on the left and right sides of wheel modules 2 and 3), and calculate their average value. Use this average value as the statistical benchmark value for the motor current of all operating electromagnetic wheels at the corresponding time, denoted as [reference value]. .

[0082] Subsequently, based on the motor current of the target working electromagnetic wheel (e.g., the left wheel of wheel module 2) at each sampling time... Compared with the statistical benchmark value The ratio of these ratios determines the relative degree of difference at that moment. Specifically, the absolute value of the natural logarithm of this ratio is taken. By employing the logarithmic absolute value function, a significant positive difference value can be output regardless of whether the electromagnetic wheel motor current is higher or lower than the overall mean. Specifically, before calculating the relative difference, a statistical benchmark value is first used. Perform a validity determination: If A value of 0 indicates that none of the working electromagnetic wheels are currently outputting power (e.g., the robot has stopped or power has been cut off due to a malfunction). In this case, the force deviation judgment fails, and subsequent difference calculations are skipped. The relative difference of the electromagnetic wheel at the current sampling time is assigned a value of 0. Similarly, when the motor current itself is 0, causing the logarithmic function to be uncalculated, the relative difference is also directly recorded as 0. Only when... The logarithmic ratio is calculated only when the value is greater than 0 and the motor current of the target working electromagnetic wheel is also greater than 0.

[0083] Next, the relative differences at each sampling time (a total of T sampling points) within the current control cycle are aggregated, specifically the mean value statistics over the time dimension, to obtain the final force deviation characteristics of the target working electromagnetic wheel. The formula is: , where ln is a logarithmic function with the natural constant as the base.

[0084] Finally, utilizing the aforementioned force deviation characteristics Constructing acceleration gain coefficients And by using the acceleration gain coefficient to adjust the preset initial value of acceleration. The amplification process is performed to obtain the target adjustment acceleration of the coil current of the left wheel of wheel module 2 in the (n+1)th control cycle. : Among them, the preset initial value of the adjustment acceleration The size can be set by the implementer based on the climbing robot's weight and workload; in this embodiment, it is preferred. A / s 2 .

[0085] When the force deviates from the characteristic A large value indicates that the electromagnetic wheel is in a severe high-load overload state or a low-load disconnection state, resulting in extreme imbalance of forces in a localized area of ​​the machine body. At this time, the control system utilizes the constructed acceleration gain coefficient... By significantly increasing the target adjustment acceleration, the coil current of the electromagnetic wheel can adjust the adsorption force at a faster rate of change, thereby enhancing the dynamic response capability of the control system to sudden changes in local load and preventing instability when crossing obstacles. Conversely, if the loads on each electromagnetic wheel are balanced and the force deviation characteristic tends to 0, there is no need to further amplify the adjustment acceleration; the preset initial value can be used directly. Perform smooth control.

[0086] Similarly, the target adjustment acceleration of other electromagnetic wheels currently in operation is independently calculated in the (n+1)th control cycle using the above process.

[0087] Based on the above steps, the target adjustment speed of the current in each working electromagnetic wheel coil is obtained. and independent target-adjusted acceleration Subsequently, based on the aforementioned target adjustment speed, target adjustment acceleration, and control cycle duration, the control jerk of the coil current is calculated and determined. The specific calculation method follows the conventional kinematic constraints of S-shaped control theory to ensure the continuity of the second derivative at trajectory switching points.

[0088] Finally, the coil current of each working electromagnetic wheel is smoothly adjusted within the (n+1)th control cycle. Specifically, the actual coil current of each working electromagnetic wheel at the end of the current nth control cycle is obtained; this actual coil current is used as the adjustment starting point, and the theoretical coil current corresponding to the current state is used as the adjustment starting point. To adjust the endpoint; adjust the global target speed calculated above. Target-specific acceleration adjustment for each electromagnetic wheel The corresponding control accelerometer is input into the S-curve control algorithm to fit and generate an S-shaped control trajectory that smoothly transitions the coil current to approach the target value. The S-curve control algorithm is a well-known existing technology, and its specific process will not be elaborated upon.

[0089] Subsequently, the control system drives the coil current of each working electromagnetic wheel to be dynamically adjusted smoothly and quickly along its corresponding S-shaped control trajectory until the theoretical coil current is reached, thereby maintaining the body's extremely high stability and safety when dealing with sudden changes in force when crossing obstacles.

[0090] During the subsequent climbing process, when it is necessary to control wheel modules 2 and 3 to lift and overcome obstacles due to power failure, the same logic loop is used to execute the above steps to dynamically adjust and control the remaining wheel modules in the working state, thereby completing the climbing robot's safe obstacle crossing on the power transmission tower and the climbing operation such as hanging and retrieving the anti-fall rope.

[0091] Based on the same inventive concept as the above method, embodiments of this application also provide a climbing robot obstacle-crossing control device based on multi-source intelligent sensing, including:

[0092] Theoretical current calculation module: acquires real-time status data of the climbing robot and determines the target basic adsorption force to maintain the stable adsorption of the climbing robot; when performing obstacle crossing action and causing some electromagnetic wheels to lose power and lift, the target basic adsorption force is redistributed to the working electromagnetic wheels that are currently maintaining the adsorption state, and the theoretical coil current of each working electromagnetic wheel is obtained.

[0093] Adjustment speed generation module: Based on the body attitude data of the current control cycle in the real-time status data, obtain the attitude deviation characteristics and changing trend of the actual body attitude relative to the reference attitude, so as to determine the target adjustment speed of the coil current in the next control cycle.

[0094] Adjustment acceleration generation module: Based on the motor current of each working electromagnetic wheel in the current control cycle from the real-time status data, obtain the force deviation characteristics reflecting the unbalanced load state of each working electromagnetic wheel, so as to determine the target adjustment acceleration of the coil current of each working electromagnetic wheel in the next control cycle;

[0095] S-shaped trajectory control module: Using the target adjustment speed and target adjustment acceleration as control parameters, it generates an S-shaped control trajectory for the corresponding working electromagnetic wheel coil current, so as to control the actual coil current of each working electromagnetic wheel to be dynamically adjusted to the theoretical coil current along the S-shaped control trajectory in the next control cycle.

Claims

1. A climbing robot obstacle-crossing control method based on multi-source intelligent perception, characterized in that, The method includes the following steps: Acquire real-time status data of the climbing robot and determine the target basic adsorption force to maintain stable adsorption of the climbing robot; When performing an obstacle-crossing maneuver that causes some electromagnetic wheels to lose power and lift, the target base attraction force is redistributed to the working electromagnetic wheels that are currently maintaining the attraction state, and the theoretical coil current of each working electromagnetic wheel is obtained. Based on the body attitude data of the current control cycle in the real-time status data, obtain the attitude deviation characteristics and changing trend of the actual body attitude relative to the reference attitude, so as to determine the target adjustment speed of the coil current in the next control cycle. Based on the motor current of each working electromagnetic wheel in the current control cycle from the real-time status data, the force deviation characteristics reflecting the unbalanced load state of each working electromagnetic wheel are obtained, so as to determine the target adjustment acceleration of the coil current of each working electromagnetic wheel in the next control cycle. Using the target adjustment speed and target adjustment acceleration as control parameters, an S-shaped control trajectory is generated for the corresponding working electromagnetic wheel coil current, so as to control the actual coil current of each working electromagnetic wheel to be dynamically adjusted to the theoretical coil current along the S-shaped control trajectory in the next control cycle. The determination of the attitude deviation characteristics includes: The instantaneous attitude deviation of the computer body attitude data in a set axis is obtained by unifying the dimensions of the instantaneous attitude deviation in different axes and extracting statistical features from the unified instantaneous attitude deviation to obtain the attitude deviation features. The determination of the target adjustment speed of the coil current for the next control cycle includes: Calculate the rate of change of the instantaneous attitude offset within the current control cycle, and extract the rate of change component characterizing the expansion of the attitude deviation as the trend of change; The attitude deviation features and the changing trend are weighted and fused into an adaptive amplification coefficient, and the preset initial value of the adjustment speed is amplified to obtain the target adjustment speed. The process of obtaining the force deviation characteristics is as follows: For each sampling moment within the current control cycle, calculate the statistical benchmark value of the motor current of all the working electromagnetic wheels, and determine the relative difference of the motor current of each working electromagnetic wheel relative to the statistical benchmark value; The relative differences at each sampling time within the current control cycle are aggregated to obtain the force deviation characteristics of each working electromagnetic wheel; The determination of the target adjustment acceleration of the current of each working electromagnetic wheel coil in the next control cycle includes: An acceleration gain coefficient is constructed using the force deviation characteristics, and the preset initial value of the adjustment acceleration is amplified by the acceleration gain coefficient to obtain the target adjustment acceleration of the current of each working electromagnetic wheel coil.

2. The obstacle-crossing control method for climbing robots based on multi-source intelligent perception as described in claim 1, characterized in that, The method of obtaining the theoretical coil current of each working electromagnetic wheel includes: The target basic adsorption force is evenly distributed to each working electromagnetic wheel that is currently maintaining the adsorption state, and the target distributed adsorption force that each working electromagnetic wheel needs to bear at the moment is determined. Based on a preset electromagnetic force conversion model and combined with the structural parameters of the working electromagnetic wheel, the target distribution adsorption force is converted into the theoretical coil current of each working electromagnetic wheel.

3. The obstacle-crossing control method for climbing robots based on multi-source intelligent perception as described in claim 1, characterized in that, The determination of the relative difference between the motor current of each working electromagnetic wheel and the statistical benchmark value includes: The relative difference is determined based on the ratio of the motor current of each working electromagnetic wheel to the average motor current of all working electromagnetic wheels at each sampling time.

4. The obstacle-crossing control method for climbing robots based on multi-source intelligent perception as described in claim 1, characterized in that, The control of dynamically adjusting the actual coil current of each working electromagnetic wheel along the S-shaped control trajectory to the theoretical coil current in the next control cycle includes: The actual coil current of the working electromagnetic wheel in the current control cycle is obtained. The actual coil current is used as the starting point of the adjustment, and the corresponding theoretical coil current is used as the ending point of the adjustment. Combined with the target adjustment speed, target adjustment acceleration, and control jerk, an S-shaped control trajectory with smooth transition of coil current is fitted and generated.

5. The obstacle-crossing control method for climbing robots based on multi-source intelligent perception as described in claim 4, characterized in that, The control jerk is determined based on the target adjustment speed, the target adjustment acceleration, and the duration of the control cycle.

6. A climbing robot obstacle-crossing control device based on multi-source intelligent perception, implementing the climbing robot obstacle-crossing control method based on multi-source intelligent perception as described in claim 1, characterized in that, The obstacle crossing control device includes: Theoretical current calculation module: acquires real-time status data of the climbing robot and determines the target basic adsorption force to maintain the stable adsorption of the climbing robot; when performing obstacle crossing action and causing some electromagnetic wheels to lose power and lift, the target basic adsorption force is redistributed to the working electromagnetic wheels that are currently maintaining the adsorption state, and the theoretical coil current of each working electromagnetic wheel is obtained. Adjustment speed generation module: Based on the body attitude data of the current control cycle in the real-time status data, obtain the attitude deviation characteristics and changing trend of the actual body attitude relative to the reference attitude, so as to determine the target adjustment speed of the coil current in the next control cycle. Adjustment acceleration generation module: Based on the motor current of each working electromagnetic wheel in the current control cycle from the real-time status data, obtain the force deviation characteristics reflecting the unbalanced load state of each working electromagnetic wheel, so as to determine the target adjustment acceleration of the coil current of each working electromagnetic wheel in the next control cycle; S-shaped trajectory control module: Using the target adjustment speed and target adjustment acceleration as control parameters, it generates an S-shaped control trajectory for the corresponding working electromagnetic wheel coil current, so as to control the actual coil current of each working electromagnetic wheel to be dynamically adjusted to the theoretical coil current along the S-shaped control trajectory in the next control cycle.

Images