Wheel-foot cooperative control method and system for wheel-foot robot

Abstract

The invention discloses a wheel-foot coordination control method and system for a wheel-foot robot, and particularly relates to the technical field of wheel-foot robots. The wheel-foot robot is divided into a plurality of monitoring areas, temperature distribution and motion load data of all the areas are obtained in real time through temperature sensors and acceleration sensors installed in a wheel type motion control system and a foot type motion control system, the data are preprocessed and analyzed, and load abnormal areas are determined; for a monitoring area with abnormal load, a fuzzy logic algorithm is applied, load distribution is adjusted in real time and energy utilization efficiency is optimized according to the abnormal degree and the movement speed of the wheel foot load, on the basis, prediction is carried out for the adjusted load state, a movement control strategy is adjusted in advance according to a prediction result, and the stability and safety of the system are ensured. Through intelligent monitoring and dynamic adjustment, the problems of uneven energy consumption and local overheating under wheel-foot separation control are solved, the service life of the system is prolonged, and the operation efficiency is improved.

Description

Technical Field

[0001] The present invention relates to the technical field of wheel-legged robots, and in particular to a wheel-legged robot wheel-legged coordinated control method and system. Background Art

[0002] With the rapid development of robotics technology, wheeled robots and legged robots, as common types of mobile robots, have been widely used in different fields. Wheeled robots are suitable for fast movement on flat ground due to their simple structure, high motion efficiency and low energy consumption; legged robots have stronger terrain adaptability, can maintain stability in rugged environments, and are suitable for movement on complex terrain. Wheeled-legged robots combine the advantages of wheeled and legged mobility, and can move at high speed on flat ground and walk flexibly on complex terrain. Although wheeled-legged robots have significant advantages in terrain adaptability and motion efficiency, their control difficulty also increases accordingly. In particular, how to achieve wheel-legged coordinated control to ensure smooth transition and efficient operation of the robot in different modes has become the focus and difficulty of current research.

[0003] Most of the existing wheel-legged robot control methods adopt the method of controlling wheeled motion and legged motion separately. This method usually regards wheeled and legged motion as independent modules, and applies different control algorithms to handle them respectively. However, the control requirements of wheeled and legged motion are different, which may lead to uneven energy consumption and heat dissipation. During long-term operation, if the system frequently switches motion modes and the thermal management system is not designed carefully, the wheel-legged separation control may cause local overheating of a system. For example, when the wheeled motion is running at high speed, the motor heats up seriously, but if it is switched to legged motion immediately afterwards, the servo motor of the legged system will also start to operate at high load, resulting in failure of the overall thermal management. Ultimately, this may cause overheating of the motor or circuit, and even cause a fire or permanent damage to the system. Summary of the invention

[0004] The purpose of the present invention is to provide a wheel-foot coordinated control method and system for a wheel-foot robot to overcome the deficiencies in the background technology.

[0005] In order to achieve the above object, the present invention provides the following technical solution: a wheel-foot coordinated control method of a wheel-foot robot, comprising the following steps:

[0006] S1: Divide the wheeled-legged robot into several monitoring areas, and obtain the temperature distribution data and motion load data in each monitoring area through the temperature sensors and acceleration sensors installed in the wheeled motion control system and the legged motion control system;

[0007] S2: preprocessing the monitored temperature distribution data and motion load data, analyzing the preprocessed temperature distribution data and motion load data, and determining a monitoring area of ​​abnormal wheel foot load according to the analysis results;

[0008] S3: For the monitoring area with abnormal wheel-foot load, the load distribution of the wheel-foot system is adjusted in real time according to the abnormal degree of wheel-foot load and movement speed through fuzzy logic algorithm to optimize the system energy utilization rate;

[0009] S4: Predict the wheel foot load change in the wheel foot load abnormality monitoring area within the adjusted fixed time period, and adjust the motion control strategy in advance according to the prediction result to prevent local overheating.

[0010] Preferably, in S2, the temperature distribution anomaly index is generated after analyzing the preprocessed temperature distribution data, and the method for obtaining the temperature distribution anomaly index is:

[0011] Set the prior distribution P(θ). If the temperature data conforms to the normal distribution, use the prior normal distribution to represent it. where μ is the initial mean temperature, σ 2 is the initial variance of the temperature; for a new temperature data point T, the likelihood function is defined according to its actual distribution, that is, the probability of observing the current temperature data. The likelihood function P(T|θ) is the probability density function based on the new temperature data T, and the expression is: According to Bayes' theorem, the posterior distribution P(θ|T) combines the prior distribution and the likelihood function of the current observation data, expressed as: P(θ|T) is the posterior distribution, that is, the probability of the temperature distribution parameter θ after the data T is observed. P(θ) is the prior distribution, which represents the temperature distribution before the observation. P(T) is the marginal likelihood of the observed temperature. The temperature distribution anomaly index is calculated as follows: Where KM is the temperature distribution anomaly index, [T min ,T max ] indicates the temperature threshold range.

[0012] Preferably, in S2, the motion load drift index is generated after analyzing the preprocessed motion load data, and the method for obtaining the motion load drift index is:

[0013] Real-time acquisition of motion load data within the Q time period is expressed as L1, L2, ..., L n ; where each L n is the load data at time n, and defines the distance metric between load data points to measure the difference between two load data points. The expression is: Among them, L i,k and Lj,k Respectively represent the load data L i and L j The kth eigenvalue of , m is the feature dimension. In the load data space, the neighborhood radius of a point is defined. If the distance between two points is less than or equal to ∈, they are regarded as neighbors of each other. Within a given radius ∈, at least MinPts neighbors are required to regard the data point as a core point and classify it into a cluster. The load data is divided into multiple clusters. Points that cannot be classified into any cluster are marked as noise points, that is, abnormal load points; clustering is performed based on density. The local density is calculated by the number of neighbors within the radius ∈. The specific formula is: Among them, |N ∈ (L i )| represents the number of data points within the radius ∈, V ∈ Represents the neighborhood volume with ∈ as radius. For the two-dimensional space of Euclidean distance, the neighborhood volume is V ∈ =π∈ 2 ; Calculate the motion load drift index, the expression is: In the formula, ρ(L i ) is the data point L i The local density, ρ mean (C) is point L i The average local density of cluster C, SD is the motion load drift index.

[0014] Preferably, in S2, the temperature distribution anomaly index and the motion load drift index are converted into a first eigenvector, and the first eigenvector is used as the input of the machine learning model. The machine learning model predicts the wheel load anomaly degree value label of each monitoring area with each group of first eigenvectors as the prediction target, and minimizes the sum of prediction errors of the wheel load anomaly degree value labels of all monitoring areas as the training target. The machine learning model is trained until the sum of prediction errors converges, and the model training is stopped. The wheel load anomaly degree value of each monitoring area is determined according to the model output results, wherein the machine learning model is a polynomial regression model.

[0015] Preferably, the acquired wheel load abnormality degree value of each monitoring area is compared with a pre-set reference threshold value of the wheel load abnormality degree value. If the wheel load abnormality degree value is greater than or equal to the reference threshold value of the wheel load abnormality degree value, it means that the wheel load abnormality degree in the monitoring area is high. At this time, a warning signal is generated and it is marked as a wheel load abnormality monitoring area; if the wheel load abnormality degree value is less than the reference threshold value of the wheel load abnormality degree value, it means that the wheel load abnormality degree in the monitoring area is low. At this time, no warning signal is generated, and it is marked as a normal wheel load monitoring area.

[0016] Preferably, in S3, for the monitoring area where the wheel foot load is abnormal, the load distribution of the wheel foot system is adjusted in real time according to the abnormal degree of the wheel foot load and the movement speed through a fuzzy logic algorithm to optimize the system energy utilization rate, specifically:

[0017] The wheel-foot load abnormality value CP and the wheel-foot movement speed E are used as the input items of fuzzy logic, and the load distribution H of the wheel-foot system is used as the output item of fuzzy logic.

[0018] Fuzzify the input and output variables, that is, convert these precise values ​​into fuzzy sets;

[0019] Construct fuzzy rules to generate corresponding outputs based on the input fuzzy sets and rule base;

[0020] The result of fuzzy reasoning is a fuzzy set, which is converted into an accurate output value, namely defuzzification, and the centroid of the fuzzy output is calculated as the final load distribution value H;

[0021] Through the centroid method, a specific load distribution value H is obtained, which represents the load distributed by the system under current conditions.

[0022] The present invention also provides a wheel-foot robot wheel-foot coordinated control system, comprising a data acquisition module, a data analysis module, a load adjustment module and a motion control module;

[0023] Data acquisition module: The wheeled-legged robot is divided into several monitoring areas. The temperature distribution data and motion load data in each monitoring area are obtained through the temperature sensors and acceleration sensors installed in the wheeled motion control system and the legged motion control system.

[0024] Data analysis module: pre-processes the monitored temperature distribution data and motion load data, analyzes the pre-processed temperature distribution data and motion load data, and determines the monitoring area of ​​abnormal wheel foot load according to the analysis results;

[0025] Load adjustment module: For monitoring areas with abnormal wheel-foot load, the load distribution of the wheel-foot system is adjusted in real time according to the abnormal degree of wheel-foot load and movement speed through fuzzy logic algorithm to optimize the system energy utilization rate;

[0026] Motion control module: predict the wheel load changes in the wheel load abnormality monitoring area within the adjusted fixed time period, and adjust the motion control strategy in advance according to the prediction results to prevent local overheating.

[0027] In the above technical solution, the technical effects and advantages provided by the present invention are:

[0028] 1. The present invention solves the problems of uneven energy consumption and local overheating in the traditional wheel-foot separation control method by performing intelligent collaborative control on the wheel-foot robot. By dividing the robot into multiple monitoring areas and collecting the temperature and motion load data of each area in real time, the system can accurately monitor the working status of each area. On this basis, the temperature distribution anomaly index and motion load drift index are used to deeply analyze the data, and the machine learning model is used to identify the areas with abnormal load. This enables the system to detect potential abnormal points in a timely manner and provides an accurate basis for subsequent real-time adjustments.

[0029] 2. The present invention adopts a fuzzy logic algorithm for the identified load abnormality area, and dynamically adjusts the load distribution of the wheel-foot system according to the real-time load abnormality degree and movement speed. This adaptive control method enables the system to balance energy consumption under different working conditions, effectively avoiding local overheating and overload. After the load is adjusted, the system can also predict future load changes, formulate and adjust the motion control strategy in advance, so as to further prevent possible overheating risks, which not only improves the energy utilization efficiency of the wheel-foot robot, but also significantly enhances the safety and stability of the system operation. BRIEF DESCRIPTION OF THE DRAWINGS

[0030] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the drawings required for use in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments recorded in the present invention. For ordinary technicians in this field, other drawings can also be obtained based on these drawings.

[0031] Figure 1 The figure is a flow chart of the method of the present invention.

[0032] Figure 2 It is a system module diagram of the present invention. DETAILED DESCRIPTION

[0033] In order to make the purpose, technical solution and advantages of the embodiments of the present invention clearer, the technical solution in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0034] Example 1, please refer to Figure 1 and Figure 2 As shown, the wheel-foot coordinated control method of a wheel-foot robot described in this embodiment includes the following steps:

[0035] S1: Divide the wheeled-legged robot into several monitoring areas, and obtain the temperature distribution data and motion load data in each monitoring area through the temperature sensors and acceleration sensors installed in the wheeled motion control system and the legged motion control system;

[0036] S2: preprocessing the monitored temperature distribution data and motion load data, analyzing the preprocessed temperature distribution data and motion load data, and determining a monitoring area of ​​abnormal wheel foot load according to the analysis results;

[0037] S3: For the monitoring area with abnormal wheel-foot load, the load distribution of the wheel-foot system is adjusted in real time according to the abnormal degree of wheel-foot load and movement speed through fuzzy logic algorithm to optimize the system energy utilization rate;

[0038] S4: Predict the wheel foot load change in the wheel foot load abnormality monitoring area within the adjusted fixed time period, and adjust the motion control strategy in advance according to the prediction result to prevent local overheating.

[0039] Among them, in S1, the wheeled-legged robot is divided into several monitoring areas, and the temperature distribution data and motion load data in each monitoring area are obtained respectively through the temperature sensors and acceleration sensors installed in the wheeled motion control system and the footed motion control system, specifically:

[0040] First, according to the physical structure of the wheeled-legged robot, it is divided into several functional areas. A typical wheeled-legged robot is mainly composed of a wheeled motion control system, a footed motion control system, a central control unit, a power transmission module, etc. Therefore, the monitoring area can be divided according to these key components.

[0041] For wheeled systems, there are usually multiple wheels or motor drives. Each motor can be considered an independent monitoring area. The specific divisions are:

[0042] Wheel motor area: The motor of each wheel is a key monitoring point and temperature sensors and power sensors need to be installed.

[0043] Axles and power transmission areas: Wheels are connected via axles or transmission mechanisms. These transmission components may also be affected by temperature and load, so these areas can be included in monitoring.

[0044] Wheel motor control module area: This area is where the motor controller is located, and its energy consumption and heat dissipation need to be monitored.

[0045] Divide the monitoring area of ​​foot movement Foot movement usually consists of multiple joints, and the motor and transmission components of each joint may generate heat and bear the movement load. Divide into the following areas:

[0046] Foot joint area: The motor and servo drive corresponding to each foot joint are key monitoring points, and temperature and acceleration sensors are installed to monitor motion loads.

[0047] Leg structure area: The connection part of the leg will also be subjected to stress, especially when the load of foot movement is large, and the temperature changes of the transmission parts need to be monitored.

[0048] Foot Motor Control Module Area: The control unit used to drive the foot joints needs to monitor its energy consumption and temperature.

[0049] In addition to the wheel-foot motion system, the central control unit and the thermal management module must also be monitored. These areas are usually responsible for managing the coordination and heat distribution of the entire system.

[0050] Central processing unit area: Monitor the temperature and load of the main controller, as it generates heat over long periods of time and is critical to system stability.

[0051] Thermal management module area: includes cooling fans, batteries, heat sinks, etc., used to monitor thermal efficiency.

[0052] Install temperature sensors Install temperature sensors on the motors, motor control modules, and joints in each monitoring area to monitor temperature data in real time. Specific sensor installation points include:

[0053] Temperature sensors are installed directly on the wheel motors and foot joint motors to monitor the operating temperature of the motors. Temperature sensors are installed on key circuit boards of each control module to prevent the control unit from overheating. These sensors send the collected temperature data to the central control unit at a fixed frequency (such as once per second), and the central system will store the data for further analysis.

[0054] Installing Acceleration Sensors Acceleration sensors are installed at the joints of the robot to obtain the motion load data in each monitoring area. The specific operations include: installing acceleration sensors in each foot joint area to monitor the acceleration changes of the joints during movement, so as to infer the actual motion load. Acceleration sensors are installed in the axle area of ​​the wheels to monitor the vibration and load of the wheels during driving. The output data of the acceleration sensor can be compared with the planned speed of the robot's movement to analyze whether there is overload operation or abnormal vibration.

[0055] The temperature distribution data and motion load data obtained by the sensor will be synchronously transmitted to the central control system via wired or wireless means. The system will synchronously process the data according to the acquisition frequency of sensors in different areas. The data will be pre-processed and input into the advanced algorithm module, and analyzed through historical data, real-time status and thermal management model to identify the temperature rise trend or excessive load. If the temperature or load exceeds the standard in a certain monitoring area, the system will issue an alarm and implement corresponding adjustment strategies, such as reducing the workload in that area or increasing the work intensity in other areas. Data fusion and decision-making will conduct a comprehensive analysis of the sensor data in each area, and the system will determine whether the current heat dissipation and load distribution are reasonable. For example: If the temperature of multiple motors rises, it may indicate that the overall energy consumption is too large. The system can reduce high-speed movement or reduce the total load through algorithms. If the acceleration of a joint is obviously abnormal, the system can determine that the joint may be overloaded or damaged, and issue a maintenance reminder.

[0056] In this application, by dividing different monitoring areas, the system can monitor the temperature and motion load status of the wheeled robot in real time to ensure that the energy consumption and heat dissipation of each area are balanced. The combination of sensor data collection and advanced control algorithms enables the robot to intelligently deal with heat dissipation problems in complex environments and prevent local overheating or system failure.

[0057] S2: Preprocessing the monitored temperature distribution data and motion load data, analyzing the preprocessed temperature distribution data and motion load data, and determining a monitoring area of ​​abnormal wheel foot load according to the analysis results.

[0058] The motor temperature and motion load data may be affected by noise, sensor failure or external interference during the collection process, resulting in incomplete or erroneous data. In the case of sensor anomalies, the temperature data or motion load data may be invalid values ​​(such as zero, negative values, or exceeding the physical upper limit). These invalid data need to be removed to avoid affecting subsequent analysis. If some data is missing, interpolation methods (such as linear interpolation and spline interpolation) can be used to fill the missing values, or the missing data segments can be deleted directly.

[0059] Since the sensors in the wheeled motion control system and the footed motion control system may collect data at different sampling rates or different timestamps, the data synchronization of different sensors must be ensured during the preprocessing process to facilitate subsequent joint analysis.

[0060] Ensure that the data from all sensors have the same time base. Linear interpolation or nearest neighbor interpolation can be performed on the sampled data from different sensors to map all data onto the same timeline. If the sampling rates of the sensors are different, it is usually necessary to downsample the high-frequency sampled data to a lower frequency to ensure that the data from different sensors are aligned in time and to ensure consistency in subsequent analysis.

[0061] Since the motor temperature and motion load data may have different value ranges (for example, the temperature range is 0°C to 100°C, while the acceleration range may be 0 to 10m / s 2 ), in order to analyze the data on a unified scale, normalization is necessary.

[0062] After analyzing the preprocessed temperature distribution data, the temperature distribution anomaly index is generated. The method for obtaining the temperature distribution anomaly index is as follows:

[0063] Set the prior distribution P(θ), where θ represents the mean and variance of the temperature. If the temperature data conforms to the normal distribution, use the prior normal distribution to represent it. where μ is the initial mean temperature, σ 2 is the initial variance of the temperature; for a new temperature data point T, the likelihood function is defined according to its actual distribution, that is, the probability of observing the current temperature data. The likelihood function P(T|θ) is the probability density function based on the new temperature data T, and the expression is: According to Bayes' theorem, the posterior distribution P(θ|T) combines the prior distribution and the likelihood function of the current observation data, expressed as: P(θ|T) is the posterior distribution, that is, the probability of the temperature distribution parameter θ after the data T is observed. P(θ) is the prior distribution, which represents the temperature distribution before the observation. P(T) is the marginal likelihood of the observed temperature, which can be obtained by integrating all possible parameter values ​​θ. The temperature distribution anomaly index is calculated as follows: Where KM is the temperature distribution anomaly index, [T min ,T max ] indicates the temperature threshold range.

[0064] The larger the temperature distribution anomaly index is, the more the temperature rise in the monitored area has deviated significantly from the normal range, indicating that the area is under a heavy load. When a motor or device is under high load, the work intensity increases, more heat is generated, and the temperature rises rapidly. An increase in the anomaly index usually means that the area may be in an overloaded operating state, with untimely heat dissipation or excessive energy consumption. If the temperature continues to rise and is not effectively controlled, it may cause the equipment to overheat or even be damaged. Therefore, a larger anomaly index usually indicates that the load is excessive and needs to be adjusted or reduced in time to avoid failures.

[0065] The smaller the temperature distribution anomaly index is, the closer the temperature change in the monitored area is to or within the normal range, which means that the load of the equipment is at a normal or low level. When the motor or component is running at a standard load, the heat generated can be effectively discharged through the system's heat dissipation mechanism, and the temperature fluctuation is small. A lower anomaly index indicates that the area is running stably, the equipment load is light, and the heat dissipation is good. This usually means that the equipment is in a healthy working state, with no overload and no risk of abnormal temperature rise.

[0066] The motion load drift index is generated after analyzing the preprocessed motion load data. The method for obtaining the motion load drift index is:

[0067] Real-time acquisition of motion load data within the Q time period is expressed as L1, L2, ..., L n ; where each L n It is the load data at time n, including multiple dimensions (such as acceleration, torque, speed, etc.), and defines the distance metric between load data points. The most commonly used is the Euclidean distance, which is used to measure the difference between two load data points. The expression is: Among them, L i,k and L j,k Respectively represent the load data L i and L j The kth eigenvalue of , m is the feature dimension, and in the load data space, the neighborhood radius of a point is defined. If the distance between two points is less than or equal to ∈, they are considered to be neighbors of each other. Within a given radius ∈, at least MinPts neighbors are required to consider the data point as a core point and classify it into a cluster. Through the DBSCAN algorithm, the load data is divided into multiple clusters, and the points in the clusters represent load data with higher density and relatively concentrated distribution. Points that cannot be classified into any cluster are marked as noise points, that is, abnormal load points.

[0068] Core point: If a point L i If a point L has at least MinPts neighbors within the radius ∈, then the point is considered a core point and belongs to a cluster. i The number of neighbors of a point is less than MinPts, but it is located in the neighborhood of a core point ∈, it is considered a boundary point. Noise points: Points that do not meet the above conditions are considered noise points, that is, abnormal points. Clustering is based on density, so the local density of points in each cluster should be relatively uniform. The local density can be calculated by the number of neighbors within the ∈ radius. The specific formula is: Among them, |N ∈ (L i )| represents the number of data points within the radius ∈, V ∈represents the neighborhood volume with ∈ as radius. For the two-dimensional space of Euclidean distance, the neighborhood volume is V ∈ =π∈ 2 ; Calculate the motion load drift index, the expression is: In the formula, ρ(L i ) is the data point L i The local density, ρ mean (C) is point L i The average local density of cluster C, SD is the motion load drift index.

[0069] The larger the motion load drift index is, the greater the deviation of the load in the monitored area from the normal state, reflecting that the load in the area is abnormal or changes dramatically. A larger drift index usually means that the area is under too high a load, or that a significant load fluctuation occurs during operation, which may be caused by sudden environmental changes, equipment aging, or control system failure. If the drift index remains at a high level, it means that the area may be overloaded, and long-term high-load operation may cause increased equipment wear or failure, which requires attention and appropriate load adjustment or maintenance measures.

[0070] When the motion load drift index is smaller, it indicates that the load in the monitoring area is relatively stable and the load level is within the normal or expected range. A smaller drift index indicates that the load changes in the monitoring area are small, the equipment is running smoothly, and it meets the design requirements of the system or the historical operation records. This usually means that the system can handle the current load efficiently and the equipment is in a healthy working state without immediate intervention. A continuous small drift index reflects that the load in the area is evenly distributed and can operate for a long time with low risk.

[0071] The temperature distribution anomaly index and the motion load drift index are converted into the first eigenvector, and the first eigenvector is used as the input of the machine learning model. The machine learning model predicts the wheel load anomaly degree value label of each monitoring area with each group of first eigenvectors as the prediction target, and minimizes the sum of prediction errors of the wheel load anomaly degree value labels of all monitoring areas as the training target. The machine learning model is trained until the sum of prediction errors converges and the model training is stopped. The wheel load anomaly degree value of each monitoring area is determined according to the model output results, wherein the machine learning model is a polynomial regression model.

[0072] The method for obtaining the abnormal degree value of the wheel load in each monitoring area is as follows: from the first eigenvector training data of the trained machine learning model, the corresponding function expression is obtained: CP = F (KM, SD); where F is the output function of the model, KM is the temperature distribution anomaly index, SD is the motion load drift index, and CP is the abnormal degree value of the wheel load in each monitoring area.

[0073] The acquired wheel load abnormality degree value of each monitoring area is compared with the pre-set reference threshold of the wheel load abnormality degree value. If the wheel load abnormality degree value is greater than or equal to the reference threshold of the wheel load abnormality degree value, it means that the wheel load abnormality degree in the monitoring area is high. At this time, an early warning signal is generated and it is marked as a wheel load abnormality monitoring area; if the wheel load abnormality degree value is less than the reference threshold of the wheel load abnormality degree value, it means that the wheel load abnormality degree in the monitoring area is low. At this time, no early warning signal is generated and it is marked as a normal wheel load monitoring area.

[0074] S3: For the monitoring area with abnormal wheel-foot load, the fuzzy logic algorithm is used to adjust the load distribution of the wheel-foot system in real time according to the abnormal degree of the wheel-foot load and the movement speed, so as to optimize the energy utilization of the system.

[0075] The wheel-foot load abnormality value CP and the wheel-foot movement speed E are used as the input items of fuzzy logic, and the load distribution H of the wheel-foot system is used as the output item of fuzzy logic.

[0076] The input variables (CP and E) are fuzzified, that is, these precise numerical values ​​are converted into fuzzy sets. This involves assigning continuous numerical inputs to a set of fuzzy sets (such as "low", "medium", "high") to represent the different states of each input variable.

[0077] Wheel load abnormality value (CP): CP indicates the abnormality of the wheel load in the monitoring area, reflecting the health status of the current system. It can be calculated by the previous machine learning model, and the range is usually [0,1], where: 0: indicates that the load abnormality is low and the system status is close to normal. 1: indicates that the load abnormality is extremely high, and the system may be overloaded or seriously abnormal.

[0078] After fuzzification, CP can be divided into three fuzzy sets: Low: The load abnormality is low. Medium: The load abnormality is moderate, and there may be slight abnormalities. High: The load abnormality is high, indicating serious abnormalities.

[0079] Wheel-foot movement speed (E): E represents the actual running speed of the wheel-foot system, which reflects the current state of movement. The range of movement speed depends on the system itself, such as from 0 (stop) to maximum speed (the highest driving speed of the system).

[0080] After fuzzification, E can be divided into three fuzzy sets: Slow: low speed state, the system moves slowly. Medium: medium speed state, the system runs at normal speed. Fast: high speed state, the system runs fast.

[0081] Fuzzy inference rules are the core of fuzzy logic control, which generates corresponding outputs based on the input fuzzy set and the rule base. The rule base consists of a set of "IF-THEN" rules to describe the relationship between input and output.

[0082] A set of fuzzy rules can be defined to map CP (load abnormality) and E (movement speed) to the output H (load distribution):

[0083] Fuzzy rule example: If CP is low and E is slow, then H should be normal allocation. If CP is low and E is fast, then H should be reduced allocation. If CP is medium and E is slow, then H should be slightly increased allocation. If CP is medium and E is fast, then H should be greatly reduced allocation. If CP is high and E is slow, then H should be increased allocation. If CP is high and E is fast, then H should be significantly reduced allocation to prevent overload.

[0084] When the load abnormality is low and the system is close to normal state, the load distribution can be reasonably adjusted according to the movement speed. If the system speed is fast, it is usually necessary to reduce the load distribution to avoid excessive energy consumption.

[0085] When the load abnormality is moderate, the system may be slightly abnormal, so a significant load reduction is required at high speed to avoid further abnormality.

[0086] When the load is abnormally high, the system is in a serious load state. In this case, if the speed is slow, the load distribution can be appropriately increased to avoid excessive energy waste. But if the speed is fast, the load must be significantly reduced to prevent overload and damage.

[0087] The result of fuzzy reasoning is a fuzzy set, which must be converted into an accurate output value, i.e., defuzzification. Commonly used defuzzification methods include the centroid method, which calculates the centroid of the fuzzy output as the final load distribution value H.

[0088] Through the centroid method, a specific load distribution value H can be obtained, which represents the load that the system should distribute under current conditions.

[0089] The output variable H represents the load that should be distributed to the wheel-foot system. Through fuzzy logic reasoning, the load distribution will be dynamically adjusted to ensure that the system can operate efficiently and avoid overload under different movement speeds and load conditions.

[0090] Reduce distribution: Indicates that load distribution should be reduced at high load or high speed to prevent overload.

[0091] Normal distribution (Normal): Indicates that the load is in normal state and the system operates at standard load.

[0092] Increase distribution: It means to appropriately increase the load distribution at low load or low speed to maintain energy utilization efficiency.

[0093] Through the above fuzzy logic control, the system can dynamically adjust the load distribution according to the abnormal degree of wheel load and movement speed monitored in real time. The following are specific ways to optimize the energy utilization of the system:

[0094] At high load and high speed, the system reduces energy consumption by reducing load distribution, avoiding unnecessary excessive energy consumption and reducing the risk of equipment damage.

[0095] At low load and low speed, the system increases load distribution appropriately to ensure that energy is not wasted and maintain a smooth load distribution.

[0096] At medium load, the load distribution is adjusted according to the speed to ensure that the system can maintain normal operation without performance loss.

[0097] In this application, the fuzzy logic algorithm is used to take the wheel foot load abnormality value and the wheel foot movement speed as input, and the load distribution value is generated, so that the load distribution of the system can be adjusted in real time. In this process, the fuzzy reasoning rules provide the system with a flexible load adjustment strategy according to different load states and speed states, optimize the energy utilization rate, and ensure the stability and efficient operation of the system.

[0098] S4: Predict the wheel foot load change in the wheel foot load abnormality monitoring area within the adjusted fixed time period, and adjust the motion control strategy in advance according to the prediction result to prevent local overheating.

[0099] Collect system load data within a fixed time period, including temperature distribution anomaly index (TDI), motion load drift index (SD), and real-time wheel-foot load distribution (H). These data can reflect the operating status and load change trend of the wheel-foot system. Collect other environmental factors that may affect the load, such as terrain complexity, mission type, ambient temperature, etc., which can help improve the accuracy of the prediction model.

[0100] Common load change prediction models include time series models and machine learning models. The following are some common prediction methods:

[0101] ARIMA (Autoregressive Integrated Moving Average): Suitable for processing trends and seasonal fluctuations in time series data. The ARIMA model can capture short-term changes and trends in wheel load. Based on historical load data, the model can predict load changes in future time periods, thereby identifying possible overload or local overheating trends.

[0102] LSTM (Long Short-Term Memory): LSTM is a recursive neural network model for processing time series data, which is particularly suitable for processing long-term dependencies and nonlinear changes in load data. The LSTM model can make accurate predictions about the long-term trend of load by learning the historical change patterns of load data.

[0103] It can capture the compound impact of different factors (such as abnormal load status and movement speed) on future loads.

[0104] The collected data is organized into the model’s input feature vector for training and prediction. Common input features include:

[0105] Temperature Distribution Anomaly Index (TDI): reflects the changes in the system temperature load and provides information on whether the load has an overheating trend.

[0106] Motion load drift index (SD): reflects the stability of the load during system motion and helps predict load fluctuations.

[0107] Load distribution value (H): The load distribution value previously adjusted by fuzzy logic as a direct reflection of the current load status of the system.

[0108] External factors: such as terrain complexity, speed changes, etc.

[0109] These input features can form the input sequence of the model through time series processing, which is convenient for subsequent load change prediction.

[0110] Time series model: For the ARIMA model, the historical load data is used to train the model, and the autoregression, moving average, and difference parameters of the model are tuned to enable it to capture the trends and fluctuations of the load data.

[0111] Machine learning model: If a deep learning model such as LSTM is used, historical load data will be used as a training set. The model will be trained through multiple iterations to learn the time correlation and change pattern of the system load.

[0112] The output is the predicted value of wheel load in the future, including:

[0113] Prediction trend of abnormal temperature distribution: It can predict whether the system will experience local overheating.

[0114] Prediction of load fluctuations: Identify whether the load will enter an unstable state.

[0115] Based on the results of the prediction model, if the load forecast value shows a rapid temperature rise and excessive load in the future, which may cause local overheating, the system will issue a warning signal.

[0116] If the predicted value of the temperature distribution anomaly index (TDI) continues to rise and approaches or exceeds the set safety threshold, it indicates that the system load will be in an abnormally high temperature state.

[0117] If the predicted value of the motion load drift index (SD) fluctuates greatly, it means that the load distribution is unstable, which may lead to excessive load in local areas.

[0118] Once future abnormal load trends are identified, the system will adjust the motion control strategy in advance based on the prediction results to prevent local overheating. These strategies may include:

[0119] Dynamically adjust the load distribution (H): Dynamically adjust the load distribution of the wheel-foot system again through fuzzy logic algorithm. For example:

[0120] If the load forecast value for a certain area is too high, reduce the load distribution in that area to alleviate its burden.

[0121] At the same time, part of the load can be transferred to areas with lower load to balance the overall system load.

[0122] If the system predicts that the load under high-speed operation will cause overheating, the control strategy can reduce the movement speed in advance to reduce heat generation. Movement speed is directly related to load, and slowing down speed helps reduce load and extend the service life of equipment. In complex terrain or tasks, the system can switch from wheeled movement to footed movement (or vice versa) in advance based on the prediction results to adapt to terrain changes and reduce the risk of overheating caused by abnormal loads.

[0123] After prediction and adjustment, the system will monitor the adjustment effect based on real-time data. If the adjusted load distribution and motion control strategy can effectively reduce the load, the system will continue to execute the current strategy. Otherwise, the system will further optimize and adjust until the load status returns to the normal range.

[0124] Through the prediction-based control strategy, the system can identify possible load problems in advance and make adjustments before the problem occurs. In this way, it can not only effectively prevent local overheating, but also optimize the energy utilization of the entire system.

[0125] Reduced risk of overheating: Early adjustments avoid equipment overheating due to high loads, thereby reducing energy waste and the risk of equipment damage.

[0126] Balanced load distribution: By intelligently adjusting load distribution, the system can balance energy consumption among multiple areas to avoid local overload and uneven load.

[0127] Improve overall operating efficiency: Load forecasting combined with motion control strategy ensures that the system can maintain an efficient and stable working state during long-term operation.

[0128] In this application, local overheating in the wheel foot load abnormality monitoring area can be effectively prevented by predicting load changes. Based on the prediction results of the temperature distribution abnormality index and the motion load drift index, the system can adjust the load distribution and motion control strategy in advance to prevent performance degradation caused by high load. This not only optimizes the energy utilization of the system, but also improves the safety and stability of equipment operation.

[0129] In this embodiment, the wheel-foot robot is first divided into multiple monitoring areas, and the temperature distribution data and motion load data in each area are obtained by installing temperature sensors and acceleration sensors in the wheel-type and foot-type motion control systems. Then, these data are preprocessed and analyzed to determine the monitoring areas with abnormal loads. For these abnormal areas, a fuzzy logic algorithm is used to adjust the load distribution of the system in real time according to the abnormal degree and motion speed of the wheel-foot load to optimize energy utilization. After the adjustment, the load changes within a fixed time period are predicted, and the motion control strategy is adjusted in advance based on the prediction results to prevent local overheating.

[0130] Embodiment 2, a wheel-foot robot wheel-foot coordinated control system described in this embodiment includes a data acquisition module, a data analysis module, a load adjustment module and a motion control module;

[0131] Data acquisition module: The wheeled-legged robot is divided into several monitoring areas. The temperature distribution data and motion load data in each monitoring area are obtained through the temperature sensors and acceleration sensors installed in the wheeled motion control system and the legged motion control system.

[0132] Data analysis module: pre-processes the monitored temperature distribution data and motion load data, analyzes the pre-processed temperature distribution data and motion load data, and determines the monitoring area of ​​abnormal wheel foot load according to the analysis results;

[0133] Load adjustment module: For monitoring areas with abnormal wheel-foot load, the load distribution of the wheel-foot system is adjusted in real time according to the abnormal degree of wheel-foot load and movement speed through fuzzy logic algorithm to optimize the system energy utilization rate;

[0134] Motion control module: predict the wheel load changes in the wheel load abnormality monitoring area within the adjusted fixed time period, and adjust the motion control strategy in advance according to the prediction results to prevent local overheating.

[0135] The above formulas are all dimensionless and numerical calculations. The formula is a formula for the most recent real situation obtained by collecting a large amount of data and performing software simulation. The preset parameters in the formula are set by technicians in this field according to actual conditions.

[0136] It should be understood that the term "and / or" in this article is only a description of the association relationship of associated objects, indicating that there can be three relationships. For example, A and / or B can represent: A exists alone, A and B exist at the same time, and B exists alone. A and B can be singular or plural. In addition, the character " / " in this article generally indicates that the associated objects before and after are in an "or" relationship, but it may also indicate an "and / or" relationship. Please refer to the context for specific understanding.

[0137] Those of ordinary skill in the art will appreciate that the units and algorithm steps of each example described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professional and technical personnel can use different methods to implement the described functions for each specific application, but such implementation should not be considered to be beyond the scope of this application.

[0138] The above description is only a specific implementation manner of the present application, but the protection scope of the present application is not limited thereto. Any technician familiar with the technical field can easily think of changes or substitutions within the technical scope disclosed in the present application, which should be included in the protection scope of the present application.

Claims

1. A wheel-foot coordinated control method for a wheel-foot robot, characterized in that: The steps include: S1: Divide the wheeled-legged robot into several monitoring areas, and obtain the temperature distribution data and motion load data in each monitoring area through the temperature sensors and acceleration sensors installed in the wheeled motion control system and the legged motion control system; S2: preprocessing the monitored temperature distribution data and motion load data, analyzing the preprocessed temperature distribution data and motion load data, and determining a monitoring area of ​​abnormal wheel foot load according to the analysis results; S3: For the monitoring area with abnormal wheel-foot load, the load distribution of the wheel-foot system is adjusted in real time according to the abnormal degree of wheel-foot load and movement speed through fuzzy logic algorithm to optimize the system energy utilization rate; S4: Predict the wheel foot load change in the wheel foot load abnormality monitoring area within the adjusted fixed time period, and adjust the motion control strategy in advance according to the prediction result to prevent local overheating.

2. A wheel-foot coordinated control method for a wheel-foot robot according to claim 1, characterized in that: In S2, the temperature distribution anomaly index is generated after analyzing the preprocessed temperature distribution data. The method for obtaining the temperature distribution anomaly index is: Set the prior distribution P(θ). If the temperature data conforms to the normal distribution, use the prior normal distribution to represent it. where μ is the initial mean temperature, σ 2 is the initial variance of the temperature; for a new temperature data point T, the likelihood function is defined according to its actual distribution, that is, the probability of observing the current temperature data. The likelihood function P(T|θ) is the probability density function based on the new temperature data T, and the expression is: According to Bayes' theorem, the posterior distribution P(θ|T) combines the prior distribution and the likelihood function of the current observation data, expressed as: P(θ|T) is the posterior distribution, that is, the probability of the temperature distribution parameter θ after the data T is observed. P(θ) is the prior distribution, which represents the temperature distribution before the observation. P(T) is the marginal likelihood of the observed temperature. The temperature distribution anomaly index is calculated as follows: Where kM is the temperature distribution anomaly index, [T min ,T max ] indicates the temperature threshold range.

3. A wheel-foot coordinated control method for a wheel-foot robot according to claim 2, characterized in that: In S2, the motion load drift index is generated after analyzing the preprocessed motion load data. The method for obtaining the motion load drift index is: Real-time acquisition of motion load data within the Q time period is expressed as L1, L2, ..., L n ; where each L n is the load data at time n, and defines the distance metric between load data points to measure the difference between two load data points. The expression is: Among them, L i,k and L j,k Respectively represent the load data L i and L j The kth eigenvalue of , m is the feature dimension. In the load data space, the neighborhood radius of a point is defined. If the distance between two points is less than or equal to ∈, they are regarded as neighbors of each other. Within a given radius ∈, at least MinPts neighbors are required to regard the data point as a core point and classify it into a cluster. The load data is divided into multiple clusters. Points that cannot be classified into any cluster are marked as noise points, that is, abnormal load points; clustering is performed based on density. The local density is calculated by the number of neighbors within the radius ∈. The specific formula is: Among them, |N ∈ (L i )| represents the number of data points within the radius ∈, V ∈ Represents the neighborhood volume with ∈ as radius. For the two-dimensional space of Euclidean distance, the neighborhood volume is V ∈ =π∈ 2 ; Calculate the motion load drift index, the expression is: In the formula, ρ(L i ) is the data point L i The local density, ρ mean (C) is point L i The average local density of cluster C, SD is the motion load drift index.

4. The method for controlling wheel-foot coordinated wheel-foot robot according to claim 3, characterized in that: In S2, the temperature distribution anomaly index and the motion load drift index are converted into the first eigenvector, and the first eigenvector is used as the input of the machine learning model. The machine learning model predicts the wheel load anomaly degree value label of each monitoring area with each group of first eigenvectors as the prediction target, and minimizes the sum of prediction errors of the wheel load anomaly degree value labels of all monitoring areas as the training target. The machine learning model is trained until the sum of prediction errors converges and the model training is stopped. The wheel load anomaly degree value of each monitoring area is determined according to the model output results, wherein the machine learning model is a polynomial regression model.

5. The method for controlling wheel-foot coordinated wheel-foot robot according to claim 4, characterized in that: The acquired wheel load abnormality value of each monitoring area is compared with the preset reference threshold of the wheel load abnormality value. If the wheel load abnormality value is greater than or equal to the reference threshold of the wheel load abnormality value, it means that the wheel load abnormality in the monitoring area is high. At this time, an early warning signal is generated and it is marked as a wheel load abnormality monitoring area; If the wheel load abnormality value is less than the reference threshold of the wheel load abnormality value, it indicates that the wheel load abnormality in the monitoring area is low. In this case, no warning signal is generated and it is marked as a normal wheel load monitoring area.

6. A wheel-foot coordinated control method for a wheel-foot robot according to claim 5, characterized in that: In S3, for the monitoring area with abnormal wheel-foot load, the fuzzy logic algorithm is used to adjust the load distribution of the wheel-foot system in real time according to the abnormal degree of wheel-foot load and the movement speed, and optimize the system energy utilization rate, specifically: The wheel-foot load abnormality value CP and the wheel-foot movement speed E are used as the input items of fuzzy logic, and the load distribution H of the wheel-foot system is used as the output item of fuzzy logic. Fuzzify the input and output variables, that is, convert these precise values ​​into fuzzy sets; Construct fuzzy rules to generate corresponding outputs based on the input fuzzy sets and rule base; The result of fuzzy reasoning is a fuzzy set, which is converted into an accurate output value, namely defuzzification, and the centroid of the fuzzy output is calculated as the final load distribution value H; Through the centroid method, a specific load distribution value H is obtained, which represents the load distributed by the system under current conditions.

7. A wheel-foot robot wheel-foot coordinated control system, used to implement a wheel-foot robot wheel-foot coordinated control method according to any one of claims 1 to 6, characterized in that: Including data acquisition module, data analysis module, load adjustment module and motion control module; Data acquisition module: The wheeled-legged robot is divided into several monitoring areas. The temperature distribution data and motion load data in each monitoring area are obtained through the temperature sensors and acceleration sensors installed in the wheeled motion control system and the legged motion control system. Data analysis module: pre-processes the monitored temperature distribution data and motion load data, analyzes the pre-processed temperature distribution data and motion load data, and determines the monitoring area of ​​abnormal wheel foot load according to the analysis results; Load adjustment module: For monitoring areas with abnormal wheel-foot load, the load distribution of the wheel-foot system is adjusted in real time according to the abnormal degree of wheel-foot load and movement speed through fuzzy logic algorithm to optimize the system energy utilization rate; Motion control module: predict the wheel load changes in the wheel load abnormality monitoring area within the adjusted fixed time period, and adjust the motion control strategy in advance according to the prediction results to prevent local overheating.

Images