A posture perception-based electric trolley case control method and system

Abstract

The present application relates to the technical field of intelligent control, in particular to a control method and system for an electric trolley case based on posture sensing. The control system for the electric trolley case based on posture sensing comprises an operation data acquisition module, a normal dragging posture fitting module and a damping torque constraint module. The present application determines the wheel-ground impact time point through vertical acceleration, combines wheel end movement, motor current and motor control input to screen normal dragging segments, constructs a normal dragging posture fitting function, and obtains the normal dragging priori posture at the wheel-ground impact moment. Then, the actual case posture is subtracted to obtain wheel-ground impact response posture data, and a damping torque constraint is constructed accordingly, so as to avoid misjudgment of wheel-ground impact as dragging enhancement, reduce false assistance, torque sudden change, pre-obstacle rush and locked-rotor risk.

Description

Technical Field

[0001] This invention relates to the field of intelligent control technology, specifically to a control method and system for an electric trolley case based on attitude perception. Background Technology

[0002] With the development of smart travel devices, electric trolley cases are increasingly being used in airports, train stations, campuses, industrial parks, and daily commutes. Compared to traditional trolley cases, electric trolley cases provide assistance through a drive wheel motor, reducing the physical exertion required when dragging heavy items. To enable electric trolley cases to automatically adjust the drive output based on the user's dragging behavior, existing technologies typically incorporate attitude sensors, angle sensors, acceleration sensors, motor current detection units, and wheel speed detection units in the case body or handle. By collecting data on the handle angle, case pitch angle, wheel end movement, or motor operating status, the system determines the user's dragging intention and adjusts the drive wheel motor's speed, torque, or level of assistance accordingly.

[0003] In existing attitude-sensing electric suitcase control schemes, a common approach is to use the pitch, tilt, or offset angles of the suitcase or handle as indicators of the user's dragging intention. When the suitcase or handle is detected to be within a preset dragging angle range, the controller determines that the user is dragging the suitcase; when the pitch or tilt angle increases, the controller further concludes that the user's dragging demand has increased and correspondingly increases the motor output to provide greater assistance to the suitcase. This type of control is feasible to a certain extent under conditions of flat surfaces, continuous dragging, and minimal external disturbances.

[0004] However, in actual use, electric suitcases often need to traverse discontinuous road surfaces such as thresholds, brick seams, tactile paving bumps, speed bumps, potholes, step edges, and small stones. Because suitcase wheels typically have small diameters and the suitcase's center of gravity is relatively high, wheel-to-ground impacts are easily generated when the wheels contact these discontinuous road surfaces. These impacts cause sudden changes in the suitcase's pitch angle, pitch velocity, vertical acceleration, wheel-end motion, and motor current. In this situation, the change in the suitcase's posture may not necessarily stem from the user's intention to actively drag it, but rather from external disturbances caused by the abrupt change in the wheel's contact with the ground.

[0005] If the controller still directly adjusts the motor output based on the measured pitch angle or pitch velocity at the moment of wheel-ground impact, it may easily misinterpret the sudden change in attitude caused by the wheel-ground impact as an increase in user drag. For example, when the wheels pass over brick seams, tactile paving bumps, or thresholds, the box may tilt backward momentarily, resulting in an increased pitch angle. If the controller directly interprets this increased pitch angle as the user needing more assistance, it may excessively increase the torque at the drive wheel end, causing the suitcase to experience problems such as incorrect assistance, sudden torque changes, forward lurching after overcoming obstacles, dragging jerks, or even motor stalling, thereby affecting the user's operating safety and dragging stability. Summary of the Invention

[0006] This invention, by collecting sequence data of the box's attitude, vertical acceleration, wheel-end motion, motor current, and motor control input, enables the controller to characterize the dragging state of the electric trolley case from multiple dimensions, including attitude, impact, wheel-end motion, motor load, and control input. By determining the wheel-to-ground impact time point based on the local extrema of vertical acceleration, it can accurately pinpoint the moment of transient disturbance when the wheel passes over discontinuous road surfaces such as thresholds, brick seams, and tactile paving bumps. By performing consistency screening on the enhanced, stable, or weakened responses of wheel-end motion, motor current, and motor control input over time, and eliminating candidate normal dragging segments contaminated by the wheel-to-ground impact time point, it can obtain a fitting function for the normal dragging attitude of the box that more closely approximates the actual normal dragging state. By substituting the wheel-to-ground impact time point... The normal dragging posture fitting function of the box is used to obtain the prior posture data of normal dragging, and the difference between the box posture data corresponding to the wheel-ground impact time point is used to obtain the wheel-ground impact response posture data. This can distinguish the posture changes caused by normal dragging by the user and the posture abrupt changes caused by wheel-ground impact. By using the damping torque constraint to constrain the operation when adjusting the wheel-end torque of the motor corresponding to the drive wheel, the excessive output of wheel-end torque in the energy-boosting direction caused by wheel-ground impact can be limited. This reduces the probability that the controller will misjudge the impact posture change as enhanced dragging by the user, thereby reducing the risk of false assist, torque abrupt change, forward lurching after overcoming obstacles, and motor stall when the electric trolley case passes through complex road surfaces. This improves the dragging stability, control safety, and scene adaptability of the electric trolley case on discontinuous road surfaces such as thresholds, brick seams, tactile paving protrusions, and speed bumps.

[0007] This invention provides a control method for an electric trolley case based on attitude perception, comprising: The following data are collected at the current monitoring time point: the box posture sequence data, vertical acceleration sequence data, wheel end motion sequence data, motor current sequence data, and motor control input sequence data during the movement of the electric trolley case. All of these data are sampled within the time interval between the previous monitoring time point and the current monitoring time point. Each of these data includes several box posture data, vertical acceleration data, wheel end motion data, motor current data, and motor control input data arranged in chronological order. The wheel-to-ground impact time point is determined based on the vertical acceleration sequence data acquired at the current monitoring time point. The normal dragging attitude fitting function of the car body is then determined based on the wheel-to-ground impact time point, the wheel end motion sequence data, the motor current sequence data, and the motor control input sequence data acquired at the current monitoring time point over the time interval. The time difference between the wheel-to-ground impact time point and the previous monitoring time point is substituted into the normal dragging attitude fitting function to obtain the prior attitude data for normal dragging. The difference between the car body attitude data corresponding to the wheel-to-ground impact time point and the prior attitude data for normal dragging is then recorded as the wheel-to-ground impact response attitude data. Based on the wheel-to-ground impact response attitude data, a damping torque constraint is constructed for the motor corresponding to the drive wheel. When the motor corresponding to the drive wheel adjusts the wheel-end torque, the constraint operation is performed through the damping torque constraint.

[0008] Preferably, determining the wheel-to-ground impact time point based on the vertical acceleration sequence data acquired at the current monitoring time point specifically includes the following steps: The sampling time point corresponding to the local maximum value in the vertical acceleration sequence data is recorded as the wheel-ground impact time point. The method for determining the local maximum value in the vertical acceleration sequence data is as follows: take the sampling time point corresponding to the vertical acceleration data in any vertical acceleration sequence data as the positioning time point, and then use the positioning time point as the center to select the vertical acceleration data in the vertical acceleration sequence data with the analysis time window to obtain the local analysis set of vertical acceleration. If the selected vertical acceleration data is not lower than all the vertical acceleration data in the local analysis set of vertical acceleration, the selected vertical acceleration data is regarded as the local maximum value in the vertical acceleration sequence data.

[0009] Preferably, the normal dragging posture fitting function of the box is determined based on the wheel-to-ground impact time point and the wheel-end motion sequence data, motor current sequence data, and motor control input sequence data obtained at the current monitoring time point over a time period. Specifically, this includes the following steps: The sampling time points corresponding to the local maximum or local minimum values ​​in the wheel-end motion sequence data are recorded as wheel speed extreme value time points. Then, based on these wheel speed extreme value time points, the time axis corresponding to the wheel-end motion sequence data is divided into several time segments. If the absolute value of the difference between the wheel-end motion data corresponding to the wheel speed extreme value time points at both ends of a time segment is higher than the wheel speed amplitude threshold (set by the operator), and the wheel-end motion data corresponding to the wheel speed extreme value time point at the left end of the time segment is lower than the wheel-end motion data corresponding to the wheel speed extreme value time point at the right end of the time segment, this time segment is recorded as a wheel speed enhancement segment. If the absolute value of the difference between the wheel end motion data corresponding to the wheel speed extreme points at both ends of the time segment is higher than the wheel speed amplitude threshold, and the wheel end motion data corresponding to the wheel speed extreme point at the left end of the time segment is higher than the wheel end motion data corresponding to the wheel speed extreme point at the right end of the time segment, the time segment is recorded as a wheel speed weakening segment. If the absolute value of the difference between the wheel end motion data corresponding to the wheel speed extreme points at both ends of the time segment is not higher than the wheel speed amplitude threshold, the time segment is recorded as a wheel speed stable segment, and adjacent wheel speed stable segments are merged and updated. The sampling time points corresponding to the local maximum or local minimum values ​​in the motor current sequence data are recorded as current extreme value time points. Then, the time axis corresponding to the motor current sequence data is divided into several time segments based on the current extreme value time points. If the absolute value of the difference between the motor current data corresponding to the current extreme value time points at both ends of the time segment is higher than the current amplitude threshold (set by the operator), and the motor current data corresponding to the current extreme value time point at the left end of the time segment is lower than the motor current data corresponding to the current extreme value time point at the right end of the time segment, the time segment is recorded as a current enhancement segment. If the absolute value of the difference between the motor current data corresponding to the current extreme value time points at both ends of the time segment is higher than the current amplitude threshold, and the motor current data corresponding to the current extreme value time point at the left end of the time segment is higher than the motor current data corresponding to the current extreme value time point at the right end of the time segment, the time segment is recorded as a current weakening segment. If the absolute value of the difference between the motor current data corresponding to the current extreme value time points at both ends of the time segment is not higher than the current amplitude threshold, the time segment is recorded as a current stabilization segment. Adjacent current stabilization segments are merged and the current stabilization segments are updated. The sampling time points corresponding to the local maximum or local minimum values ​​in the motor control input sequence data are recorded as control extreme value time points. Based on these control extreme value time points, the time axis corresponding to the motor control input sequence data is divided into several time segments. If the absolute value of the difference between the motor control input data corresponding to the control extreme value time points at both ends of a time segment is higher than the control amplitude threshold (set by the operator), and the motor control input data corresponding to the control extreme value time point at the left end of the time segment is lower than the motor control input data corresponding to the control extreme value time point at the right end of the time segment, the time segment is recorded as a control enhancement segment. If the absolute value of the difference between the motor control input data corresponding to the control extreme value time points at both ends of a time segment is higher than the control amplitude threshold, and the motor control input data corresponding to the control extreme value time point at the left end of the time segment is higher than the motor control input data corresponding to the control extreme value time point at the right end of the time segment, the time segment is recorded as a control weakening segment. If the absolute value of the difference between the motor control input data corresponding to the control extreme value time points at both ends of a time segment is not higher than the control amplitude threshold, the time segment is recorded as a control stabilization segment. Adjacent control stabilization segments are then merged, and the control stabilization segments are updated. The time axis corresponding to the motor control input sequence data is divided into several local time periods according to the local analysis window. The time length of the local analysis window is set by the operator. If the local time period corresponds to the wheel speed enhancement segment, current enhancement segment, and control enhancement segment, the local time period is recorded as an enhancement segment, and adjacent enhancement segments are merged. If the local time period corresponds to the wheel speed stabilization segment, current stabilization segment, and control stabilization segment, the local time period is recorded as a stabilization segment, and adjacent stabilization segments are merged. If the local time period corresponds to the wheel speed reduction segment, current reduction segment, and control reduction segment, the local time period is recorded as a reduction segment, and adjacent reduction segments are merged. Enhancement segments, stabilization segments, and reduction segments are all candidate normal drag segments. The candidate normal drag segments are filtered based on the wheel-ground impact time point to obtain normal drag segments. Linear fitting is performed on the box posture data in the normal drag segment of the box posture sequence data to construct a fitting function for the normal drag posture of the box.

[0010] Preferably, the candidate normal drag segments are screened based on the wheel-ground impact time point to obtain normal drag segments, specifically including the following steps: If the wheel impact time point appears in the candidate normal drag segment, or if the wheel impact time point appears within the response time window range of the corresponding time period at the left end of the candidate normal drag segment, delete the candidate normal drag segment and record the remaining candidate normal drag segment as a normal drag segment.

[0011] Preferably, the damping torque constraint of the motor corresponding to the drive wheel is constructed based on the wheel-to-ground impact response attitude data, specifically including the following steps: The wheel-to-ground impact response attitude data and vertical acceleration sequence data are concatenated and then fed into the damping torque constraint analysis network for processing, and the corresponding damping torque constraint is output. Training the damping torque constraint analysis network involves the following steps: Several training samples for damping torque constraint analysis were obtained. These training samples included the output of the concatenated wheel-to-ground impact response attitude data and vertical acceleration sequence data. The training samples were then labeled using damping torque constraints. All labeled training samples were combined into a training set for damping torque constraint analysis. The damping torque constraint analysis network was then trained using this training set.

[0012] Preferably, the damping torque constraint analysis network employs a multilayer perceptron.

[0013] This invention provides a posture-aware-based electric trolley case control system, comprising: The data acquisition module is used to collect the following sequence data during the movement of the electric trolley case at the current monitoring time point: case attitude sequence data, vertical acceleration sequence data, wheel end motion sequence data, motor current sequence data, and motor control input sequence data. The vertical acceleration sequence data, wheel end motion sequence data, motor current sequence data, and motor control input sequence data are all sequence data sampled within the time interval between the previous monitoring time point and the current monitoring time point. Furthermore, the case attitude sequence data, vertical acceleration sequence data, wheel end motion sequence data, motor current sequence data, and motor control input sequence data each include several case attitude data, vertical acceleration data, wheel end motion data, motor current data, and motor control input data arranged in chronological order. The normal dragging attitude fitting module is used to determine the wheel-to-ground impact time point based on the vertical acceleration sequence data acquired at the current monitoring time point, and to determine the normal dragging attitude fitting function of the box body based on the response of the wheel-to-ground impact time point, the wheel end motion sequence data, the motor current sequence data, and the motor control input sequence data acquired at the current monitoring time point over a time period. The time difference between the wheel-to-ground impact time point and the previous monitoring time point is substituted into the normal dragging attitude fitting function of the box body to obtain the normal dragging prior attitude data. The difference between the box body attitude data corresponding to the wheel-to-ground impact time point and the normal dragging prior attitude data is recorded as the wheel-to-ground impact response attitude data. The damping torque constraint module is used to construct the damping torque constraint of the motor corresponding to the drive wheel based on the wheel-to-ground impact response attitude data. When the motor corresponding to the drive wheel adjusts the wheel-end torque, the constraint operation is performed through the damping torque constraint.

[0014] The present invention has the following advantages: This invention, by collecting sequence data of the box's attitude, vertical acceleration, wheel-end motion, motor current, and motor control input, enables the controller to characterize the dragging state of the electric trolley case from multiple dimensions, including attitude, impact, wheel-end motion, motor load, and control input. By determining the wheel-to-ground impact time point based on the local extrema of vertical acceleration, it can accurately pinpoint the moment of transient disturbance when the wheel passes over discontinuous road surfaces such as thresholds, brick seams, and tactile paving bumps. By performing consistency screening on the enhanced, stable, or weakened responses of wheel-end motion, motor current, and motor control input over time, and eliminating candidate normal dragging segments contaminated by the wheel-to-ground impact time point, it can obtain a fitting function for the normal dragging attitude of the box that more closely approximates the actual normal dragging state. By substituting the wheel-to-ground impact time point... The normal dragging posture fitting function of the box is used to obtain the prior posture data of normal dragging, and the difference between the box posture data corresponding to the wheel-ground impact time point is used to obtain the wheel-ground impact response posture data. This can distinguish the posture changes caused by normal dragging by the user and the posture abrupt changes caused by wheel-ground impact. By using the damping torque constraint to constrain the operation when adjusting the wheel-end torque of the motor corresponding to the drive wheel, the excessive output of wheel-end torque in the energy-boosting direction caused by wheel-ground impact can be limited. This reduces the probability that the controller will misjudge the impact posture change as enhanced dragging by the user, thereby reducing the risk of false assist, torque abrupt change, forward lurching after overcoming obstacles, and motor stall when the electric trolley case passes through complex road surfaces. This improves the dragging stability, control safety, and scene adaptability of the electric trolley case on discontinuous road surfaces such as thresholds, brick seams, tactile paving protrusions, and speed bumps. Attached Figure Description

[0015] Figure 1 This is a schematic diagram of the electric trolley case control system based on attitude perception used in an embodiment of the present invention. Detailed Implementation

[0016] To enable those skilled in the art to better understand the technical solutions of this invention, the technical solutions of this invention will be clearly and completely described below with reference to the accompanying drawings in the embodiments of this invention.

[0017] Example 1: A control method for an electric trolley case based on attitude perception, comprising: At the current monitoring time point, the following data were collected during the movement of the electric trolley case: case attitude sequence data, vertical acceleration sequence data, wheel end motion sequence data, motor current sequence data, and motor control input sequence data. The case attitude sequence data characterizes the spatial attitude state of the electric trolley case relative to the ground, specifically including pitch angle and pitch rate, which can be obtained through an IMU installed on the case. Vertical acceleration characterizes the vertical motion of the case relative to the ground and can be obtained through three-axis acceleration transformation acquired by the IMU. Wheel end motion sequence data characterizes the motion of the electric trolley case, specifically including the angular velocity of the drive wheel corresponding to the axle, which can be obtained through a magnetic encoder installed on the corresponding axle of the drive wheel. The motor current sequence data characterizes the motion of the drive wheel. The operating status of the drive motor corresponding to the wheel includes the current data of the drive motor. The motor control input sequence data is used to characterize the control status of the drive motor, specifically including the wheel end torque input by the controller of the drive motor to the drive motor. The vertical acceleration sequence data, wheel end motion sequence data, motor current sequence data, and motor control input sequence data are all sequence data sampled within the time interval between the previous monitoring time point and the current monitoring time point. Moreover, the box attitude sequence data, vertical acceleration sequence data, wheel end motion sequence data, motor current sequence data, and motor control input sequence data each include several box attitude data, vertical acceleration data, wheel end motion data, motor current data, and motor control input data arranged in chronological order. The wheel-to-ground impact time point is determined based on the vertical acceleration sequence data acquired at the current monitoring time point. The normal dragging posture fitting function of the suitcase is then determined based on the response of the wheel-to-ground impact time point, the wheel-end motion sequence data, the motor current sequence data, and the motor control input sequence data acquired at the current monitoring time point over a time period. Wheel-to-ground impact refers to the transient disturbance generated when the wheels of the electric suitcase come into contact with discontinuous structures or obstacles on the ground during dragging. For example, when passing over thresholds, brick seams, tactile paving bumps, speed bumps, potholes, step edges, or small stones, the wheels experience vertical impact, sudden changes in rolling resistance, or short-term jamming, further causing abnormal changes in the suitcase's pitch angle, pitch velocity, vertical acceleration, wheel-end motion, and motor current. The time difference between the wheel-to-ground impact time point and the previous monitoring time point is substituted into the normal dragging posture fitting function to obtain the normal dragging prior posture data. The difference between the suitcase posture data corresponding to the wheel-to-ground impact time point and the normal dragging prior posture data is then recorded as the wheel-to-ground impact response posture data. Based on the attitude data of wheel-to-ground impact response, a damping torque constraint is constructed for the motor corresponding to the drive wheel. When the motor corresponding to the drive wheel adjusts the wheel-end torque, the constraint operation is performed through the damping torque constraint. When the wheel-to-ground impact causes a sudden change in pitch angle or pitch rate, the damping torque constraint will reduce the upper limit of the wheel-end torque in the "energy enhancement direction" to prevent the controller from misjudging the attitude change caused by the impact as user drag enhancement, and thus excessively increasing the output of the wheel-end torque. Determining the wheel-to-ground impact time point based on the vertical acceleration sequence data obtained at the current monitoring time point includes the following steps: The sampling time point corresponding to the local maximum value in the vertical acceleration sequence data is recorded as the wheel-ground impact time point. The method for determining the local maximum value in the vertical acceleration sequence data is as follows: take the sampling time point corresponding to the vertical acceleration data in any vertical acceleration sequence data as the positioning time point, and then use the positioning time point as the center to select the vertical acceleration data in the vertical acceleration sequence data with the analysis time window to obtain the local analysis set of vertical acceleration. The time length of the analysis time window is set by the operator. If the selected vertical acceleration data is not lower than all the vertical acceleration data in the local analysis set of vertical acceleration, the selected vertical acceleration data is regarded as the local maximum value in the vertical acceleration sequence data. Based on the wheel-to-ground impact time point and the wheel-end motion sequence data, motor current sequence data, and motor control input sequence data obtained at the current monitoring time point, the normal dragging posture fitting function of the box is determined over the time period. The specific steps include the following: The sampling time point corresponding to the local maximum or local minimum value in the wheel-end motion sequence data is recorded as the wheel speed extreme value time point. The method for determining the local maximum or local minimum value in the wheel-end motion sequence data is as follows: Using the sampling time point corresponding to any wheel-end motion data in the wheel-end motion sequence data as the positioning time point, and then using the positioning time point as the center, a bounding box is drawn around the wheel-end motion data in the wheel-end motion sequence data to obtain the local analysis set of wheel speed. If the selected wheel-end motion data is not lower than all wheel-end motion data in the local analysis set of wheel speed, the selected wheel-end motion data is considered a local maximum value in the wheel-end motion sequence data. If the selected wheel-end motion data is not higher than all wheel-end motion data in the local analysis set of wheel speed, the selected wheel-end motion data is considered a local minimum value in the wheel-end motion sequence data. Then, based on the wheel speed extreme value time point, the time axis corresponding to the wheel-end motion sequence data is divided into several time segments. If the absolute value of the difference between the wheel-end motion data corresponding to the wheel speed extreme points at both ends of the time segment is higher than the wheel speed amplitude threshold (set by the operator), and the wheel-end motion data corresponding to the wheel speed extreme point at the left end of the time segment is lower than the wheel-end motion data corresponding to the wheel speed extreme point at the right end of the time segment, the time segment is recorded as a wheel speed enhancement segment. If the absolute value of the difference between the wheel-end motion data corresponding to the wheel speed extreme points at both ends of the time segment is higher than the wheel speed amplitude threshold, and the wheel-end motion data corresponding to the wheel speed extreme point at the left end of the time segment is higher than the wheel-end motion data corresponding to the wheel speed extreme point at the right end of the time segment, the time segment is recorded as a wheel speed reduction segment. If the absolute value of the difference between the wheel-end motion data corresponding to the wheel speed extreme points at both ends of the time segment is not higher than the wheel speed amplitude threshold, the time segment is recorded as a wheel speed stability segment, and adjacent wheel speed stability segments are merged and updated. The sampling time point corresponding to the local maximum or local minimum value in the motor current sequence data is recorded as the current extreme value time point. The method for determining the local maximum or local minimum value in the motor current sequence data is as follows: Using the sampling time point corresponding to any motor current data in the motor current sequence data as the positioning time point, and centering on the positioning time point, a bounding box is drawn around the motor current data in the motor current sequence data to obtain the local current analysis set. If the selected motor current data is not lower than all motor current data in the local current analysis set, the selected motor current data is considered a local maximum value in the motor current sequence data; if the selected motor current data is not higher than all motor current data in the local current analysis set, the selected motor current data is considered a local minimum value in the motor current sequence data. Then, based on the current extreme value time points, the time axis corresponding to the motor current sequence data is divided into several time segments. If the absolute value of the difference between the motor current data corresponding to the current extreme points at both ends of the time segment is higher than the current amplitude threshold (set by the operator), and the motor current data corresponding to the current extreme point at the left end of the time segment is lower than the motor current data corresponding to the current extreme point at the right end of the time segment, the time segment is recorded as a current enhancement segment. If the absolute value of the difference between the motor current data corresponding to the current extreme points at both ends of the time segment is higher than the current amplitude threshold, and the motor current data corresponding to the current extreme point at the left end of the time segment is higher than the motor current data corresponding to the current extreme point at the right end of the time segment, the time segment is recorded as a current reduction segment. If the absolute value of the difference between the motor current data corresponding to the current extreme points at both ends of the time segment is not higher than the current amplitude threshold, the time segment is recorded as a current stability segment, and adjacent current stability segments are merged and updated. The sampling time point corresponding to the local maximum or local minimum value in the motor control input sequence data is recorded as the control extreme value time point. The method for determining the local maximum or local minimum value in the motor control input sequence data is as follows: Using the sampling time point corresponding to any motor control input data in the motor control input sequence data as the positioning time point, and then using the positioning time point as the center, a bounding box is drawn around the motor control input data in the motor control input sequence data to obtain the control local analysis set. If the selected motor control input data is not lower than all motor control input data in the control local analysis set, the selected motor control input data is considered as the local maximum value in the motor control input sequence data. If the selected motor control input data is not higher than all motor control input data in the control local analysis set, the selected motor control input data is considered as the local minimum value in the motor control input sequence data. Then, the time axis corresponding to the motor control input sequence data is divided according to the control extreme value time point. For several time segments, if the absolute value of the difference between the motor control input data corresponding to the control extreme time points at both ends of the time segment is higher than the control amplitude threshold (set by the operator), and the motor control input data corresponding to the control extreme time point at the left end of the time segment is lower than the motor control input data corresponding to the control extreme time point at the right end of the time segment, the time segment is recorded as a control enhancement segment. If the absolute value of the difference between the motor control input data corresponding to the control extreme time points at both ends of the time segment is higher than the control amplitude threshold, and the motor control input data corresponding to the control extreme time point at the left end of the time segment is higher than the motor control input data corresponding to the control extreme time point at the right end of the time segment, the time segment is recorded as a control reduction segment. If the absolute value of the difference between the motor control input data corresponding to the control extreme time points at both ends of the time segment is not higher than the control amplitude threshold, the time segment is recorded as a control stability segment. Adjacent control stability segments are then merged, and the control stability segments are updated. The time axis corresponding to the motor control input sequence data is divided into several local time periods according to the local analysis window. The time length of the local analysis window is set by the operator. If the local time period corresponds to the wheel speed enhancement segment, current enhancement segment, and control enhancement segment, the local time period is recorded as an enhancement segment, and adjacent enhancement segments are merged. If the local time period corresponds to the wheel speed stabilization segment, current stabilization segment, and control stabilization segment, the local time period is recorded as a stabilization segment, and adjacent stabilization segments are merged. If the local time period corresponds to the wheel speed reduction segment, current reduction segment, and control reduction segment, the local time period is recorded as a reduction segment, and adjacent reduction segments are merged. Enhancement segments, stabilization segments, and reduction segments are all candidate normal drag segments. The candidate normal drag segments are filtered based on the wheel-ground impact time point to obtain normal drag segments. Linear fitting is performed on the box posture data in the normal drag segment of the box posture sequence data to construct the box normal drag posture fitting function; Based on the time point of wheel-ground impact, candidate normal drag segments are filtered to obtain normal drag segments. The specific steps include the following: If the wheel impact time point appears in the candidate normal drag segment, or if the wheel impact time point appears within the response time window range of the corresponding time period at the left end of the candidate normal drag segment, delete the candidate normal drag segment and record the remaining candidate normal drag segment as a normal drag segment. The damping torque constraint of the motor corresponding to the drive wheel is constructed based on the wheel-to-ground impact response attitude data, specifically including the following steps: The wheel-to-ground impact response attitude data and vertical acceleration sequence data are concatenated and then fed into the damping torque constraint analysis network for processing, and the corresponding damping torque constraint is output. The damping torque constraint includes the maximum pitch angle and the maximum pitch angular velocity under wheel-to-ground impact; the damping torque constraint analysis network adopts a multilayer sensor. Training the damping torque constraint analysis network involves the following steps: Several training samples for damping torque constraint analysis were obtained. These training samples included the output of a concatenated data set of wheel-to-ground impact response attitude data and vertical acceleration sequence data. These training samples were constructed by operators based on actual driving experiments of an electric rolling suitcase on discontinuous road surfaces. The training samples were labeled with damping torque constraints designed by operators based on expert experience to prevent excessive increase in wheel-end torque due to wheel-to-ground impact. All labeled training samples were combined into a training set for damping torque constraint analysis. The damping torque constraint analysis network was trained using this training set. During training, the training samples were used as input, and the labeled damping torque constraints were used as the target output. A loss value was constructed based on the difference between the predicted output and the target output of the damping torque constraint analysis network. The MSE (Minimum Sum Equation) approach was used to minimize the loss value. The parameters of the damping torque constraint analysis network were adjusted through gradient backpropagation.

[0018] This application acquires sequence data of the box's attitude, vertical acceleration, wheel-end motion, motor current, and motor control input. This enables the controller to characterize the dragging state of the electric trolley case from multiple dimensions, including attitude, impact, wheel-end motion, motor load, and control input. By determining the wheel-to-ground impact time point based on the local extrema of vertical acceleration, it can accurately locate the moment when transient disturbances occur when the wheels pass over discontinuous road surfaces such as thresholds, brick seams, and tactile paving bumps. By performing consistency screening on the enhanced, stable, or weakened responses of wheel-end motion, motor current, and motor control input over time, and eliminating candidate normal dragging segments contaminated by the wheel-to-ground impact time point, it can obtain a fitting function for the normal dragging attitude of the box that more closely approximates the actual normal dragging state. By substituting the wheel-to-ground impact time point into... The normal dragging posture fitting function of the box is used to obtain the prior posture data of normal dragging, and the difference between the box posture data corresponding to the wheel-ground impact time point is used to obtain the wheel-ground impact response posture data. This can distinguish the posture changes caused by normal dragging by the user and the posture abrupt changes caused by wheel-ground impact. By using the damping torque constraint to constrain the operation when adjusting the wheel-end torque of the motor corresponding to the drive wheel, the excessive output of wheel-end torque in the energy-boosting direction caused by wheel-ground impact can be limited. This reduces the probability that the controller will misjudge the impact posture change as enhanced dragging by the user, thereby reducing the risk of false assist, torque abrupt change, forward lurching after overcoming obstacles, and motor stall when the electric trolley case passes through complex road surfaces. This improves the dragging stability, control safety, and scene adaptability of the electric trolley case on discontinuous road surfaces such as thresholds, brick seams, tactile paving protrusions, and speed bumps.

[0019] Example 2: A control system for an electric trolley case based on attitude perception, such as... Figure 1 As shown, it includes: The data acquisition module is used to collect the following data at the current monitoring time point: suitcase attitude sequence data, vertical acceleration sequence data, wheel end motion sequence data, motor current sequence data, and motor control input sequence data during the electric suitcase's movement. The suitcase attitude sequence data characterizes the spatial attitude state of the electric suitcase relative to the ground, specifically including pitch angle and pitch rate, which can be obtained through an IMU installed on the suitcase. Vertical acceleration characterizes the vertical motion of the suitcase relative to the ground and can be obtained through three-axis acceleration transformation acquired by the IMU. Wheel end motion sequence data characterizes the motion of the electric suitcase, specifically including the angular velocity of the drive wheel corresponding to the axle, which can be obtained through a magnetic encoder installed on the corresponding axle of the drive wheel. The motor current sequence data is obtained through... To characterize the operating status of the drive motor corresponding to the drive wheel, specifically including the drive motor current data, the motor control input sequence data is used to characterize the control status of the drive motor, specifically including the wheel end torque input by the controller of the drive motor to the drive motor; the vertical acceleration sequence data, wheel end motion sequence data, motor current sequence data and motor control input sequence data are all sequence data sampled within the time interval between the previous monitoring time point and the current monitoring time point, and the box attitude sequence data, vertical acceleration sequence data, wheel end motion sequence data, motor current sequence data and motor control input sequence data each include several box attitude data, vertical acceleration data, wheel end motion data, motor current data and motor control input data arranged in chronological order; The normal dragging posture fitting module is used to determine the wheel-to-ground impact time point based on the vertical acceleration sequence data acquired at the current monitoring time point, and to determine the normal dragging posture fitting function of the box based on the response of the wheel-to-ground impact time point, the wheel end motion sequence data, the motor current sequence data, and the motor control input sequence data acquired at the current monitoring time point over a time period. Wheel-to-ground impact refers to the transient disturbance generated when the wheels of the electric trolley case come into contact with discontinuous structures or obstacles on the ground during dragging. For example, when passing through thresholds, brick seams, tactile paving bumps, speed bumps, potholes, step edges, or small stones, the wheels are subjected to vertical impact, sudden changes in rolling resistance, or short-term jamming, which further causes abnormal changes in the box's pitch angle, pitch angular velocity, vertical acceleration, wheel end motion, and motor current. The time difference between the wheel-to-ground impact time point and the previous monitoring time point is substituted into the normal dragging posture fitting function of the box to obtain the normal dragging prior posture data. The difference between the box posture data corresponding to the wheel-to-ground impact time point and the normal dragging prior posture data is recorded as the wheel-to-ground impact response posture data. The damping torque constraint module is used to construct the damping torque constraint of the motor corresponding to the drive wheel based on the wheel-to-ground impact response attitude data. When the motor corresponding to the drive wheel adjusts the wheel-end torque, the constraint operation is performed through the damping torque constraint.

[0020] It should be understood that those skilled in the art can make improvements or modifications based on the above description, and all such improvements and modifications should fall within the protection scope of the appended claims. Parts not described in detail in this specification are prior art known to those skilled in the art.

Claims

1. A control method for an electric trolley case based on attitude perception, characterized in that, include: The following data are collected at the current monitoring time point: the box posture sequence data, vertical acceleration sequence data, wheel end motion sequence data, motor current sequence data, and motor control input sequence data during the movement of the electric trolley case. All of these data are sampled within the time interval between the previous monitoring time point and the current monitoring time point. Each of these data includes several box posture data, vertical acceleration data, wheel end motion data, motor current data, and motor control input data arranged in chronological order. The wheel-to-ground impact time point is determined based on the vertical acceleration sequence data acquired at the current monitoring time point. The normal dragging attitude fitting function of the car body is then determined based on the wheel-to-ground impact time point, the wheel end motion sequence data, the motor current sequence data, and the motor control input sequence data acquired at the current monitoring time point over the time interval. The time difference between the wheel-to-ground impact time point and the previous monitoring time point is substituted into the normal dragging attitude fitting function to obtain the prior attitude data for normal dragging. The difference between the car body attitude data corresponding to the wheel-to-ground impact time point and the prior attitude data for normal dragging is then recorded as the wheel-to-ground impact response attitude data. Based on the wheel-to-ground impact response attitude data, a damping torque constraint is constructed for the motor corresponding to the drive wheel. When the motor corresponding to the drive wheel adjusts the wheel-end torque, the constraint operation is performed through the damping torque constraint.

2. The control method for an electric trolley case based on attitude perception according to claim 1, characterized in that, Determining the wheel-to-ground impact time point based on the vertical acceleration sequence data obtained at the current monitoring time point includes the following steps: The sampling time point corresponding to the local maximum value in the vertical acceleration sequence data is recorded as the wheel-ground impact time point. The method for determining the local maximum value in the vertical acceleration sequence data is as follows: take the sampling time point corresponding to the vertical acceleration data in any vertical acceleration sequence data as the positioning time point, and then use the positioning time point as the center to select the vertical acceleration data in the vertical acceleration sequence data with the analysis time window to obtain the local analysis set of vertical acceleration. If the selected vertical acceleration data is not lower than all the vertical acceleration data in the local analysis set of vertical acceleration, the selected vertical acceleration data is regarded as the local maximum value in the vertical acceleration sequence data.

3. The control method for an electric trolley case based on attitude perception according to claim 2, characterized in that, Based on the wheel-to-ground impact time point and the wheel-end motion sequence data, motor current sequence data, and motor control input sequence data obtained at the current monitoring time point, the normal dragging posture fitting function of the box is determined over a time period. The specific steps include the following: The sampling time points corresponding to the local maximum or local minimum values ​​in the wheel-end motion sequence data are recorded as wheel speed extreme value time points. Then, based on these wheel speed extreme value time points, the time axis corresponding to the wheel-end motion sequence data is divided into several time segments. If the absolute value of the difference between the wheel-end motion data corresponding to the wheel speed extreme value time points at both ends of a time segment is higher than the wheel speed amplitude threshold (set by the operator), and the wheel-end motion data corresponding to the wheel speed extreme value time point at the left end of the time segment is lower than the wheel-end motion data corresponding to the wheel speed extreme value time point at the right end of the time segment, this time segment is recorded as a wheel speed enhancement segment. If the absolute value of the difference between the wheel end motion data corresponding to the wheel speed extreme points at both ends of the time segment is higher than the wheel speed amplitude threshold, and the wheel end motion data corresponding to the wheel speed extreme point at the left end of the time segment is higher than the wheel end motion data corresponding to the wheel speed extreme point at the right end of the time segment, the time segment is recorded as a wheel speed weakening segment. If the absolute value of the difference between the wheel end motion data corresponding to the wheel speed extreme points at both ends of the time segment is not higher than the wheel speed amplitude threshold, the time segment is recorded as a wheel speed stable segment, and adjacent wheel speed stable segments are merged and updated. The sampling time points corresponding to the local maximum or local minimum values ​​in the motor current sequence data are recorded as current extreme value time points. Then, the time axis corresponding to the motor current sequence data is divided into several time segments based on the current extreme value time points. If the absolute value of the difference between the motor current data corresponding to the current extreme value time points at both ends of the time segment is higher than the current amplitude threshold (set by the operator), and the motor current data corresponding to the current extreme value time point at the left end of the time segment is lower than the motor current data corresponding to the current extreme value time point at the right end of the time segment, the time segment is recorded as a current enhancement segment. If the absolute value of the difference between the motor current data corresponding to the current extreme value time points at both ends of the time segment is higher than the current amplitude threshold, and the motor current data corresponding to the current extreme value time point at the left end of the time segment is higher than the motor current data corresponding to the current extreme value time point at the right end of the time segment, the time segment is recorded as a current weakening segment. If the absolute value of the difference between the motor current data corresponding to the current extreme value time points at both ends of the time segment is not higher than the current amplitude threshold, the time segment is recorded as a current stabilization segment. Adjacent current stabilization segments are merged and the current stabilization segments are updated. The sampling time points corresponding to the local maximum or local minimum values ​​in the motor control input sequence data are recorded as control extreme value time points. Based on these control extreme value time points, the time axis corresponding to the motor control input sequence data is divided into several time segments. If the absolute value of the difference between the motor control input data corresponding to the control extreme value time points at both ends of a time segment is higher than the control amplitude threshold (set by the operator), and the motor control input data corresponding to the control extreme value time point at the left end of the time segment is lower than the motor control input data corresponding to the control extreme value time point at the right end of the time segment, the time segment is recorded as a control enhancement segment. If the absolute value of the difference between the motor control input data corresponding to the control extreme value time points at both ends of a time segment is higher than the control amplitude threshold, and the motor control input data corresponding to the control extreme value time point at the left end of the time segment is higher than the motor control input data corresponding to the control extreme value time point at the right end of the time segment, the time segment is recorded as a control weakening segment. If the absolute value of the difference between the motor control input data corresponding to the control extreme value time points at both ends of a time segment is not higher than the control amplitude threshold, the time segment is recorded as a control stabilization segment. Adjacent control stabilization segments are then merged, and the control stabilization segments are updated. The time axis corresponding to the motor control input sequence data is divided into several local time periods according to the local analysis window. The time length of the local analysis window is set by the operator. If the local time period corresponds to the wheel speed enhancement segment, current enhancement segment, and control enhancement segment, the local time period is recorded as an enhancement segment, and adjacent enhancement segments are merged. If the local time period corresponds to the wheel speed stabilization segment, current stabilization segment, and control stabilization segment, the local time period is recorded as a stabilization segment, and adjacent stabilization segments are merged. If the local time period corresponds to the wheel speed reduction segment, current reduction segment, and control reduction segment, the local time period is recorded as a reduction segment, and adjacent reduction segments are merged. Enhancement segments, stabilization segments, and reduction segments are all candidate normal drag segments. The candidate normal drag segments are filtered based on the wheel-ground impact time point to obtain normal drag segments. Linear fitting is performed on the box posture data in the normal drag segment of the box posture sequence data to construct a fitting function for the normal drag posture of the box.

4. The control method for an electric trolley case based on attitude perception according to claim 3, characterized in that, Based on the time point of wheel-ground impact, candidate normal drag segments are filtered to obtain normal drag segments. The specific steps include the following: If the wheel impact time point appears in the candidate normal drag segment, or if the wheel impact time point appears within the response time window range of the corresponding time period at the left end of the candidate normal drag segment, delete the candidate normal drag segment and record the remaining candidate normal drag segment as a normal drag segment.

5. The control method for an electric trolley case based on attitude perception according to claim 4, characterized in that, The damping torque constraint of the motor corresponding to the drive wheel is constructed based on the wheel-to-ground impact response attitude data, specifically including the following steps: The wheel-to-ground impact response attitude data and vertical acceleration sequence data are concatenated and then fed into the damping torque constraint analysis network for processing, and the corresponding damping torque constraint is output. Training the damping torque constraint analysis network involves the following steps: Several training samples for damping torque constraint analysis were obtained. These training samples included the output of the concatenated wheel-to-ground impact response attitude data and vertical acceleration sequence data. The training samples were then labeled using damping torque constraints. All labeled training samples were combined into a training set for damping torque constraint analysis. The damping torque constraint analysis network was then trained using this training set.

6. The control method for an electric trolley case based on attitude perception according to claim 5, characterized in that, The damping torque constraint analysis network employs a multilayer perceptron.

7. A control system for an electric trolley case based on attitude perception, characterized in that, The system employs a posture-aware electric trolley case control method according to any one of claims 1-6, comprising: The data acquisition module is used to collect the following sequence data during the movement of the electric trolley case at the current monitoring time point: case attitude sequence data, vertical acceleration sequence data, wheel end motion sequence data, motor current sequence data, and motor control input sequence data. The vertical acceleration sequence data, wheel end motion sequence data, motor current sequence data, and motor control input sequence data are all sequence data sampled within the time interval between the previous monitoring time point and the current monitoring time point. Furthermore, the case attitude sequence data, vertical acceleration sequence data, wheel end motion sequence data, motor current sequence data, and motor control input sequence data each include several case attitude data, vertical acceleration data, wheel end motion data, motor current data, and motor control input data arranged in chronological order. The normal dragging attitude fitting module is used to determine the wheel-to-ground impact time point based on the vertical acceleration sequence data acquired at the current monitoring time point, and to determine the normal dragging attitude fitting function of the box body based on the response of the wheel-to-ground impact time point, the wheel end motion sequence data, the motor current sequence data, and the motor control input sequence data acquired at the current monitoring time point over a time period. The time difference between the wheel-to-ground impact time point and the previous monitoring time point is substituted into the normal dragging attitude fitting function of the box body to obtain the normal dragging prior attitude data. The difference between the box body attitude data corresponding to the wheel-to-ground impact time point and the normal dragging prior attitude data is recorded as the wheel-to-ground impact response attitude data. The damping torque constraint module is used to construct the damping torque constraint of the motor corresponding to the drive wheel based on the wheel-to-ground impact response attitude data. When the motor corresponding to the drive wheel adjusts the wheel-end torque, the constraint operation is performed through the damping torque constraint.

Images