Stair climbing load-bearing robot control system of multi-degree-of-freedom driving wheel train

Abstract

The invention discloses a stair climbing load-bearing robot control system of a multi-degree-of-freedom driving wheel train, and relates to the technical field of robot control, historical garbage search information of a stair climbing load-bearing robot in a sub-region is obtained, search path abnormal points are marked based on the historical garbage search information, and the search path abnormal points are found; whenever the stair climbing load-bearing robot enters a search path abnormal point, the stair climbing load-bearing robot senses the environment and the posture of the robot according to the adjusted sensing frequency, the balance state change trend is predicted based on the environment and posture sensing result, and when it is predicted that the balance state change trend develops towards an inclined state, the stair climbing load-bearing robot searches the abnormal point of the path. And each driving wheel train of the stair climbing load-bearing robot is controlled to keep the balance state of the stair climbing load-bearing robot. The control system can conduct toppling prediction according to the climbing posture of the stair climbing load-bearing robot, the inclination angle can be adjusted in time before the stair climbing load-bearing robot turns over, and the stability and safety of operation of the stair climbing load-bearing robot are guaranteed.

Description

Technical Field

[0001] The invention relates to the technical field of robot control, and in particular to a control system for a stair-climbing load-bearing robot with a multi-degree-of-freedom driving wheel system. Background Art

[0002] Stair climbing load-bearing robots are a type of autonomous robots that can climb stairs in different environments. They are mainly used in industries, construction, rescue, homes, and other fields. With the progress of society and the development of intelligent technology, the application of stair climbing robots is becoming more and more extensive.

[0003] As a stair-climbing load-bearing robot operating in an area, the stair-climbing load-bearing robot needs to search for garbage in multiple buildings in the area and transport the loaded garbage from the lower or upper floors to the ground through the stairs. In actual applications, the stair-climbing load-bearing robot usually makes adjustments only when it senses that it is tilted. However, when the tilt angle is too large, the stair-climbing load-bearing robot will not be able to adjust to a balanced state. At this time, if the stair-climbing load-bearing robot overturns, it may not only cause damage to the stair-climbing load-bearing robot, but also cause the mission to be terminated, resulting in the inability to clean up the loaded garbage in time, bringing potential hidden dangers of garbage operations.

[0004] Based on this, the present invention proposes a stair-climbing load-bearing robot control system with a multi-degree-of-freedom drive wheel system, which can predict the tipping over of the stair-climbing load-bearing robot according to the climbing posture of the stair-climbing load-bearing robot, so that the tilt angle can be adjusted in time before the stair-climbing load-bearing robot falls over, thereby ensuring the stability and safety of the stair-climbing load-bearing robot's operation. Summary of the invention

[0005] The purpose of the present invention is to provide a stair climbing load-bearing robot control system with a multi-degree-of-freedom driving wheel system to overcome the shortcomings of the background technology.

[0006] In order to achieve the above object, the present invention provides the following technical solution: a control method for a stair climbing load-bearing robot with a multi-degree-of-freedom driving gear train, the control method comprising the following steps:

[0007] The control system obtains garbage search area information, divides the garbage search area into multiple sub-areas, initially numbers each sub-area, establishes an initial sub-area list, and for each sub-area, obtains a preset search route of the stair-climbing load-bearing robot in the sub-area;

[0008] Obtain historical garbage search information of the stair climbing load-bearing robot in the sub-area, mark abnormal points of the search path based on the historical garbage search information, and adjust the perception frequency of the stair climbing load-bearing robot whenever the stair climbing load-bearing robot enters the abnormal point of the search path;

[0009] The stair-climbing load-bearing robot perceives the environment and its own posture according to the adjusted perception frequency, and predicts the trend of change of the balance state based on the environment and its own posture perception results. When the predicted trend of change of the balance state is toward a tilted state, the driving wheel systems of the stair-climbing load-bearing robot are automatically controlled to maintain the balance state of the stair-climbing load-bearing robot.

[0010] In a preferred embodiment, obtaining historical garbage search information of the stair-climbing load-bearing robot in a sub-area and marking abnormal points of the search path based on the historical garbage search information include the following steps:

[0011] When the stair climbing load-bearing robot completes a search in a sub-area, the control system obtains abnormal conditions encountered by the load-bearing robot on the preset search route, including pauses, detours, tilts, and shakes;

[0012] Obtain the historical garbage search information of the stair climbing load-bearing robot in the sub-area, integrate the multiple garbage search information, and mark the locations where the stair climbing load-bearing robot pauses, detours, tilts or shakes during the search process in the sub-area as initial abnormal points;

[0013] For each initial outlier point, the number of outlier records of the initial outlier point is obtained. When the number of outlier records of the initial outlier point is less than 3 times, the initial outlier point is deleted from the sub-region model. When the number of outlier records of the initial outlier point is greater than or equal to 3 times, the initial outlier point is updated to a search path outlier point.

[0014] In a preferred embodiment, whenever the stair-climbing load-bearing robot enters an abnormal point in the search path, adjusting the sensing frequency of the stair-climbing load-bearing robot comprises the following steps:

[0015] When the stair climbing load-bearing robot enters the abnormal point of the search path, the control system calculates the adjustment factor according to the number of abnormal records, and the expression is: In the formula, adjus is the adjustment factor, n is the type of abnormal record, yc i represents the number of occurrences of the i-th abnormal record, ω i Represents the weight of the i-th abnormal record;

[0016] The sensing frequency of the stair climbing load-bearing robot is adjusted by the obtained adjustment factor. The adjustment algorithm is: new =fq old (1+log5(adjust+1)), where fq new is the adjusted perceived frequency, fq old is the perceived frequency before adjustment, adjus is the adjustment factor;

[0017] After the perception frequency of the stair-climbing load-bearing robot is corrected by the adjustment algorithm, when the stair-climbing load-bearing robot enters an abnormal point in the search path, it senses according to the adjusted perception frequency corresponding to the abnormal point in the search path.

[0018] In a preferred embodiment, predicting the balance state change trend based on the environment and the self-posture perception results includes the following steps:

[0019] The slope deviation, flatness standard deviation, slope angle increase rate, and friction decrease rate obtained in real time are used as input features, and the input features are input into the trained random forest model, and the random forest model outputs the target value;

[0020] The target value is compared with the preset target threshold. The target threshold is used to predict the trend of equilibrium state change. If the target value is less than or equal to the target threshold, the equilibrium state change trend is predicted to develop toward the equilibrium state. If the target value is greater than the target threshold, the equilibrium state change trend is predicted to develop toward the tilted state.

[0021] In a preferred embodiment, random forest model training includes the following steps:

[0022] Collect historical data and organize them into training sets, including:

[0023] Input feature subset X = {Δα,σ,ω(t),dF friction}, Δα is the slope deviation, σ is the flatness standard deviation, ω(t) is the rate of increase of the inclination angle, dF friction is the friction force decreasing rate;

[0024] Target value Y = S balance , S balance is the predicted value;

[0025] Random forests generate multiple decision trees based on random subsets of features. Each decision tree splits and makes decisions based on different random selections. For each decision tree, feature values ​​are selected when nodes split, and decisions are made based on feature values.

[0026] Random forest generates multiple decision trees based on different subsets of input features. When splitting, the decision tree divides the data into different subsets according to different thresholds of the features, and finally generates leaf nodes. Each leaf node represents the prediction result based on the input features.

[0027] In the prediction stage, random forest will perform weighted average or voting on the output results of all decision trees to obtain the final prediction value. Assuming there are T trees, the prediction value is: In the formula, S balance is the predicted value, T is the number of decision trees, S i is the prediction result of the i-th decision tree.

[0028] In a preferred embodiment, the stair-climbing load-bearing robot perceives the environment and its own posture according to the adjusted perception frequency, including the following steps:

[0029] The stair climbing load-bearing robot perceives the environment and its own posture according to the adjusted perception frequency, scans the stairs in front through the laser radar and visual sensor, and measures the slope of the stairs and detects the flatness of the stair tread surface;

[0030] When the stair climbing load-bearing robot starts climbing, it senses the current tilt angle through the gyroscope and tilt sensor, and obtains the friction between the traveling wheel and the ground through the wheel train sensor;

[0031] The actual slope is subtracted from the standard slope to obtain the slope deviation, and the rate of increase of the inclination angle and the rate of decrease of the friction force are calculated within the monitoring period. The logic for obtaining the rate of increase of the inclination angle is as follows: subtract the inclination angle at the current moment from the inclination angle at the previous moment to obtain the inclination angle difference, and compare the inclination angle difference with the previous monitoring period to obtain the rate of increase of the inclination angle. The logic for calculating the rate of decrease of the friction force is as follows: subtract the friction force at the current moment from the friction force at the previous moment to obtain the friction difference, and compare the friction difference with the previous monitoring period to obtain the rate of decrease of the friction force.

[0032] Select n measuring points on a step surface and mark the heights of the n measuring points as h1, h2, ..., h n ; Then the calculation expression of flatness standard deviation is: In the formula, σ is the flatness standard deviation, n is the number of measurement points, and h is i is the height of the ith measurement point, is the average height. The larger the flatness standard deviation, the more uneven the step surface is.

[0033] In a preferred embodiment, the control system obtains garbage search area information, divides the garbage search area into multiple sub-areas, and after initially numbering each sub-area, establishes an initial sub-area list, including the following steps:

[0034] Obtaining garbage search area information, including building layout, stair distribution, and load garbage location;

[0035] Sub-area division includes division based on physical characteristics, division based on search tasks, and division based on distance:

[0036] Divide by physical features: Divide the garbage search area into multiple sub-areas based on the physical layout of the building, each sub-area corresponds to a floor, a flight of stairs, or the area where a specific equipment is located;

[0037] Division by search task: Divide the garbage search area into corresponding sub-areas according to the inspection scope of the search task;

[0038] Divide by distance: divide the garbage search area into several sub-areas based on the distance;

[0039] Assign a unique number to each sub-area, and record the number, location information, area, and search target parameters of each sub-area in the management system to form an initial sub-area list.

[0040] A control system for a stair climbing load-bearing robot with a multi-degree-of-freedom drive gear train, including a sub-area division module, an abnormal point marking module, a perception frequency adjustment module, a tilt prediction module, and a control module;

[0041] Sub-area division module: obtains garbage search area information, divides the garbage search area into multiple sub-areas, creates an initial sub-area list after initially numbering each sub-area, and obtains the preset search route of the stair climbing load-bearing robot in the sub-area;

[0042] Anomaly marking module: obtains the historical garbage search information of the stair climbing load-bearing robot in the sub-area, and marks the abnormal points of the search path based on the historical garbage search information;

[0043] Perception frequency adjustment module: Whenever the stair climbing load-bearing robot enters an abnormal point in the search path, the perception frequency of the stair climbing load-bearing robot is adjusted;

[0044] Tilt prediction module: The stair climbing load-bearing robot perceives the environment and its own posture according to the adjusted perception frequency, and predicts the change trend of the balance state based on the perception results of the environment and its own posture;

[0045] Control module: When it is predicted that the balance state changes in a trend toward a tilted state, each driving wheel system of the stair climbing load-bearing robot is automatically controlled to maintain the balance state of the stair climbing load-bearing robot.

[0046] In the above technical solution, the technical effects and advantages provided by the present invention are:

[0047] 1. The present invention obtains the historical garbage search information of the stair climbing load-bearing robot in the sub-area, marks the abnormal points of the search path based on the historical garbage search information, and adjusts the perception frequency of the stair climbing load-bearing robot whenever the stair climbing load-bearing robot enters the abnormal point of the search path. The stair climbing load-bearing robot perceives the environment and its own posture according to the adjusted perception frequency, and predicts the trend of the balance state change based on the environment and its own posture perception results. When the predicted balance state change trend develops toward the tilted state, each driving wheel system of the stair climbing load-bearing robot is automatically controlled to maintain the balance state of the stair climbing load-bearing robot. The control system can predict the tipping according to the climbing posture of the stair climbing load-bearing robot, so that the tilt angle can be adjusted in time before the stair climbing load-bearing robot falls over, ensuring the stability and safety of the stair climbing load-bearing robot.

[0048] 2. After the present invention corrects the perception frequency of the stair-climbing load-bearing robot through an adjustment algorithm, when the stair-climbing load-bearing robot enters an abnormal point in the search path, it performs perception based on the adjusted perception frequency corresponding to the abnormal point in the search path. Therefore, during the search process of the stair-climbing load-bearing robot, a preset perception frequency can be adopted for the general search route, which can reduce energy consumption while maintaining normal perception. Increasing the perception frequency for the abnormal point in the search path can effectively improve the accuracy of the prediction results of the balance state change trend, thereby further ensuring the stability of the operation of the stair-climbing load-bearing robot. BRIEF DESCRIPTION OF THE DRAWINGS

[0049] 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.

[0050] Figure 1 The present invention is a flow chart of the method. DETAILED DESCRIPTION

[0051] 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.

[0052] Example 1: Please refer to Figure 1 As shown, the control method of the stair climbing load-bearing robot with a multi-degree-of-freedom driving gear train described in this embodiment includes the following steps:

[0053] The control system obtains garbage search area information, divides the garbage search area into multiple sub-areas, and after initially numbering each sub-area, establishes an initial sub-area list. For each sub-area, a preset search route of the stair-climbing load-bearing robot in the sub-area is obtained, and the preset search route includes a stair-climbing route and an equipment search route. The historical garbage search information of the stair-climbing load-bearing robot in the sub-area is obtained, and abnormal points of the search path are marked based on the historical garbage search information. Whenever the stair-climbing load-bearing robot enters an abnormal point of the search path, the perception frequency of the stair-climbing load-bearing robot is adjusted. The stair-climbing load-bearing robot perceives the environment and its own posture according to the adjusted perception frequency, and predicts the balance state change trend based on the environment and its own posture perception results. When the predicted balance state change trend is toward a tilted state, each drive wheel system of the stair-climbing load-bearing robot is automatically controlled to maintain the balance state of the stair-climbing load-bearing robot.

[0054] This application obtains the historical garbage search information of the stair climbing load-bearing robot in the sub-area, marks the abnormal points of the search path based on the historical garbage search information, and adjusts the perception frequency of the stair climbing load-bearing robot whenever the stair climbing load-bearing robot enters the abnormal point of the search path. The stair climbing load-bearing robot perceives the environment and its own posture according to the adjusted perception frequency, and predicts the trend of the balance state change based on the environment and its own posture perception results. When the predicted balance state change trend develops toward a tilted state, each driving wheel system of the stair climbing load-bearing robot is automatically controlled to maintain the balance state of the stair climbing load-bearing robot. The control system can predict the tipping over according to the climbing posture of the stair climbing load-bearing robot, so that the tilt angle can be adjusted in time before the stair climbing load-bearing robot falls over, ensuring the stability and safety of the stair climbing load-bearing robot.

[0055] Embodiment 2: The control system obtains garbage search area information, divides the garbage search area into multiple sub-areas, and after initially numbering each sub-area, establishes an initial sub-area list, including the following steps:

[0056] Through sensors, maps, cameras and other equipment, detailed information about the garbage search area can be obtained, including building layout, stair distribution, load garbage location, etc.

[0057] Sub-area division includes division based on physical characteristics, division based on search tasks, and division based on distance:

[0058] Divide by physical features: Divide the garbage search area into multiple sub-areas based on the physical layout of the building (such as floors, stair locations, and room distribution). Each sub-area can correspond to a floor, a flight of stairs, or the area where a specific device is located.

[0059] Division by search task: Different inspection ranges are divided into corresponding sub-areas according to different search tasks, such as the location of load garbage inspection points and equipment maintenance frequency.

[0060] Divide by distance: Taking into account the optimization of the search path, the garbage search area is divided into several sub-areas based on the distance, ensuring that the search route in each sub-area has a reasonable length.

[0061] Assign a unique number to each sub-area to facilitate accurate positioning and scheduling in subsequent search tasks. For example, the floor number plus the room number can be used (such as "1-01" represents the 01 area on the first floor).

[0062] Optimize numbering logic: Ensure clear numbering logic based on the geographical location and adjacent relationships of the sub-areas to facilitate the planning and adjustment of search paths.

[0063] The number, location information, area, search target and other parameters of each sub-area are recorded in the system to form an initial sub-area list.

[0064] For each sub-area, obtaining a preset search route of the stair-climbing load-bearing robot in the sub-area, the preset search route including a stair-climbing route and an equipment search route, includes the following steps:

[0065] Use sensors, lidar or map data to model the sub-areas in three dimensions, identify the specific locations of structures such as stairs, platforms, and equipment, identify the location of stairs, stair slope, number of steps, handrail location, etc. in the sub-area, and integrate this information into the sub-area model to determine the location of equipment that needs to be searched in the sub-area, including load garbage, monitoring devices, etc., and mark their coordinates in the sub-area model;

[0066] Define several key points in the stair area as key locations in the stair climbing route (such as the starting point, turning point, and end point of the stairs). These points will guide the robot to climb the stairs smoothly. Based on the device location, each device is defined as a search point. When the robot reaches this location, it will perform the corresponding device search task (such as reading sensor data or performing device status check).

[0067] According to the physical structure of the stairs in the sub-area, the robot's climbing route on the stairs is planned to ensure that the route is smooth and the robot is not affected by stair obstacles or slopes. When planning, it is necessary to consider the robot's driving speed, the need to adjust the drive wheel system, etc. Starting from the end of the stair climbing route, combined with the location of the equipment search point, the robot is planned to reach each equipment search point in the sub-area. The search efficiency is guaranteed, and the stair climbing route and the equipment search route are seamlessly connected to ensure that the robot can quickly switch to the equipment search task after completing the stair climbing, and return to the initial position after successfully completing all tasks.

[0068] Obtaining historical garbage search information of the stair-climbing load-bearing robot in the sub-area, and marking abnormal points of the search path based on the historical garbage search information, including the following steps:

[0069] When the stair climbing load-bearing robot completes a search in a sub-area, the control system obtains abnormal conditions encountered by the load-bearing robot on the preset search route, including pauses, detours, tilts, and shakes;

[0070] Obtain the historical multiple garbage search information of the stair climbing load-bearing robot in the sub-area, integrate the multiple garbage search information, and mark the places where the stair climbing load-bearing robot pauses, detours, tilts or shakes during the search process in the sub-area as initial abnormal points (that is, in the multiple search processes, any place where there is a pause, detour, tilt or shake is an initial abnormal point);

[0071] It should be noted that when there are few pauses, detours, tilts or jitters during a sub-area search, it may be just an accidental event. At this time, marking will increase the operating burden and energy consumption of the stair climbing load-bearing robot. Therefore, in this application, it is necessary to optimize the initial abnormal points to obtain the search path abnormal points;

[0072] For each initial outlier point, the number of abnormal records of the initial outlier point is obtained (the number of abnormal records includes the number of pauses, detours, tilts and jitters). When the number of abnormal records of the initial outlier point is less than 3 times, the initial outlier point is deleted from the sub-region model. When the number of abnormal records of the initial outlier point is greater than or equal to 3 times, the initial outlier point is updated to a search path outlier point.

[0073] Whenever the stair climbing load-bearing robot enters an abnormal point in the search path, the sensing frequency of the stair climbing load-bearing robot is adjusted, including the following steps:

[0074] When the stair climbing load-bearing robot enters the abnormal point of the search path, the control system calculates the adjustment factor according to the number of abnormal records, and the expression is: Where adjus is the adjustment factor, n is the type of abnormal record, and n=4, representing pause, detour, tilt, and jitter, respectively, and yc i represents the number of occurrences of the i-th abnormal record, ω i Represents the weight of the i-th abnormal record.

[0075] The larger the adjustment factor, the more serious the abnormality of the search path abnormal point, and the more it is necessary to increase the perception frequency of the stair climbing load-bearing robot. The perception frequency of the stair climbing load-bearing robot is adjusted by the obtained adjustment factor. The adjustment algorithm is: new =fq old (1+log5(adjust+1)), where fq new is the adjusted perceived frequency, fq old is the perceived frequency before adjustment, adjus is the adjustment factor;

[0076] After the perception frequency of the stair-climbing load-bearing robot is corrected by the adjustment algorithm, when the stair-climbing load-bearing robot enters an abnormal point in the search path, it senses according to the adjusted perception frequency corresponding to the abnormal point in the search path.

[0077] After the present application corrects the perception frequency of the stair-climbing load-bearing robot through an adjustment algorithm, when the stair-climbing load-bearing robot enters an abnormal point in the search path, it performs perception based on the adjusted perception frequency corresponding to the abnormal point in the search path. Thus, during the search process of the stair-climbing load-bearing robot, a preset perception frequency can be adopted for the general search route, which can reduce energy consumption while maintaining normal perception. Increasing the perception frequency for the abnormal point in the search path can effectively improve the accuracy of the prediction results of the balance state change trend, thereby further ensuring the stability of the operation of the stair-climbing load-bearing robot.

[0078] The stair climbing load-bearing robot perceives the environment and its own posture according to the adjusted perception frequency, including the following steps:

[0079] The stair climbing load-bearing robot perceives the environment and its own posture according to the adjusted perception frequency, scans the stairs in front through the laser radar and visual sensor, and measures the slope of the stairs and detects the flatness of the stair tread surface;

[0080] When the stair climbing load-bearing robot starts climbing, it senses the current tilt angle through the gyroscope and tilt sensor, and obtains the friction between the traveling wheel and the ground through the wheel train sensor;

[0081] The actual slope is subtracted from the standard slope to obtain the slope deviation, and the rate of increase of the inclination angle and the rate of decrease of the friction force are calculated within the monitoring period. The logic for obtaining the rate of increase of the inclination angle is as follows: subtract the inclination angle at the current moment from the inclination angle at the previous moment to obtain the inclination angle difference, and compare the inclination angle difference with the previous monitoring period to obtain the rate of increase of the inclination angle. The logic for calculating the rate of decrease of the friction force is as follows: subtract the friction force at the current moment from the friction force at the previous moment to obtain the friction difference, and compare the friction difference with the previous monitoring period to obtain the rate of decrease of the friction force.

[0082] Select n measuring points on a step surface and mark the heights of the n measuring points as h1, h2, ..., h n ; Then the calculation expression of flatness standard deviation is: In the formula, σ is the flatness standard deviation, n is the number of measurement points, and h is i is the height of the ith measurement point, is the average height. The larger the flatness standard deviation, the more uneven the step surface is.

[0083] Predicting the balance state change trend based on the environment and the self-posture perception results includes the following steps:

[0084] The actual slope is subtracted from the standard slope to obtain the slope deviation, and the rate of increase of the inclination angle and the rate of decrease of the friction force are calculated within the monitoring period. The logic for obtaining the rate of increase of the inclination angle is as follows: subtract the inclination angle at the current moment from the inclination angle at the previous moment to obtain the inclination angle difference, and compare the inclination angle difference with the previous monitoring period to obtain the rate of increase of the inclination angle. The logic for calculating the rate of decrease of the friction force is as follows: subtract the friction force at the current moment from the friction force at the previous moment to obtain the friction difference, and compare the friction difference with the previous monitoring period to obtain the rate of decrease of the friction force.

[0085] Select n measuring points on a step surface and mark the heights of the n measuring points as h1, h2, ..., h n ; Then the calculation expression of flatness standard deviation is: In the formula, σ is the flatness standard deviation, n is the number of measurement points, and h is i is the height of the ith measurement point, is the average height. The larger the flatness standard deviation, the more uneven the step surface is.

[0086] The slope deviation, flatness standard deviation, inclination angle increase rate, and friction force decrease rate obtained in real time are used as input features, and the input features are input into the trained random forest model. The random forest model outputs the target value, that is, the balance state change trend of the stair climbing load-bearing robot;

[0087] The target value is compared with the preset target threshold. The target threshold is used to predict the trend of equilibrium state change. If the target value is less than or equal to the target threshold, the equilibrium state change trend is predicted to develop toward the equilibrium state. If the target value is greater than the target threshold, the equilibrium state change trend is predicted to develop toward the tilted state.

[0088] Random forest model training consists of the following steps:

[0089] Collect historical data and organize them into training sets, including:

[0090] Input feature subset X = {Δα,σ,ω(t),dF friction}, Δα is the slope deviation, σ is the flatness standard deviation, ω(t) is the rate of increase of the inclination angle, dF friction is the friction force decreasing rate;

[0091] Target value Y = S balance , S balance The predicted value, i.e., the degree of change in the robot's balance state, can be annotated through historical data;

[0092] Random forest will generate multiple decision trees based on random subsets of features. Each decision tree splits and makes decisions based on different random selections. For each decision tree, the feature value (such as the maximum slope deviation, the maximum flatness standard deviation, the maximum rate of increase of the inclination angle, and the maximum rate of decrease of friction) is selected when the node is split, and the decision is made based on the feature value;

[0093] Random forest generates multiple decision trees based on different subsets of input features. When splitting, the decision tree divides the data into different subsets according to different thresholds of the features, and finally generates leaf nodes. Each leaf node represents the prediction result based on the input features.

[0094] In the prediction stage, random forest will perform weighted average or voting on the output results of all decision trees to obtain the final prediction value. Assuming there are T trees, the prediction value is: In the formula, S balance is the predicted value, T is the number of decision trees, S i is the prediction result of the i-th decision tree.

[0095] When it is predicted that the balance state change trend develops toward the tilted state, each driving wheel system of the stair climbing load-bearing robot is automatically controlled to maintain the balance state of the stair climbing load-bearing robot, including the following steps:

[0096] According to the output of the equilibrium state prediction model, determine the tilt direction (such as front, back, left, right);

[0097] The system adjusts the power output of different drive wheels according to the tilt direction. The specific steps are as follows:

[0098] Analyze the tilt angle and decide which drive wheel or wheels need to be adjusted. For example, if the robot tilts to the right, the power of the left drive wheel may need to be increased, while the power of the right drive wheel needs to be reduced; adjust the speed of each drive wheel according to the feedback of different drive wheels. For example, increase the power of the wheel on the opposite side of the tilt direction and reduce the power of the wheel in the tilt direction to restore the robot to balance. For example, when the robot's tilt angle is forward, the speed of the front drive wheel can be appropriately reduced, and the power of the rear drive wheel can be increased to balance the front and rear tilts. If the tilt is caused by a decrease in friction, the contact stability between the robot and the stairs can be increased by adjusting the torque of the drive wheel or decelerating to avoid slipping.

[0099] During the adjustment of the drive wheel system, the robot's posture and tilt angle changes are monitored in real time, new sensor data is acquired, and fed back to the control system. If the robot returns to a balanced state, the adjustment process of the drive wheel ends automatically.

[0100] The gyroscope, accelerometer, inclinometer and other devices installed on the robot provide real-time feedback on the changes in the robot's current posture and evaluate whether it has achieved balance. Based on the feedback results, if the balance has not been restored, continue to fine-tune the power of the drive wheels until it is detected that the robot has returned to a balanced state.

[0101] After adjusting each drive wheel system, the system confirms that the robot is in a balanced state again. When the robot returns to a normal search posture, the system stops correcting the drive wheels and the robot continues its search mission. If the robot still cannot restore balance after multiple adjustments, the system will send out an alarm signal and the robot will automatically shut down to avoid further risk of loss of control. At this time, the system can record fault information and remind search personnel to perform maintenance.

[0102] Embodiment 3: The control system of the stair climbing load-bearing robot with a multi-degree-of-freedom drive wheel train described in this embodiment includes a sub-area division module, an abnormal point marking module, a perception frequency adjustment module, a tilt prediction module, and a control module;

[0103] Sub-area division module: obtains garbage search area information, divides the garbage search area into multiple sub-areas, creates an initial sub-area list after initially numbering each sub-area, obtains the preset search route of the stair climbing load-bearing robot in the sub-area, the preset search route includes the stair climbing route and the equipment search route, and sends the sub-area list to the abnormal point marking module;

[0104] Anomaly marking module: obtains historical garbage search information of the stair climbing load-bearing robot in the sub-area, marks the abnormal points of the search path based on the historical garbage search information, and sends the abnormal points of the search path to the perception frequency adjustment module;

[0105] Perception frequency adjustment module: Whenever the stair climbing load-bearing robot enters an abnormal point in the search path, the perception frequency of the stair climbing load-bearing robot is adjusted, and the perception frequency of the stair climbing load-bearing robot after adjustment is sent to the tilt prediction module;

[0106] Tilt prediction module: The stair climbing load-bearing robot perceives the environment and its own posture according to the adjusted perception frequency, predicts the balance state change trend based on the environment and its own posture perception results, and sends the balance state change trend prediction results to the control module;

[0107] Control module: When it is predicted that the balance state changes in a trend toward a tilted state, each driving wheel system of the stair climbing load-bearing robot is automatically controlled to maintain the balance state of the stair climbing load-bearing robot.

[0108] 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.

[0109] In the description of this specification, the description with reference to the terms "one embodiment", "example", "specific example", etc. means that the specific features, structures, materials or characteristics described in conjunction with the embodiment or example are included in at least one embodiment or example of the present invention. In this specification, the schematic representation of the above terms does not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials or characteristics described can be combined in any one or more embodiments or examples in a suitable manner.

[0110] The preferred embodiments of the present invention disclosed above are only used to help explain the present invention. The preferred embodiments do not describe all the details in detail, nor do they limit the invention to only specific implementation methods. Obviously, many modifications and changes can be made according to the content of this specification. This specification selects and specifically describes these embodiments in order to better explain the principles and practical applications of the present invention, so that those skilled in the art can understand and use the present invention well. The present invention is limited only by the claims and their full scope and equivalents.

Claims

1. A control method for a stair climbing load-bearing robot with a multi-degree-of-freedom drive gear train, characterized in that: The control method comprises the following steps: The control system obtains garbage search area information, divides the garbage search area into multiple sub-areas, initially numbers each sub-area, establishes an initial sub-area list, and for each sub-area, obtains a preset search route of the stair-climbing load-bearing robot in the sub-area; Obtain historical garbage search information of the stair climbing load-bearing robot in the sub-area, mark abnormal points of the search path based on the historical garbage search information, and adjust the perception frequency of the stair climbing load-bearing robot whenever the stair climbing load-bearing robot enters the abnormal point of the search path; The stair-climbing load-bearing robot perceives the environment and its own posture according to the adjusted perception frequency, and predicts the trend of change of the balance state based on the environment and its own posture perception results. When the predicted trend of change of the balance state is toward a tilted state, the driving wheel systems of the stair-climbing load-bearing robot are automatically controlled to maintain the balance state of the stair-climbing load-bearing robot.

2. The control method of the stair climbing load-bearing robot with a multi-degree-of-freedom drive gear train according to claim 1 is characterized in that: Obtaining historical garbage search information of the stair-climbing load-bearing robot in the sub-area, and marking abnormal points of the search path based on the historical garbage search information, including the following steps: When the stair climbing load-bearing robot completes a search in a sub-area, the control system obtains abnormal conditions encountered by the load-bearing robot on the preset search route, including pauses, detours, tilts, and shakes; Obtain the historical garbage search information of the stair climbing load-bearing robot in the sub-area, integrate the multiple garbage search information, and mark the locations where the stair climbing load-bearing robot pauses, detours, tilts or shakes during the search process in the sub-area as initial abnormal points; For each initial outlier point, the number of outlier records of the initial outlier point is obtained. When the number of outlier records of the initial outlier point is less than 3 times, the initial outlier point is deleted from the sub-region model. When the number of outlier records of the initial outlier point is greater than or equal to 3 times, the initial outlier point is updated to a search path outlier point.

3. The control method of the stair climbing load-bearing robot with a multi-degree-of-freedom drive gear train according to claim 2 is characterized in that: Whenever the stair climbing load-bearing robot enters an abnormal point in the search path, the sensing frequency of the stair climbing load-bearing robot is adjusted, including the following steps: When the stair climbing load-bearing robot enters an abnormal point in the search path, the control system calculates the adjustment factor according to the number of abnormal records, and the expression is: In the formula, adjus is the adjustment factor, n is the type of abnormal record, yc i represents the number of occurrences of the i-th abnormal record, ω i Represents the weight of the i-th abnormal record; The sensing frequency of the stair climbing load-bearing robot is adjusted by the obtained adjustment factor. The adjustment algorithm is: new =fq old (1+log5(adjust+1)), where fq new is the adjusted perceived frequency, fq old is the perceived frequency before adjustment, adjus is the adjustment factor; After the perception frequency of the stair-climbing load-bearing robot is corrected by the adjustment algorithm, when the stair-climbing load-bearing robot enters an abnormal point in the search path, it senses according to the adjusted perception frequency corresponding to the abnormal point in the search path.

4. The control method of the stair climbing load-bearing robot with a multi-degree-of-freedom drive gear train according to claim 3 is characterized in that: Predicting the balance state change trend based on the environment and the self-posture perception results includes the following steps: The slope deviation, flatness standard deviation, slope angle increase rate, and friction decrease rate obtained in real time are used as input features, and the input features are input into the trained random forest model, and the random forest model outputs the target value; The target value is compared with the preset target threshold. The target threshold is used to predict the trend of equilibrium state change. If the target value is less than or equal to the target threshold, the equilibrium state change trend is predicted to develop toward the equilibrium state. If the target value is greater than the target threshold, the equilibrium state change trend is predicted to develop toward the tilted state.

5. The control method of the stair climbing load-bearing robot with a multi-degree-of-freedom drive gear train according to claim 4 is characterized in that: Random forest model training consists of the following steps: Collect historical data and organize them into training sets, including: Input feature subset X = {Δα,σ,ω(t),dF friction }, Δα is the slope deviation, σ is the flatness standard deviation, ω(t) is the rate of increase of the inclination angle, dF friction is the friction force decreasing rate; Target value Y = S balance , S balance is the predicted value; Random forests generate multiple decision trees based on random subsets of features. Each decision tree splits and makes decisions based on different random selections. For each decision tree, feature values ​​are selected when nodes split, and decisions are made based on feature values. Random forest generates multiple decision trees based on different subsets of input features. When splitting, the decision tree divides the data into different subsets according to different thresholds of the features, and finally generates leaf nodes. Each leaf node represents the prediction result based on the input features. In the prediction stage, random forest will perform weighted average or voting on the output results of all decision trees to obtain the final prediction value. Assuming there are T trees, the prediction value is: In the formula, S balance is the predicted value, T is the number of decision trees, S i is the prediction result of the i-th decision tree.

6. The control method of the stair climbing load-bearing robot with a multi-degree-of-freedom drive gear train according to claim 5 is characterized in that: The stair climbing load-bearing robot perceives the environment and its own posture according to the adjusted perception frequency, including the following steps: The stair climbing load-bearing robot perceives the environment and its own posture according to the adjusted perception frequency, scans the stairs in front through the laser radar and visual sensor, and measures the slope of the stairs and detects the flatness of the stair tread surface; When the stair climbing load-bearing robot starts climbing, it senses the current tilt angle through the gyroscope and tilt sensor, and obtains the friction between the traveling wheel and the ground through the wheel train sensor; The actual slope is subtracted from the standard slope to obtain the slope deviation, and the rate of increase of the inclination angle and the rate of decrease of the friction force are calculated within the monitoring period. The logic for obtaining the rate of increase of the inclination angle is as follows: subtract the inclination angle at the current moment from the inclination angle at the previous moment to obtain the inclination angle difference, and compare the inclination angle difference with the previous monitoring period to obtain the rate of increase of the inclination angle. The logic for calculating the rate of decrease of the friction force is as follows: subtract the friction force at the current moment from the friction force at the previous moment to obtain the friction difference, and compare the friction difference with the previous monitoring period to obtain the rate of decrease of the friction force. Select n measuring points on a step surface and mark the heights of the n measuring points as h1, h2, ..., h n ; Then the calculation expression of flatness standard deviation is: In the formula, σ is the flatness standard deviation, n is the number of measurement points, and h is i is the height of the ith measurement point, is the average height. The larger the flatness standard deviation, the more uneven the step surface is.

7. The control method of the stair climbing load-bearing robot with a multi-degree-of-freedom drive gear train according to claim 6 is characterized in that: The control system obtains garbage search area information, divides the garbage search area into multiple sub-areas, and after initially numbering each sub-area, establishes an initial sub-area list, including the following steps: Obtaining garbage search area information, including building layout, stair distribution, and load garbage location; Sub-area division includes division based on physical characteristics, division based on search tasks, and division based on distance: Divide by physical features: Divide the garbage search area into multiple sub-areas based on the physical layout of the building, each sub-area corresponds to a floor, a flight of stairs, or the area where a specific equipment is located; Division by search task: Divide the garbage search area into corresponding sub-areas according to the inspection scope of the search task; Divide by distance: divide the garbage search area into several sub-areas based on the distance; Assign a unique number to each sub-area, and record the number, location information, area, and search target parameters of each sub-area in the management system to form an initial sub-area list.

8. A stair climbing load-bearing robot control system with a multi-degree-of-freedom drive gear train, used to implement the control method according to any one of claims 1 to 7, characterized in that: It includes a sub-area division module, an abnormal point marking module, a perception frequency adjustment module, a tilt prediction module, and a control module; Sub-area division module: obtains garbage search area information, divides the garbage search area into multiple sub-areas, creates an initial sub-area list after initially numbering each sub-area, and obtains the preset search route of the stair climbing load-bearing robot in the sub-area; Anomaly marking module: obtains the historical garbage search information of the stair climbing load-bearing robot in the sub-area, and marks the abnormal points of the search path based on the historical garbage search information; Perception frequency adjustment module: Whenever the stair climbing load-bearing robot enters an abnormal point in the search path, the perception frequency of the stair climbing load-bearing robot is adjusted; Tilt prediction module: The stair climbing load-bearing robot perceives the environment and its own posture according to the adjusted perception frequency, and predicts the change trend of the balance state based on the perception results of the environment and its own posture; Control module: When it is predicted that the balance state changes in a trend toward a tilted state, each driving wheel system of the stair climbing load-bearing robot is automatically controlled to maintain the balance state of the stair climbing load-bearing robot.

Images