An intelligent motion control method for a four-wheel differential robot

Abstract

The present invention provides an intelligent motion control method for controlling a four-wheel differential robot, belonging to the field of artificial intelligence mobile robots. This application utilizes multiple sensor signals, comprehensively combines real-time online weather forecast information, performs information fusion based on the existing computing system carried by the robot, quantifies and outputs the decision result after fusion to the control of the robot motion control system, selects the controller control algorithm according to the quantified decision result, and performs the final motion control. It solves the problems that the existing four-wheel differential chassis of the robot in the power distribution network engineering environment, facing complex road paving conditions, complex traffic conditions and complex weather, a single motion control strategy will cause mechanical structure wear due to torque matching; the problem of reducing the effective working time due to a large amount of ineffective energy consumption; the problem of weakening the working environment adaptability due to the inability to adapt to weather changes.

Description

Technical Field

[0001] The present invention belongs to the field of artificial intelligence mobile robots, and particularly relates to an intelligent motion control method for a four-wheel differential robot. Background Art

[0002] With the development of artificial intelligence, computer, and control technologies, as well as the development and improvement of modern mobile robot technologies, mobile robots can move from structured environments such as laboratories to complex and comprehensive environments such as battlefields, mining areas, and engineering operation sites. The task requirements for mobile robots have also evolved from simple artificial intelligence services in the past to more complex tasks such as autonomous navigation, fault detection, and patrol reconnaissance. Therefore, mobile robots are widely used in social life, such as the use of intelligent public service robots, intelligent nursing robots, fire rescue robots, etc.

[0003] Under the guidance of the "Robot Industry Development Plan (2016 - 2020)", mobile robots (AGVs) are being increasingly applied to various engineering environments with harsh geographical conditions, complex working conditions, and variable weather, including applications in the distribution network engineering environment. In the existing distribution network engineering environment, the road surface paving conditions are poor, the construction site working conditions are complex, and the weather conditions change beyond human control. At this time, the motion control part, which is the core foundation of the distribution network engineering robot, becomes particularly important in the overall work.

[0004] Looking at all kinds of existing four-wheel differential motion chassis of robots on the market, most of them are designed for paved roads. Due to their simple application environment, the control strategies of their corresponding control systems are very single, and they cannot adjust their own motion modes and control strategies according to the external road surface paving conditions, the complexity of the construction site working conditions, and the real-time weather conditions. As a result, it is difficult to meet the various challenges faced in the complex distribution network environment. Eventually, problems such as mechanical structure wear caused by unadjustable torque matching, reduced effective working time caused by a large amount of ineffective energy consumption when the motor is on standby, and weakened working environment adaptability caused by the inability to adapt to weather changes occur for mobile robots in the distribution network working environment. Summary of the Invention

[0005] In order to overcome the deficiencies of the above-mentioned prior art, the present invention provides an intelligent motion control method for a four-wheel differential robot.

[0006] To achieve the above object, the present invention provides the following technical solutions:

[0007] An intelligent motion control method for a four-wheel differential robot, comprising the following steps:

[0008] Obtain the multi-source heterogeneous information of the multi-sensors of the robot in various working conditions at the power distribution network engineering site, and combine all the multi-source heterogeneous information into a data set;

[0009] Establish the relationship between the data characteristics of the data acquired by each sensor during the movement of each working condition and the change of the robot's motion state;

[0010] Adjust the dynamic model through the relationship between the data characteristics and the change of the robot's motion state; Set the corresponding robot motion control algorithm according to each adjusted dynamic model. Each adjusted dynamic model and the corresponding robot motion control algorithm are the artificial neural network model;

[0011] Input the data set into the artificial neural network model, train the artificial neural network model, continuously optimize the network during the training process, and output the control decision and the normalization parameter;

[0012] Intelligently control the movement of the robot by adjusting the normalization parameter and the control decision in each dynamic model.

[0013] Preferably, the multi-source heterogeneous information includes:

[0014] The acceleration of each wheel axle of the robot's four-wheel differential chassis, collected by the inertial sensor;

[0015] The angular velocity of each wheel axle of the robot's four-wheel differential chassis, collected by the inertial sensor;

[0016] The inertial value of the robot, collected by the inertial sensor;

[0017] The pose of the robot estimated by the wheel odometer of the robot, estimated by the wheel odometer;

[0018] The pressure data of each tire of the robot's four-wheel differential chassis, collected by the pressure sensor;

[0019] Visible light image data, collected by the visible light vision collector;

[0020] Infrared image data, collected by the infrared camera;

[0021] Robot position data, provided by GNSS;

[0022] Weather data, directly obtained through the Internet.

[0023] Preferably, it further includes preprocessing the multi-source heterogeneous information. The steps of preprocessing the multi-source heterogeneous information include:

[0024] Perform curve fitting processing on the acceleration, angular velocity of each wheel axle, the pressure data of each wheel tire and the inertial value of the robot by the interpolation method;

[0025] Process the visible light image data by downsampling, smoothing filtering, and generating an image pyramid in sequence;

[0026] Process the infrared image data by blind pixel compensation, median filtering, and histogram equalization in sequence.

[0027] Preferably, the specific steps of curve fitting by the interpolation method include:

[0028] Take n + 1 points to be fitted and set them to satisfy the function , specifically: , …, ;

[0029] By For Fit the function of the n + 1 - point part:

[0030]

[0031] Obtain n + 1 linear equations about , and by And Solve ; Substitute the solved Into ;

[0032] Verify and compare the error between the function And the actual value through the verification program. When the error meets the verification requirements, the function Is the fitting function of the function ; Otherwise, repeat the value - taking and fitting operations of the function .

[0033] Preferably, the steps of pre - processing the multi - source heterogeneous information further include:

[0034] Perform time registration on each group of multi - source heterogeneous information;

[0035] Perform spatial registration on each group of multi - source heterogeneous information.

[0036] Preferably, the specific steps of performing time registration on each group of multi - source heterogeneous information include:

[0037] Take the acquisition time of the visible light image data as the interpolation time;

[0038] Obtain the previous - frame data and the next - frame data of the interpolation time of the remaining motion data as the first - frame data and the second - frame data of the motion data respectively;

[0039] Compare the acquisition time of the first - frame data of each remaining motion data with the interpolation time;

[0040] Compare the acquisition time of the second-frame data of each remaining motion data with the interpolation time;

[0041] For a remaining motion data, if the acquisition times of both the first-frame data and the second-frame data are later than the interpolation time, end the time registration; if the acquisition time of the first-frame data is earlier than the interpolation time and the acquisition time of the second-frame data is later than the interpolation time, judge the time difference between the acquisition times of the first-frame data and the second-frame data. When the time difference is greater than the threshold, end the time registration, otherwise perform interpolation; if the acquisition times of both the first-frame data and the second-frame data are earlier than the interpolation time, replace the first-frame data with the second-frame data, and use the data of the frame after the replaced first-frame data as the second-frame data, and repeat the comparison with the interpolation time;

[0042] Among them, the steps of the interpolation include:

[0043] Calculate the value to be inserted through the following formula ;

[0044]

[0045] Insert the value to be inserted Insert;

[0046] Among them, is the value of the first-frame data of a motion data; is the value of the second-frame data of a motion data; is the time corresponding to the first-frame data of a motion data; is the time corresponding to the second-frame data of a motion data; is the interpolation time.

[0047] Preferably, the specific steps of performing spatial registration on each group of multi-source heterogeneous information include:

[0048] Analyze the GNSS satellite signal, convert the standard GPGGA statement information into longitude and latitude information and satellite number information, and obtain the robot coordinates through the longitude and latitude information and the satellite number information;

[0049] Establish a reference coordinate system based on the robot coordinates;

[0050] Taking the reference coordinate system as the standard, convert the fixed coordinate systems of multiple sensors to determine the tf transformation relationship between the multiple sensors and the reference coordinate system.

[0051] Preferably, the steps of establishing the connection between the data characteristics obtained by each sensor during the motion process of each working condition and the change of the robot motion state include:

[0052] Integrate and double integrate the acceleration of each wheel axle of the robot's four-wheel differential chassis respectively to obtain the rotational speed and moving distance of each wheel within a period of time;

[0053] Integrate the angular velocity of each wheel axle of the robot's four-wheel differential chassis to obtain the rotation angle of each wheel within a period of time;

[0054] Using the rotation angles of each wheel and the rotational speed and moving distance of each wheel within a period of time, apply the four-wheel differential dynamics model to calculate the linear velocity and angular velocity of the robot's four-wheel differential chassis within a period of time;

[0055] Estimate the pose of the robot estimated by the inertial sensor through the linear velocity and angular velocity of the robot's four-wheel differential chassis within a period of time;

[0056] Combining the pose of the robot estimated by the inertial sensor, the pose of the robot estimated by the wheel odometer, and the pressure data of each wheel tire of the robot's four-wheel differential chassis, judge the road surface paving condition and road cleanliness of the robot at the power distribution project site.

[0057] Preferably, the steps of adjusting the power model through the relationship between data characteristics and changes in the robot's motion state include:

[0058] Integrate and double integrate the acceleration of each wheel axle of the robot's four-wheel differential chassis respectively to obtain the rotational speed and moving distance of each wheel within a period of time;

[0059] Integrate the angular velocity of each wheel axle of the robot's four-wheel differential chassis to obtain the rotation angle of each wheel within a period of time;

[0060] Using the rotation angles of each wheel and the rotational speed and moving distance of each wheel within a period of time, apply the four-wheel differential dynamics model to calculate the linear velocity and angular velocity of the robot's four-wheel differential chassis within a period of time;

[0061] Establish the power model of the robot's four-wheel differential chassis through the following formula,

[0062]

[0063] where, is the linear velocity of the robot's four-wheel differential chassis, is the angular velocity of the robot's four-wheel differential chassis, is the wheelbase between the wheels of the robot's four-wheel differential chassis, represents the motion state of the left front wheel of the robot's four-wheel differential chassis, represents the motion state of the right front wheel of the robot's four-wheel differential chassis, represents the motion state of the left rear wheel of the robot's four-wheel differential chassis, Indicates the motion state of the right rear wheel of the robot's four-wheel differential chassis;

[0064] Derivation of the inertial sensor's inertial numerical value to calculate the vehicle's angular velocity ;

[0065] The dynamic analysis model is processed by using the variable friction coefficient through the following formula:

[0066]

[0067] in,

[0068]

[0069] The corresponding robot motion control algorithm is set according to each adjusted power model.

[0070] Preferably, the specific steps of intelligently controlling the movement of the robot include:

[0071] Normalize the data set;

[0072] Inputting the normalized data set into a preset artificial neural network model to train the artificial neural network model;

[0073] Lightweight the trained artificial neural network model;

[0074] The processed artificial neural network model is transplanted to the robot data processing core, and motion experiments are carried out in the distribution network project site environment. The normalization parameters and control decisions in the artificial neural network model are adjusted through the experimental data.

[0075] The intelligent motion control method of a four-wheel differential robot provided by the present invention has the following beneficial effects: the present application uses multi-sensor information fusion technology to comprehensively judge the IMU, various cameras, pressure sensors, GNSS satellite signals and real-time regional weather forecast information collected in the distribution network engineering environment, and obtains the quantitative indicators of the road paving conditions, construction site conditions and weather conditions of the corresponding working environment, and finally selects the best motion control strategy to make the mobile robot chassis adapt to the real-time distribution network engineering working environment, creating prerequisites for the application of mobile robots in such working environments. The present application is based on the changes in the specific working environment of the four-wheel differential chassis of the mobile robot in the distribution network engineering, and intelligently selects a motion control algorithm that is adapted to it, which solves the problems of mechanical loss, poor energy efficiency and inability to adapt to complex weather conditions to a certain extent, and paves the way for the application of mobile robots based on four-wheel differential motion systems in more fields. BRIEF DESCRIPTION OF THE DRAWINGS

[0076] To more clearly illustrate the embodiments of the present invention and their design solutions, the accompanying drawings required for this embodiment will be briefly introduced below. The drawings in the following description are only partial embodiments of the present invention. For those of ordinary skill in the art, other drawings can be obtained based on these drawings without creative efforts.

[0077] Figure 1 It is a flowchart of an intelligent motion control method for a four-wheel differential robot in Embodiment 1 of the present invention;

[0078] Figure 2 It is a flowchart of time registration for each group of multi-source heterogeneous information in Embodiment 1 of the present invention. Detailed implementation manners

[0079] In order to enable those skilled in the art to better understand the technical solutions of the present invention and implement them, the present invention will be described in detail below with reference to the accompanying drawings and specific embodiments. The following embodiments are only used to more clearly illustrate the technical solutions of the present invention and cannot be used to limit the protection scope of the present invention.

[0080] Embodiment 1

[0081] The present invention provides an intelligent motion control method for a four-wheel differential robot. Refer to Figure 1 , including the following steps: obtaining multi-source heterogeneous information on the motion states of the multi-sensors of the robot under various working conditions at the power distribution network construction site, and combining all the multi-source heterogeneous information into a data set; establishing the relationship between the data characteristics of the data acquired by each sensor during the motion process of each working condition and the change in the motion state of the robot; adjusting the dynamic model through the relationship between the data characteristics and the change in the motion state of the robot; setting the corresponding robot motion control algorithms according to the adjusted dynamic models, and the adjusted dynamic models and the corresponding robot motion control algorithms are artificial neural network models; inputting the data set into the artificial neural network model, training the artificial neural network model, continuously optimizing the network during the training process, and outputting control decisions and normalization parameters; and intelligently controlling the motion of the robot by adjusting the normalization parameters and control decisions in each dynamic model.

[0082] In this embodiment, the multi-source heterogeneous information includes: the acceleration of each wheel axle of the four-wheel differential chassis of the robot, collected by an inertial sensor; the angular velocity of each wheel axle of the four-wheel differential chassis of the robot, collected by an inertial sensor; the inertial value of the robot, collected by an inertial sensor; the pose of the robot estimated by the wheel odometer of the robot, obtained by wheel odometer estimation; the pressure data of each tire of the four-wheel differential chassis of the robot, collected by a pressure sensor; visible light image data, collected by a visible light vision collector; infrared image data, collected by an infrared camera; the position data of the robot, provided by GNSS; and weather data, directly obtained through the Internet.

[0083] In this embodiment, it further includes preprocessing multi-source heterogeneous information. Specifically, the steps of preprocessing multi-source heterogeneous information include: performing curve fitting on the acceleration, angular velocity of each wheel axle, the pressure data of each wheel tire, and the inertial value of the robot through the interpolation method; the images directly collected by various vision sensors are too prone to noise points. This is mainly because in the early design, when the CMOS processes rapidly changing images, overheating will occur due to too frequent current changes. At the same resolution, the price of CMOS is cheaper than that of CCD, but the image quality generated by CMOS devices is lower than that of CCD. The visible light image data is processed in turn by means of downsampling, smoothing filtering, and generating an image pyramid; the infrared image data is processed in turn by means of blind pixel compensation, median filtering, and histogram equalization.

[0084] For inertial sensors, pressure sensors, wheel odometers, etc., during the data acquisition process, the collected data will be distorted due to errors such as insertion error, application error, characteristic error, dynamic error, and environmental error. Therefore, it is necessary to perform filtering processing on them. Using the interpolation method for curve fitting has a good effect in processing the above sensor data. The interpolation method for curve fitting yields a continuous curve, making the points on the curve conform to the actual situation as much as possible. Using the interpolation method can better solve this problem. The so-called interpolation method is to obtain a function from several measured points so that these points are all on the function. There are Lagrange interpolation method, Newton interpolation method, spline interpolation method, etc. The characteristics of the above sensors are basically within the entire measurement range, and a low-order polynomial can be used for fitting. In this embodiment, the specific steps of performing curve fitting through the interpolation method include:

[0085] Taking n + 1 points to be fitted and setting them to satisfy the function , specifically: , … .

[0086] Through to fit the function for the part of the n + 1 points:

[0087]

[0088] Obtaining n + 1 linear equations about , and through and solving ; substituting the solved into .

[0089] Verifying and comparing the function through the verification program The error from the actual value. If the error meets the verification requirements, then this function is the fitting function ; otherwise, repeat the value-taking and fitting operations for the function .

[0090] During the fitting process, determine the degree of the polynomial according to the required approximation accuracy. At the same time, the selection of interpolation nodes is highly related to the error magnitude of the polynomial. Increase the value density at locations with high non-linearity.

[0091] Refer to Figure 2 , and the steps for preprocessing multi-source heterogeneous information further include: performing time registration on each group of multi-source heterogeneous information, and performing spatial registration on each group of multi-source heterogeneous information.

[0092] The specific steps for performing time registration on each group of multi-source heterogeneous information include: taking the acquisition time of the visible light image data as the interpolation time; obtaining the previous frame data and the subsequent frame data of the interpolation time of the remaining motion data respectively as the first frame data and the second frame data of this motion data; comparing the acquisition time of the first frame data of each remaining motion data with the interpolation time; comparing the acquisition time of the second frame data of each remaining motion data with the interpolation time; for a remaining motion data, if the acquisition times of both the first frame data and the second frame data are later than the interpolation time, then end the time registration; if the acquisition time of the first frame data is earlier than the interpolation time and the acquisition time of the second frame data is later than the interpolation time, then judge the time difference between the acquisition times of the first frame data and the second frame data. When the time difference is greater than the threshold, end the time registration, otherwise perform interpolation; if the acquisition times of both the first frame data and the second frame data are earlier than the interpolation time, replace the first frame data with the second frame data, and take the data of the frame subsequent to the replaced first frame data as the second frame data, and repeat the comparison with the interpolation time.

[0093] Among them, the steps for interpolation include: calculating the value to be inserted through the following formula ;

[0094]

[0095] Insert the value to be inserted ; where is the value of the first frame data of a motion data; is the value of the second frame data of a motion data; is the time corresponding to the first frame data of a motion data; is the time corresponding to the second frame data of a motion data; is the interpolation time.

[0096] In this embodiment, the specific steps for spatially registering each group of multi-source heterogeneous information include: parsing GNSS satellite signals, converting the standard GPGGA statement information into longitude and latitude information and satellite number information, and obtaining the robot coordinates through the longitude and latitude information and satellite number information; establishing a reference coordinate system based on the robot coordinates; taking the reference coordinate system as the standard, converting the fixed coordinate systems of multiple sensors, and determining the tf transformation relationship between the multiple sensors and the reference coordinate system.

[0097] Specifically, the steps for establishing the connection between the data characteristics obtained by each sensor during the movement of each working condition and the change in the robot's motion state include: respectively integrating and double-integrating the acceleration of each wheel axle of the robot's four-wheel differential chassis to obtain the rotational speed and moving distance of each wheel within a period of time; integrating the angular velocity of each wheel axle of the robot's four-wheel differential chassis to obtain the rotation angle of each wheel within a period of time; using the rotation angles of each wheel and the rotational speed and moving distance of each wheel within a period of time, applying the four-wheel differential dynamics model to calculate the linear velocity and angular velocity of the robot's four-wheel differential chassis within a period of time; estimating the pose of the robot estimated by the inertial sensor through the linear velocity and angular velocity of the robot's four-wheel differential chassis within a period of time; combining the pose of the robot estimated by the inertial sensor, the pose of the robot estimated by the wheel odometer, and the pressure data of each tire of each wheel of the robot's four-wheel differential chassis to determine the road surface paving condition and road cleanliness at the on-site distribution network project where the robot is located.

[0098] In this embodiment, the steps for adjusting the power model through the connection between the sensor working characteristics and the change in the robot's motion state include: respectively integrating and double-integrating the acceleration of each wheel axle of the robot's four-wheel differential chassis to obtain the rotational speed and moving distance of each wheel within a period of time; integrating the angular velocity of each wheel axle of the robot's four-wheel differential chassis to obtain the rotation angle of each wheel within a period of time; using the rotation angles of each wheel and the rotational speed and moving distance of each wheel within a period of time, applying the four-wheel differential dynamics model to calculate the linear velocity and angular velocity of the robot's four-wheel differential chassis within a period of time; establishing the power model of the robot's four-wheel differential chassis through the following formula,

[0099]

[0100] where, is the linear velocity of the robot's four-wheel differential chassis, is the angular velocity of the robot's four-wheel differential chassis, is the wheelbase between the wheels of the robot's four-wheel differential chassis, represents the motion state of the left front wheel of the robot's four-wheel differential chassis, represents the motion state of the right front wheel of the robot's four-wheel differential chassis, Indicates the motion state of the left rear wheel of the robot's four-wheel differential chassis, Indicates the motion state of the right rear wheel of the robot's four-wheel differential chassis; the angular velocity of the vehicle body is calculated by taking the derivative of the inertial value of the inertial sensor ; Since there are various road conditions at the distribution network project site, a variable friction coefficient is used to process the dynamic analysis model through the following formula,

[0101]

[0102] wherein,

[0103]

[0104] According to each adjusted dynamic model, a corresponding robot motion control algorithm is set.

[0105] Specifically, the specific steps for intelligent control of the robot's motion include: normalizing the data set; inputting the normalized data set into a preset artificial neural network model to train the artificial neural network model; lightweight processing the trained artificial neural network model; transplanting the processed artificial neural network model to the robot data processing core, and conducting motion experiments respectively in the distribution network project site environment, and adjusting the normalization parameters and control decisions in the artificial neural network model through the experimental data.

[0106] The above embodiments are only preferred specific embodiments of the present invention, and the protection scope of the present invention is not limited thereto. Any simple changes or equivalent replacements of the technical solutions that can be obviously obtained by those skilled in the art within the technical scope disclosed by the present invention all belong to the protection scope of the present invention.

Claims

1. An intelligent motion control method for a four-wheel differential robot, characterized in that, it includes the following steps: Obtain multi-source heterogeneous information on the motion states of the robot's multi-sensors under various working conditions at the power distribution network engineering site, and combine all the multi-source heterogeneous information into a data set; Establish the relationship between the data characteristics obtained by each sensor during the motion process of each working condition and the change in the robot's motion state; Adjust the dynamic model through the relationship between the data characteristics and the change in the robot's motion state; Set the corresponding robot motion control algorithm according to each adjusted dynamic model, and each adjusted dynamic model and the corresponding robot motion control algorithm are artificial neural network models; Input the data set into the artificial neural network model, train the artificial neural network model, continuously optimize the network during the training process, and output control decisions and normalization parameters; Intelligently control the motion of the robot by adjusting the normalization parameters and control decisions in each dynamic model.

2. The intelligent motion control method for a four-wheel differential robot according to claim 1, characterized in that, the multi-source heterogeneous information includes: The acceleration of each wheel axle of the robot's four-wheel differential chassis, collected by an inertial sensor; The angular velocity of each wheel axle of the robot's four-wheel differential chassis, collected by an inertial sensor; The inertial value of the robot, collected by an inertial sensor; The pose of the robot estimated by the wheel odometer of the robot, obtained by estimation with the wheel odometer; The pressure data of each tire of the robot's four-wheel differential chassis, collected by a pressure sensor; Visible light image data, collected by a visible light vision collector; Infrared image data, collected by an infrared camera; Robot position data, provided by GNSS; Weather data, directly obtained through the Internet.

3. The intelligent motion control method for a four-wheel differential robot according to claim 2, characterized in that, it further includes preprocessing the multi-source heterogeneous information, and the steps of preprocessing the multi-source heterogeneous information include: Perform curve fitting on the acceleration, angular velocity of each wheel axle, the pressure data of each tire, and the inertial value of the robot through the interpolation method; Process the visible light image data successively by downsampling, smoothing filtering, and generating an image pyramid; Process the infrared image data successively by blind pixel compensation, median filtering, and histogram equalization.

4. The intelligent motion control method for a four-wheel differential robot according to claim 3, characterized in that, The specific steps of performing curve fitting through the interpolation method include: Take n + 1 points to be fitted and set them to satisfy the function , specifically: , …, ; By fitting the function at n + 1 point parts: Obtain n + 1 linear equations regarding through and Solve ; Substitute the solved into ; Verify the comparison function through the verification program and the error from the actual value. When the error meets the verification requirements, the function is the fitting function ; otherwise, repeat the value-taking and fitting operations for the function .

5. The intelligent motion control method for a four-wheel differential robot according to claim 3, characterized in that, The steps of preprocessing the multi-source heterogeneous information further include: Perform time registration on each group of multi-source heterogeneous information; Perform spatial registration on each group of multi-source heterogeneous information.

6. The intelligent motion control method for a four-wheel differential robot according to claim 5, characterized in that, The specific steps of performing time registration on each group of multi-source heterogeneous information include: Take the acquisition time of the visible light image data as the interpolation time; Obtain the previous frame data and the next frame data of the interpolation time of the remaining motion data respectively as the first frame data and the second frame data of the motion data; Compare the acquisition time of the first frame data of each remaining motion data with the interpolation time; Compare the acquisition time of the second frame data of each remaining motion data with the interpolation time; For a remaining motion data, if the acquisition times of both the first frame data and the second frame data are later than the interpolation time, end the time registration; if the acquisition time of the first frame data is earlier than the interpolation time and the acquisition time of the second frame data is later than the interpolation time, judge the time difference between the acquisition times of the first frame data and the second frame data. When the time difference is greater than the threshold, end the time registration, otherwise interpolate; if the acquisition times of both the first frame data and the second frame data are earlier than the interpolation time, replace the first frame data with the second frame data, and use the data of the frame after the replaced first frame data as the second frame data, and repeat the comparison with the interpolation time; Among them, the steps of the interpolation include: Calculate the value to be inserted by the following formula ; The value to be inserted Insert; wherein, is the value of the first frame data of a motion data; is the value of the second frame data of a motion data; is the time corresponding to the first frame data of a motion data; is the time corresponding to the second frame data of a motion data; is the interpolation time.

7. According to an intelligent motion control method for a four-wheel differential robot as described in claim 5, characterized in that, the specific steps of performing spatial registration on each group of multi-source heterogeneous information include: Analyze the GNSS satellite signal, convert the standard GPGGA statement information into longitude and latitude information and satellite number information, and obtain the robot coordinates through the longitude and latitude information and the satellite number information; Establish a reference coordinate system based on the robot coordinates; Taking the reference coordinate system as the standard, convert the fixed coordinate systems of multiple sensors to determine the tf transformation relationship between the multiple sensors and the reference coordinate system.

8. According to an intelligent motion control method for a four-wheel differential robot as described in claim 2, characterized in that, the steps of establishing the connection between the data characteristics obtained by each sensor during the motion process of each working condition and the change of the robot motion state include: Integrate and double-integrate the acceleration of each wheel axle of the four-wheel differential chassis of the robot respectively to obtain the rotational speed and moving distance of each wheel within a period of time; Integrate the angular velocity of each wheel axle of the four-wheel differential chassis of the robot to obtain the rotation angle of each wheel within a period of time; Utilize the rotation angles of each wheel and the rotational speed and moving distance of each wheel within a period of time, and apply the four-wheel differential dynamics model to calculate the linear velocity and angular velocity of the four-wheel differential chassis of the robot within a period of time; Estimate the pose of the robot estimated by the inertial sensor through the linear velocity and angular velocity of the four-wheel differential chassis of the robot within a period of time; Combined with the pose of the robot estimated by the inertial sensor, the pose of the robot estimated by the wheel odometer, and the pressure data of each tire of the four-wheel differential chassis of the robot, judge the road surface paving condition and road cleanliness of the power distribution project site where the robot is located.

9. According to an intelligent motion control method for a four-wheel differential robot as described in claim 2, characterized in that, the steps of adjusting the power model through the connection between the data characteristics and the change of the robot motion state include: The acceleration of each wheel axle of the robot's four-wheel differential chassis is integrated and quadratically integrated to obtain the rotation speed and moving distance of each wheel in a time period; Integrate the angular velocity of each wheel axle of the robot's four-wheel differential chassis to obtain the rotation angle of each wheel within a time period; The four-wheel differential dynamics model is applied to calculate the linear velocity and angular velocity of the robot's four-wheel differential chassis within a time period using the rotation angle of each wheel and the rotation speed and movement distance of each wheel within a time period. The dynamic model of the robot's four-wheel differential chassis is established by the following formula: Among them, is the linear velocity of the four-wheel differential chassis of the robot, is the angular velocity of the four-wheel differential chassis of the robot, is the wheelbase between the wheels of the four-wheel differential chassis of the robot, represents the motion state of the left front wheel of the four-wheel differential chassis of the robot, represents the motion state of the right front wheel of the four-wheel differential chassis of the robot, represents the motion state of the left rear wheel of the four-wheel differential chassis of the robot, represents the motion state of the right rear wheel of the four-wheel differential chassis of the robot; Derive the inertial values of the inertial sensor to calculate the vehicle body angular velocity ; The dynamic analysis model is processed by using the variable friction coefficient through the following formula: in, The corresponding robot motion control algorithm is set according to each adjusted power model.

10. The intelligent motion control method of a four-wheel differential robot according to claim 2, It is characterized in that The specific steps for intelligently controlling the robot's motion include: Normalize the data set; Inputting the normalized data set into a preset artificial neural network model to train the artificial neural network model; Lightweight the trained artificial neural network model; The processed artificial neural network model is transplanted to the robot data processing core, and motion experiments are carried out in the distribution network project site environment. The normalization parameters and control decisions in the artificial neural network model are adjusted through the experimental data.

Images