Real-time monitoring and anti-tipping early warning method for forklift tilt angle based on multi-sensor fusion

Abstract

The present application belongs to the technical field of forklift safety monitoring, and discloses a forklift inclination real-time monitoring and anti-rollover early warning method based on multi-sensor fusion. The front terrain, vehicle operation, vehicle body movement, gantry load, wheel relative contact height and wheel-ground normal force are obtained, and a related contact event group corresponding to the same local elevation feature is established. The change of the reference wheel-ground normal force caused by the vehicle operation and the gantry load is separated, the actual contact time and the actual contact displacement are determined, and the predicted contact event of the wheel that has not passed through the local elevation feature is corrected. The future wheel contact state is recursively calculated, the future effective supporting state is converted by combining the mechanical limiting state of the swing type steering bridge, the supporting reference surface, the supporting unloading margin and the candidate rollover boundary are determined, and the anti-rollover early warning information is output according to the change of the supporting unloading margin and the directed effective inclination.

Description

Technical Field

[0001] This invention belongs to the field of forklift safety monitoring technology, specifically a method for real-time monitoring and anti-tipping early warning of forklift tilt angle based on multi-sensor fusion. Background Technology

[0002] During the operation of a forklift carrying goods, its stability is affected by factors such as travel speed, acceleration and deceleration, steering state, mast tilt angle, fork lifting height, cargo weight, and left and right load distribution. Existing anti-tipping monitoring typically collects vehicle tilt angle, angular velocity, wheel-ground normal force, or mast load, and compares the current monitoring values ​​with preset thresholds, fixed stability boundaries, or pre-established vehicle models. Some solutions also use forward-looking 3D sensors to obtain road surface contours, thereby predicting vehicle posture and wheel load changes when the wheels pass through bumps, depressions, step edges, or slope transition areas.

[0003] When a forklift traverses a road surface with local elevation differences, changes in vehicle acceleration / deceleration, steering, and mast load state will cause load transfer between wheels. The contact between the wheels and the local elevation difference will also cause changes in wheel-to-ground normal force and vehicle motion response. These two types of changes superimpose each other over time. When existing solutions directly identify the road surface effect based on the increase in wheel-to-ground normal force or changes in vehicle posture, it is difficult to determine whether the corresponding change is caused by the vehicle's operating conditions, changes in mast load, or contact with the local elevation difference. Furthermore, the forward-looking terrain data also has position and height measurement errors. When the same local elevation difference acts on different wheels in sequence, the actual contact information formed by the first contacting wheel usually does not correspond to the expected contact event of the subsequent wheels, resulting in the continuous propagation of prediction errors in contact time, contact displacement, and subsequent wheel-to-ground normal force.

[0004] When the sway steering axle of a counterbalance forklift has not reached the mechanical limit, the two steering wheels transmit support to the frame through the axle hinge center. After reaching the mechanical limit, the limit position forms an additional support constraint between the frame and the vehicle body, and the effective support point and effective support area of ​​the vehicle body change accordingly. Using a fixed stable triangle, a fixed support polygon, or judging solely based on the current tilt angle cannot accurately reflect the dynamic changes in wheel contact state, axle support constraint, and tilt boundary during the continuous action of local height difference. This can easily lead to deviations in the calculation of support unloading degree, inaccurate identification of tilt direction, and delayed trend warning. Summary of the Invention

[0005] The technical problem this invention aims to solve is that when a forklift passes through a local elevation difference road surface formed by convex, concave, step edges, or slope transition areas, the changes in wheel-to-ground normal force caused by the vehicle's acceleration / deceleration, steering, and mast load state are superimposed with the changes in wheel-to-ground normal force generated when the wheels pass through the local elevation difference. It is difficult to accurately identify the actual support state changes caused by the local elevation difference based on the wheel-to-ground normal force. At the same time, the position and height errors in the forward-looking terrain data will be transmitted to other wheels that have not yet passed through the same local elevation difference feature. Furthermore, the vehicle body support constraints are different before and after the swing steering axle reaches the mechanical limit. Using the fixed support relationship and the current tilt angle for judgment makes it difficult to reflect the support unloading state and tilting direction during the subsequent wheel contact with the local elevation difference.

[0006] To address the aforementioned technical problems, this invention provides a method for real-time monitoring and anti-tipping warning of forklift tilt angle based on multi-sensor fusion. This method acquires terrain data in front of the forklift, vehicle operation data, vehicle body motion data, mast load data, current relative contact height of each wheel, and current wheel-ground normal force, all with a unified time reference. It also pre-establishes the mapping relationship between wheel contact positions and effective support points of the vehicle body, as well as the adjacency relationship between effective support points of the vehicle body.

[0007] The terrain data in front of the forklift includes the spatial coordinates of the terrain sampling points collected by the forward-looking 3D sensor in the vehicle coordinate system, as well as the terrain normal obtained by fitting adjacent terrain sampling points; the vehicle operation data includes driving speed, driving direction, longitudinal acceleration, lateral acceleration, steering angle and yaw rate; the vehicle body motion data includes vehicle body vertical acceleration, roll rate, pitch rate, roll acceleration, pitch acceleration and vehicle body attitude; the mast load data includes mast tilt angle, fork lifting height, total cargo load and left and right fork loads.

[0008] The wheel-to-ground normal force of each wheel is collected by wheel-mounted sensors. When indirect detection is used, the wheel-to-ground normal force of the corresponding wheel is calculated based on the axle strain, steering axle support force, or the calibration relationship between the support cylinder pressure connected to the wheel support force transmission chain and the wheel-to-ground normal force. The current relative contact height of each wheel is the height of the wheel contact position relative to the reference plane of the vehicle coordinate system, which is determined based on the current vehicle attitude, wheel installation coordinates, wheel effective rolling radius, and the deformation of the wheel ground support structure under the current wheel-to-ground normal force.

[0009] Based on the installation coordinates of each wheel in the vehicle coordinate system, as well as the expected driving speed, expected driving direction and expected steering state in the warning time domain, the movement trajectory of each wheel contact position is determined; the movement trajectory is laterally extended according to the ground contact width of the corresponding wheel to form the predicted wheel track of each wheel.

[0010] Select terrain sampling points located in front of the driving direction in the area to be identified and whose plane fitting residual is not greater than the reference terrain judgment threshold, and fit them to form a reference terrain surface; based on the height deviation of the remaining terrain sampling points from the reference terrain surface, the angle between the terrain normal and the reference terrain surface normal, and the adjacent distance between terrain sampling points, merge the spatially continuous terrain sampling points that meet the terrain change judgment conditions into local elevation difference features.

[0011] When the same local elevation difference feature intersects with the predicted wheel tracks of at least two wheels, the local elevation difference feature is identified as a predicted contact event at the entry boundary, exit boundary, and internal height change position along each predicted wheel track. The predicted contact time is determined based on the position of the predicted contact event in the corresponding predicted wheel track and the expected moving speed of the corresponding wheel.

[0012] For predicted contact events corresponding to the boundary of a local elevation difference feature, the directional height difference between the boundary and the reference terrain surface is determined as the expected contact displacement. For predicted contact events corresponding to the boundary departure and internal elevation change locations, the directional height difference between the current predicted contact event location and the previous predicted contact event location on the same predicted wheel track is determined as the expected contact displacement. When the expected contact displacement is positive, it indicates that the wheel contact position has increased relative to the previous contact state. When the expected contact displacement is negative, it indicates that the wheel contact position has decreased relative to the previous contact state.

[0013] Predicted contact events formed on different wheel tracks with the same local elevation difference feature are arranged according to the expected contact time to form a group of associated contact events. Each predicted contact event in the group of associated contact events corresponds to a determined wheel, expected contact time and expected contact displacement, which is used to establish the event correspondence between the wheel that makes contact first and other wheels that have not yet passed through the same local elevation difference feature.

[0014] On a known smooth road surface, the longitudinal acceleration, lateral acceleration, steering angle, yaw rate, mast tilt angle, fork lifting height, total cargo load, and load difference between the left and right forks of the forklift are changed respectively, and the wheel-to-ground normal force of each wheel is collected. The calibration conditions are divided according to the total cargo load range, fork lifting height range, and steering angle range, and the segmented benchmark response relationship between the vehicle operation data, mast load data and the change of wheel-to-ground normal force of each wheel under each calibration condition is established.

[0015] During forklift operation, based on the current total load, fork lifting height, and steering angle, the corresponding segmented reference response relationship is selected. The current vehicle operation data and mast load data are input into the selected segmented reference response relationship to determine the change in the reference wheel-to-ground normal force of each wheel caused by changes in vehicle acceleration / deceleration, steering, and mast load state when the ground height relationship remains unchanged.

[0016] Based on the predicted contact time of the predicted contact event, the contact time window of the predicted contact event is determined according to the error of the predicted contact position, the error of the predicted moving speed and the error of the data time synchronization. Within the contact time window, the difference between the measured wheel-to-ground normal force change of each wheel and the corresponding reference wheel-to-ground normal force change is determined, and this difference is determined as the contact force increment caused by local height difference contact.

[0017] Based on the position of the wheel corresponding to the predicted contact event relative to the center of mass of the load system formed by the forklift and the goods, and the positive and negative directions of the expected contact displacement, the corresponding expected vehicle motion response direction is determined. When the absolute value of the contact force increment of the corresponding wheel reaches the contact force judgment threshold, and at least one of the vehicle vertical acceleration, roll rate, and pitch rate reaches the corresponding motion response judgment threshold, and the positive and negative directions of the vehicle motion data are consistent with the expected vehicle motion response direction, actual contact is determined to have occurred, and the moment when the above conditions begin to be met simultaneously is determined as the actual contact moment.

[0018] The change in vertical support resultant force is determined by summing the contact force increments of each wheel; the change in roll support moment is determined by the contact force increments of each wheel and the lateral lever arm of the corresponding wheel relative to the center of mass of the load system; the change in pitch support moment is determined by the contact force increments of each wheel and the longitudinal lever arm of the corresponding wheel relative to the center of mass of the load system; the changes in vertical support resultant force, roll support moment, and pitch support moment are normalized according to the standard deviations of the changes in vertical support resultant force, roll support moment, and pitch support moment in the repeated loading calibration test, and the three normalized changes are combined to form the measured contact response vector.

[0019] Through loading tests on wheel ground bearing structures, known contact displacements were applied to each wheel, and the changes in the wheel-ground normal force of each wheel were collected. The changes in the vertical support resultant force, the roll support moment, and the pitch support moment corresponding to the unit contact displacement under different wheel positions and gantry load conditions were determined, and the unit contact displacement response relationship was established.

[0020] When there is only one predicted contact event within the contact time window corresponding to the actual contact moment, the unit contact displacement response vector is selected according to the corresponding wheel and the current gantry load state. The contact displacement that minimizes the squared residual of the product of the unit contact displacement response vector and the contact displacement to be determined relative to the measured contact response vector is then determined as the actual contact displacement.

[0021] When there are more than two predicted contact events within the same contact time window, the unit contact displacement response vectors of the wheels corresponding to each predicted contact event are used to form a contact displacement response matrix. The contact displacement vector that minimizes the squared residual of the product of the contact displacement response matrix and the contact displacement vector to be determined relative to the measured contact response vector is then obtained. Each component in the contact displacement vector is then determined as the actual contact displacement of the corresponding predicted contact event.

[0022] The positive and negative directions of the actual contact displacement are limited to be consistent with the positive and negative directions of the corresponding expected contact displacement. When there are two or more contact displacement vectors that satisfy the minimum residual condition, the contact displacement vector with the smallest sum of squared deviations from the corresponding expected contact displacement vector is selected to determine the actual contact displacement corresponding to each predicted contact event.

[0023] The vehicle's direction of travel is taken as the positive direction of the contact position parameter. The time deviation between the actual contact time and the corresponding expected contact time is determined, and the wheel speed of the corresponding wheel between the two times is integrated to obtain the contact position deviation with a directional sign; when the actual contact time is later than the expected contact time, the contact position deviation is a positive value; when the actual contact time is earlier than the expected contact time, the contact position deviation is a negative value; the contact position deviation is used to correct the boundary position parameter of the local elevation difference feature along the corresponding predicted wheel track direction.

[0024] Determine the displacement deviation between the actual contact displacement and the corresponding expected contact displacement. Using the actual contact position as a constraint, use the displacement deviation to correct the height parameter of the local elevation difference feature at the corresponding predicted wheel track. The position parameter of the local elevation difference feature is used to characterize the position of the entry boundary, exit boundary and internal height change position along the predicted wheel track direction. The height parameter is used to characterize the directional height of the local elevation difference feature at the intersection of each predicted wheel track.

[0025] Based on the contact change position between the corrected local elevation difference feature and the predicted wheel tracks of other wheels that have not yet passed through the local elevation difference feature, the expected contact time and expected contact displacement of the predicted contact events corresponding to the other wheels are re-determined; when actual contact occurs again in the associated contact event group, the predicted contact events corresponding to other wheels that have not yet passed through the local elevation difference feature are updated again based on the contact position deviation and displacement deviation determined by the last actual contact; predicted contact events that have already occurred and predicted contact events that do not belong to the same associated contact event group are not updated with the actual contact result.

[0026] Based on the mast tilt angle, fork lifting height, total cargo load, and left and right fork loads, determine the mass and center of gravity of the load system consisting of the forklift and cargo. Based on the mass, center of gravity, current travel speed range, and fork lifting height range of the load system, select the corresponding response matrix from the segmented wheel contact displacement and wheel-ground normal force response relationship formed by the wheel ground support structure loading test and the local elevation difference road travel test.

[0027] Predicted contact events with overlapping expected contact time windows are grouped into the same state update node; a contact displacement increment vector is established based on the predicted contact events and their corresponding wheels in the state update node, and the expected contact displacement of each predicted contact event is set as the vector component of the corresponding wheel, while the corresponding component of the wheel not included in the state update node is set to zero.

[0028] Input the contact displacement increment vector into the selected response matrix to obtain the wheel-to-ground normal force increment generated by each wheel at the current state update node; based on the time interval between the previous state update node and the current state update node, and according to the dynamic response attenuation relationship formed by the local elevation difference road driving test, attenuate the wheel-to-ground normal force deviation caused by the local elevation difference in the previous future wheel contact state; the dynamic response attenuation relationship is selected based on the load range, driving speed range and the time interval between adjacent state update nodes.

[0029] Based on the vehicle operation data and gantry load data corresponding to the current state update node, the nodal reference wheel-to-ground normal force of each wheel is determined through the segmented reference response relationship; the attenuated wheel-to-ground normal force deviation and the wheel-to-ground normal force increment generated by the current state update node are superimposed on the nodal reference wheel-to-ground normal force to obtain the expected wheel-to-ground normal force of each wheel; the contact displacement increment vector is accumulated to the relative contact height in the previous future wheel contact state to obtain the expected relative contact height of each wheel; the above process is repeated according to the time sequence of each state update node to form the future wheel contact state sequence.

[0030] For wheels that transmit support forces to the frame through rigid axles or independent support structures, the contact position of the wheel is determined as the effective support point of the vehicle body, and the expected relative contact height and expected wheel-to-ground normal force of the wheel are determined as the effective support height and effective support force of the corresponding effective support point of the vehicle body.

[0031] For two steering wheels that transmit support forces to the chassis via a swing steering axle, the expected axle sway angle is determined based on the expected relative contact height of the two steering wheels, the wheelbase between the two steering wheels, the coordinates of the axle hinge center, and the geometric constraints of the swing steering axle.

[0032] When the expected axle sway angle does not reach the axle mechanical limit angle, the projection position of the hinge center of the swing steering axle in the transverse and longitudinal planes of the vehicle body is determined as the plane position of the equivalent support point of the vehicle body. The effective support height of the equivalent support point of the vehicle body is determined based on the expected relative contact height of the two steering wheels and the geometric constraint relationship of the swing steering axle. The sum of the expected wheel-to-ground normal forces of the two steering wheels is determined as the effective support force of the equivalent support point of the vehicle body.

[0033] When the expected axle sway angle reaches the axle mechanical limit angle, the hinge center of the swing steering axle and the limit position where the limit contact occurs are respectively determined as the effective support points of the vehicle body; the effective support height of the hinge center is determined based on the expected relative contact height of the two steering wheels and the geometric dimensions of the swing steering axle, and the effective support height of the limit position is determined based on the geometric coordinates when the limit structure contacts the vehicle frame.

[0034] The sum of the expected wheel-to-ground normal forces of the two steering wheels is taken as the sum of the effective support forces borne by the articulation center and the limiting position. Based on the support torque generated by the two steering wheels relative to the articulation center and the distance between the articulation center and the limiting position, the effective support forces borne by the articulation center and the limiting position are determined according to the support force balance relationship and the support torque balance relationship.

[0035] Based on the effective support points of the vehicle body formed by rigid axles or independent support structures, and the equivalent support points of the vehicle body formed by swing steering axles or the effective support points of the vehicle body after limiting, an effective support area is constructed at the projection position in the transverse and longitudinal planes of the vehicle body. The adjacency relationship between the effective support points of the vehicle body is determined according to the circumferential arrangement order on the boundary of the effective support area, thus forming the corresponding future effective support state.

[0036] The critical effective support force is determined by a calibration test in which the wheel gradually changes from a ground-to-ground state to a ground-to-ground state, and is not less than the equivalent support force corresponding to the upper limit of the wheel-to-ground normal force detection noise. For the vehicle body equivalent support point or limit position, the ground-to-ground judgment force threshold corresponding to the wheel is converted into the corresponding critical effective support force according to the mapping relationship between the wheel contact position and the vehicle body effective support point.

[0037] When there are three effective support points of the car body that are not collinear and the effective support force is not less than the critical effective support force, the support reference plane is determined based on the position and effective support height of the three effective support points of the car body; when there are more than three effective support points of the car body that meet the above conditions, the support reference plane is fitted based on the position and effective support height of each effective support point of the car body; when there are fewer than three effective support points of the car body that meet the above conditions and are not collinear, the corresponding future effective support state is determined as the critical instability state.

[0038] Based on the current vehicle body attitude, vehicle body angular velocity, expected longitudinal acceleration, expected lateral acceleration, expected steering state, and mast load state, and according to the vehicle body attitude response relationship formed by the smooth road test, the predicted vehicle body attitude when the forklift reaches each future effective support state is determined. The vehicle body attitude response relationship is used to determine the changes in vehicle body roll attitude and pitch attitude caused by changes in vehicle acceleration / deceleration, steering, and mast load state, without repeatedly including attitude changes that have already been represented by the effective support height due to local height differences.

[0039] The effective tilt angle is determined based on the lateral and longitudinal attitude differences of the predicted vehicle body attitude relative to the support reference plane; the difference between the effective support force of each effective support point of the vehicle body and the corresponding critical effective support force is determined as the support unloading margin; one or more effective support points of the vehicle body whose difference from the minimum support unloading margin is not greater than the parallel judgment threshold are determined as candidate unloading support points.

[0040] For each candidate unloading support point, the candidate unloading support point is removed in the support reference plane, and the remaining support area is constructed based on the projection position of the remaining effective support points of the vehicle body that maintain effective support. The boundary of the remaining support area outside the candidate unloading support point is determined as the corresponding candidate overturning boundary. When there are two or more remaining support area boundaries that meet the conditions, the remaining support area boundary with the smallest distance from the candidate unloading support point to the boundary is determined as the candidate overturning boundary.

[0041] Determine the direction perpendicular to the candidate overturning boundary and pointing from the inside of the remaining support area to the outside, and determine the projection of the effective tilt angle in this direction as the directional effective tilt angle.

[0042] The first warning threshold, the second warning threshold, and the parallel judgment threshold are determined through forklift tilting test and local elevation difference road driving test, and are set separately according to the total load range, fork lifting height range, driving speed range, and steering angle range. The second warning threshold is less than the first warning threshold.

[0043] When the support unloading margin of any candidate unloading support point decreases sequentially in at least two consecutive future effective support states and is less than or equal to the first warning threshold, and the corresponding directional effective tilt angle increases sequentially in the at least two consecutive future effective support states, trend warning information is output.

[0044] When the unloading margin of any effective support point of the vehicle body is less than or equal to the second warning threshold, or when the corresponding future effective support state is determined to be a critical instability state, an emergency warning message is output.

[0045] When the actual contact time is outside the contact time window of the corresponding predicted contact event, or when the absolute value of the difference between the actual contact displacement and the corresponding expected contact displacement is greater than the correction deviation threshold, the corresponding local elevation difference feature is determined as an unverified local elevation difference feature.

[0046] Based on the current height parameters of unverified local elevation differences, the upper limit of terrain height measurement error, and the lower limit of terrain height measurement error, the first expected contact displacement corresponding to the upper limit of terrain height measurement error and the second expected contact displacement corresponding to the lower limit of terrain height measurement error are determined respectively.

[0047] The corresponding predicted contact events are updated using the first and second predicted contact displacements respectively. Following the same future state recursion and effective support state transition process, a first future effective support state sequence and a second future effective support state sequence are formed.

[0048] In the first and second future effective support state sequences, the minimum support unloading margin of all vehicle body effective support points at all state update nodes is determined respectively. The smaller of the two minimum support unloading margins is determined as the support unloading margin corresponding to the unverified local elevation difference feature, and a warning judgment is performed based on the support unloading margin.

[0049] The beneficial effects of this invention are as follows:

[0050] 1. This invention establishes a segmented benchmark response relationship between vehicle operation data, mast load data, and the reference wheel-to-ground normal force of each wheel. First, it determines the wheel-to-ground normal force changes caused by vehicle acceleration / deceleration, steering, fork lifting, and load offset when the ground height relationship remains constant. Then, it separates the contact force increment corresponding to local height difference contact from the measured wheel-to-ground normal force changes. Combined with the vertical, lateral, and pitch motion responses of the vehicle body, it determines the actual contact time and actual contact displacement, thereby distinguishing between the vehicle's own load transfer and the contact effect of road height difference. This process avoids judging the road effect solely based on sudden changes in wheel-to-ground normal force or the current vehicle body tilt angle, which helps reduce contact misjudgments caused by steering, acceleration / deceleration, and high-position loading conditions. It also provides a data basis with a clear physical source for subsequent correction of local height difference position, height, and wheel support status.

[0051] 2. This invention organizes the entry boundary, exit boundary, and internal height abrupt change positions of the same local elevation difference feature on the predicted wheel tracks of different wheels into a group of associated contact events arranged according to the expected contact time. The position parameters of the local elevation difference feature along the predicted wheel track direction are corrected by the actual contact time deviation, and the height parameters of the local elevation difference feature are corrected by the actual contact displacement deviation. This achieves the purpose of using the actual response of the wheel that has made contact first to successively correct the predicted events of other wheels that have not yet passed through the local elevation difference feature. For the working condition where multiple wheels make contact with the local elevation difference at the same or similar times, the contact displacement of each wheel is jointly determined by the contact displacement response matrix, and the dynamic deviation of the wheel-to-ground normal force is attenuated according to the time interval of the state update node. This ensures that the future relative contact height and the future wheel-to-ground normal force can correspond to the occurrence sequence of contact events and the dynamic response process of the vehicle, reducing the impact of the continuous transmission of forward terrain error between different wheels.

[0052] 3. This invention converts the future wheel contact state into a future effective support state corresponding to the actual support structure of the forklift. Based on whether the swing steering axle has reached the mechanical limit, it forms an equivalent support point of the vehicle body with the axle hinge center, or forms an effective support point of the vehicle body with the axle hinge center and the limit position. This achieves the purpose of updating the effective support area, support reference surface, and candidate overturning boundary as the mechanical support constraint changes. On this basis, the overturning trend is judged by the continuous change of the support unloading margin of the effective support point of the vehicle body and the directional effective tilt angle towards the outside of the candidate overturning boundary. When the local height difference feature has not passed the actual contact verification, different future effective support state sequences are formed based on the upper and lower limits of terrain height measurement error. A smaller support unloading margin is used for judgment, achieving the purpose of early warning that takes into account the support unloading degree, overturning direction, and terrain measurement uncertainty. Attached Figure Description

[0053] Figure 1 This is a flowchart of the real-time monitoring and anti-tipping early warning method for forklift tilt angle based on multi-sensor fusion of the present invention;

[0054] Figure 2 This is a flowchart illustrating the future effective support state transition and early warning judgment process of the present invention.

[0055] Figure 3 This is a flowchart illustrating the construction and correction process for associated contact events in this invention.

[0056] Figure 4 This is a schematic diagram illustrating the technical principle of the forklift tilt angle real-time monitoring and anti-tipping early warning method based on multi-sensor fusion of the present invention. Detailed Implementation

[0057] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0058] like Figures 1 to 4 As shown, this invention provides a method for real-time monitoring and anti-tipping warning of forklift tilt angle based on multi-sensor fusion. This invention is applied to the monitoring of tilt angle and anti-tipping warning of forklifts when they pass through local bumps, depressions, step edges, ditches and slope transition areas in warehouse aisles, factory roads, loading and unloading sites or outdoor hard working surfaces.

[0059] The following embodiment uses a counterbalanced forklift with a rated load of 3000 kg, two drive wheels, and two steering wheels as an example. The two drive wheels transmit the supporting force to the frame through a rigid drive axle, and the two steering wheels are mounted on a swing steering axle. The swing steering axle is connected to the frame through the axle hinge center, and mechanical limit positions are set on the left and right sides of the swing steering axle.

[0060] Example 1: Monitoring device, coordinate system and data synchronization

[0061] The forklift is equipped with a data processing device, a forward-looking 3D sensor, a vehicle operation data interface, a vehicle motion sensor, a mast status sensor, a fork load sensor, and a wheel-to-ground normal force detection device.

[0062] The data processing device receives terrain data in front of the forklift, vehicle operation data, vehicle body motion data, mast load data, and wheel-to-ground normal force data of each wheel, and stores the forklift wheelbase, track width, wheel mounting coordinates, effective rolling radius of the wheels, axle articulation center coordinates, mechanical limit position coordinates, axle mechanical limit angle, and various test calibration data.

[0063] In one computational example, the data update frequency of the forward-looking 3D sensor is 20 Hz, the sampling frequency of vehicle operation data is 100 Hz, the sampling frequency of vehicle body motion data is 200 Hz, the sampling frequency of gantry load data is 50 Hz, and the sampling frequency of wheel-to-ground normal force data is 100 Hz. The data processing device uses a uniform calculation cycle of 10 milliseconds.

[0064] Establish a navigation coordinate system and a vehicle coordinate system; the vertical axis of the navigation coordinate system is aligned with the opposite direction of gravity; the vehicle coordinate system is fixed to the frame, with the longitudinal axis pointing in the forward direction of the forklift, the lateral axis pointing to the left side of the forklift, and the vertical axis pointing above the frame.

[0065] The forward-looking 3D sensor acquires the spatial coordinates of terrain sampling points in the sensor coordinate system; based on the installation position and orientation of the forward-looking 3D sensor relative to the vehicle coordinate system, the terrain sampling points are transformed into the vehicle coordinate system.

[0066] Vehicle operation data includes driving speed, driving direction, longitudinal acceleration, lateral acceleration, steering angle, and yaw rate; vehicle body motion data includes vehicle body vertical acceleration, roll rate, pitch rate, roll acceleration, pitch acceleration, and vehicle body attitude; mast load data includes mast tilt angle, fork lifting height, total cargo load, and left and right fork loads.

[0067] Each data point includes a sampling time. For data with a sampling frequency higher than the unified calculation frequency, the sample value closest to the current calculation time is selected. For data with a sampling frequency lower than the unified calculation frequency, linear interpolation is performed based on two sample values ​​before and after the current calculation time. When the time interval between the two sample values ​​exceeds the maximum interpolation time allowed for the corresponding data, the data at that calculation time is marked as invalid and will not be used to update the predicted contact event.

[0068] The wheel-to-ground normal force is preferentially collected by wheel-mounted sensors installed on the wheel support transmission path; if no wheel-mounted sensors are installed, the wheel-to-ground normal force is calculated based on the calibration relationship between the axle strain and the wheel-to-ground normal force; if the forklift is indeed equipped with a hydraulic support component that transmits the wheel support force, the wheel-to-ground normal force can also be calculated based on the support cylinder pressure, the effective pressure area of ​​the support cylinder, and the mechanical force transmission ratio.

[0069] Let the first The coordinates of the center of each wheel in the vehicle coordinate system are: The attitude transformation matrix from the vehicle coordinate system to the navigation coordinate system is: The vertical unit vector of the navigation coordinate system is , No. The effective rolling radius of each wheel is , No. The wheel ground bearing structure under the current wheel-ground normal force The vertical deformation below is Then the first The vertical coordinates of each wheel contact position relative to the vehicle body origin are:

[0070]

[0071] In the formula, For the first The vertical coordinates of the contact position of each wheel relative to the origin of the vehicle body; For the first The coordinates of the center of each wheel in the vehicle coordinate system; This is the attitude transformation matrix from the vehicle coordinate system to the navigation coordinate system; This is the vertical unit vector of the navigation coordinate system; For the first The effective rolling radius of each wheel; For the first The wheel ground bearing structure under wheel-to-ground normal force Vertical deformation below; superscript This indicates the matrix transpose.

[0072] For wheels mounted on a swing steering axle, the wheel center coordinates change with the axle sway angle. The data processing device updates the coordinates of the two steering wheel centers in the vehicle coordinate system in real time based on the current axle sway angle and the geometric constraints of the swing steering axle.

[0073] The current reference height is the average of the vertical coordinates of all wheel contact points.

[0074]

[0075] In the formula, This is the current reference altitude; This represents the total number of wheels; For the first The vertical coordinates of the contact position of each wheel relative to the origin of the vehicle body.

[0076] No. The current relative contact height of each wheel is:

[0077]

[0078] In the formula, For the first The current relative contact height of each wheel.

[0079] Example 2: Foundation calibration of sensor and wheel grounding support structure

[0080] After the forward-looking 3D sensor is installed, collect no less than 5 sets of terrain data on a known flat road surface; fit a terrain plane based on each set of terrain sampling points, and calculate the height residual from the terrain sampling points to the fitted plane and the change in terrain normal.

[0081] Let the upper limit of the absolute value of the height residual obtained from repeated measurements of a smooth road surface be . The upper limit of the variation of the terrain normal angle is The height change judgment threshold used for local height difference identification Threshold for determining slope changes They respectively satisfy:

[0082]

[0083]

[0084] In the formula, Threshold for determining height changes; Threshold for determining slope changes; This is the upper limit of the measurement error for the height of a smooth road surface; This represents the upper limit of the normal variation in the terrain of a smooth road surface.

[0085] In one calculation example, the upper limit of the absolute value of the height residual of the smooth road surface is 3 mm, the upper limit of the change in the terrain normal angle is 1.5 degrees, the height change judgment threshold is set to 6 mm, and the slope change judgment threshold is set to 3 degrees.

[0086] Loading and unloading tests were conducted on each wheel ground bearing structure. The loading displacements were set sequentially to 0 mm, 10 mm, 20 mm, 30 mm, and 40 mm. Each loading displacement was held for at least 3 seconds, and the wheel-to-ground normal force and the vertical displacement of the wheel center relative to the frame were collected. Each loading displacement was repeated at least 3 times.

[0087] Establish the vertical deformation relationship based on the loaded displacement and the wheel-ground normal force. When the difference in vertical deformation between the loading and unloading processes exceeds the deformation hysteresis judgment value, loading deformation relationship and unloading deformation relationship are established respectively.

[0088] Calibrate the mechanical limit angle of the axle; slowly change the relative height of the left and right steering wheels, and record the axle sway angle when the mechanical limit position begins to contact the frame; record the positive mechanical limit angle and the negative mechanical limit angle respectively.

[0089] In a structural calculation example, the positive mechanical limit angle is 5.0 degrees, the negative mechanical limit angle is -5.0 degrees, the effective lateral distance from the axle hinge center to the left mechanical limit position is 180 mm, and the effective lateral distance from the axle hinge center to the right mechanical limit position is 180 mm.

[0090] Example 3: Formation of Local Elevation Difference Features and Associated Contact Event Groups

[0091] Based on the installation coordinates of each wheel in the vehicle coordinate system, the expected driving speed, the expected driving direction, the steering angle, and the yaw rate, the movement trajectory of each wheel in the warning time domain is determined.

[0092] When the forklift is traveling in a straight line, the wheel movement trajectory extends along the expected travel direction; when the forklift is turning, the instantaneous rotation center of the vehicle is determined based on the steering angle and wheelbase, and then the circular movement trajectory of each wheel is determined based on the position of each wheel relative to the instantaneous rotation center; the movement trajectory is extended to both sides according to the ground contact width of the corresponding wheel to form the predicted wheel track.

[0093] Candidate reference terrain sampling points are selected along the front side of the area to be identified along the driving direction, and a reference terrain surface is fitted:

[0094]

[0095] In the formula, , and These represent the longitudinal, lateral, and vertical coordinates of the terrain sampling point in the vehicle coordinate system, respectively. The longitudinal slope coefficient of the reference terrain surface; The lateral slope coefficient of the reference terrain surface; The intercept is used as a reference terrain surface.

[0096] No. The directional height deviation of each terrain sampling point relative to the reference terrain surface is:

[0097]

[0098] In the formula, , and For the first Spatial coordinates of each terrain sampling point; For the first The directional height deviation of each terrain sampling point relative to the reference terrain surface.

[0099] No. The angle between the unit terrain normal at each terrain sampling point and the unit normal of the reference terrain surface is:

[0100]

[0101] In the formula, For the first Unit terrain normal of each terrain sampling point; The unit normal for the reference terrain surface; It is the angle between two unit normals.

[0102] when Not less than the height change determination threshold ,or Not less than the threshold for judging slope change At that time, the first The topographic sampling points were determined as topographic change sampling points.

[0103] Topographic change sampling points with a spatial distance not greater than the adjacency distance threshold and continuous height change direction are merged to form local elevation difference features; the adjacency distance threshold is determined based on the average spacing of the topographic sampling points; in a calculation example, the average spacing of the topographic sampling points is 10 mm, and the adjacency distance threshold is set to 20 mm.

[0104] When the same local elevation feature intersects with the predicted wheel tracks of at least two wheels, the entry boundary, exit boundary, and internal height change position of the local elevation feature along each predicted wheel track are determined respectively.

[0105] Each entry point, exit point, and internal height change location is identified as a predicted contact event. Each predicted contact event is recorded with local height difference feature identifier, corresponding wheel identifier, event type, contact change location, expected contact time, and expected contact displacement.

[0106] Let the wheel track length from the predicted contact event to the current contact position of the corresponding wheel be... The predicted speed of the wheel along the predicted wheel track is The current time is Then the expected contact time satisfy:

[0107]

[0108] In the formula, To predict the wheel track length from the contact event to the current contact position of the corresponding wheel; For the corresponding wheel in time The predicted speed of movement along the predicted wheel track; The current moment; This represents the expected contact time.

[0109] For predicted contact events corresponding to entering the boundary, the directional height difference between the entering boundary and the reference terrain surface is determined as the expected contact displacement; for predicted contact events corresponding to leaving the boundary and internal height abrupt change locations, the directional height difference between the current predicted contact event location and the previous predicted contact event location of the same wheel is determined as the expected contact displacement.

[0110] When the expected contact displacement is positive, it indicates that the wheel contact position has increased; when the expected contact displacement is negative, it indicates that the wheel contact position has decreased. By setting entry and exit events separately, the relative contact height that increases after the wheel enters the local height difference area can be restored accordingly when it leaves the local height difference area.

[0111] All predicted contact events corresponding to the same local elevation difference feature are arranged from earliest to latest according to the expected contact time to form a group of associated contact events; different groups of associated contact events are established for different local elevation difference features.

[0112] Example 4: Calibration of the normal force response relationship between the reference wheel and the ground

[0113] On a flat road surface without local elevation differences, the forklift is calibrated under baseline operating conditions; the calibration conditions include at least:

[0114] The total cargo load is 0 kg, 500 kg, 1000 kg, 1500 kg, 2000 kg and 2500 kg;

[0115] Fork lifting heights are 300 mm, 1000 mm, 1800 mm and 2500 mm;

[0116] The gantry tilt angles are -3 degrees, 0 degrees, and 3 degrees;

[0117] The load differences between the left and right forks are 0%, 10%, and 20% of the total cargo load, respectively.

[0118] The driving speeds are 0.5 meters per second, 1.0 meter per second, and 1.5 meters per second;

[0119] Steering angles are 0 degrees, 10 degrees, and 20 degrees;

[0120] The longitudinal and lateral accelerations are set in stages within the allowable operating range of the forklift.

[0121] The above parameter combinations do not need to be fully permuted; representative calibration points covering no load, partial load, near rated load, low-lift, high-lift, straight-line travel, and turning travel can be selected according to the forklift's permissible working conditions.

[0122] The reference response input vector is composed of the vehicle operating state and the gantry load state:

[0123]

[0124] In the formula, It is longitudinal acceleration; It is lateral acceleration; This refers to the steering angle; This refers to the yaw rate; For the gantry at a forward angle; This refers to the lifting height of the forks. Total cargo load; This represents the difference in load between the left and right forks.

[0125] The calibration areas are divided according to the total cargo load range, fork lifting height range, and steering angle range, and segmented reference response relationships are established:

[0126]

[0127] In the formula, This is the vector composed of the reference wheel-ground normal forces of each wheel; For the first Each calibration area corresponds to

[0128] The segmented reference response relationship; This is the current calibration area number.

[0129] Adjacent calculation time and The change in the normal force between the reference wheels and the ground is:

[0130]

[0131] In the formula, This is the reference response input vector at the current calculation time. This is the reference response input vector from the previous calculation time step; The change in the normal force on the reference wheel is the amount of force.

[0132] Let the measured change in the wheel-to-ground normal force be... Then the increment of the contact force is:

[0133]

[0134] In the formula, This is to measure the change in the normal force between the wheel and the ground. This is the incremental contact force after deducting the effects of vehicle operating status and gantry load status.

[0135] Example 5: Calibration of contact displacement response relationship

[0136] When the forklift is stationary or in a stable low-speed state, a known vertical contact displacement is applied to each wheel; the vertical contact displacement can be formed by a loading platform of known height, a lifting test platform, or a calibration block of known height.

[0137] In a calibration example, positive contact displacements of 10 mm, 20 mm, 30 mm and 40 mm, and negative contact displacements of -10 mm, -20 mm and -30 mm were applied, respectively; positive contact displacement indicates that the wheel contact position is raised, and negative contact displacement indicates that the wheel contact position is lowered.

[0138] Calibration was performed under conditions with a total load of 500 kg, 1500 kg, and 2500 kg, and fork lifting heights of 300 mm, 1500 mm, and 2500 mm, respectively; each condition was repeated at least 3 times.

[0139] The change in vertical support resultant force is determined by the sum of the incremental contact forces of each wheel:

[0140]

[0141] In the formula, This represents the change in the resultant force of the vertical support. For the first The increase in contact force of each wheel; This represents the total number of wheels.

[0142] The change in the tilt support moment is:

[0143]

[0144] The change in pitch support moment is:

[0145]

[0146] In the formula, This represents the change in the tilt support moment. This represents the change in pitch support torque. For the first The longitudinal lever arm of each wheel contact position relative to the center of mass of the load system formed by the forklift and the goods; For the first The lateral lever arm of each wheel contact position relative to the center of mass of the load system.

[0147] The standard deviations of the changes in the resultant force of the vertical support, the change in the tilt support moment, and the change in the pitch support moment were determined through repeated loading tests. , and The measured contact response vector is:

[0148]

[0149] In the formula, This is the normalized measured contact response vector; This represents the standard deviation of the variation in the resultant force of the vertical support. This represents the standard deviation of the variation in the tilt support moment. This represents the standard deviation of the pitch support moment variation.

[0150] Using the same normalized scale, the changes in vertical support resultant force, tilt support moment, and pitch support moment generated by unit contact displacement are combined to form a unit contact displacement response vector.

[0151] Unit contact displacement response vectors are established for different wheel sizes, different total load ranges of goods, and different fork lifting height ranges, instead of using a single fixed response vector to cover all working conditions.

[0152] Example 6: Determination of contact time window and actual contact time

[0153] Let the upper limit of the positional error of the local elevation feature along the predicted wheel track direction be . The lower limit of the expected moving speed of the wheel during the contact period is: The upper limit of data time synchronization error is Then the half-width of the contact time window is:

[0154]

[0155] In the formula, The contact time window is half the width; This is the upper limit of the position error; This represents the lower limit of the projected movement speed; This represents the upper limit of data time synchronization error.

[0156] In a calculation example, the upper limit of positional error is 50 mm, the lower limit of expected moving speed is 0.8 m / s, and the upper limit of time synchronization error is 20 ms. Then the half-width of the contact time window is:

[0157]

[0158] The contact time window width was set to 0.09 seconds during the calculation.

[0159] Through repeated driving tests on smooth roads, the noise range of contact force increment, vehicle vertical acceleration, roll rate, and pitch rate was statistically analyzed.

[0160] In one calculation example, the noise limit for the absolute value of the contact force increment is 0.08 kN, and the contact force judgment threshold is set to 0.12 kN; the upper limit for the vertical acceleration fluctuation of the vehicle body is 0.10 m / s², and the vertical acceleration judgment threshold of the vehicle body is set to 0.15 m / s².

[0161] When the contact force condition and at least one vehicle motion response condition are maintained continuously for 3 unified calculation cycles, the actual contact condition is determined to be valid; when the unified calculation cycle is 10 milliseconds, the condition is maintained continuously for 30 milliseconds.

[0162] Example 7: Inversion of actual contact displacement corresponding to a single predicted contact event

[0163] When there is only one predicted contact event within the contact time window, the actual contact displacement is determined according to the least squares relation with signed constraints:

[0164]

[0165] And satisfy:

[0166]

[0167] In the formula, For the first The unit contact displacement response vector corresponding to each wheel; The contact displacement is yet to be determined; This represents the actual contact displacement. For the expected contact displacement; The L2 norm of a vector.

[0168] When the unit contact displacement response vector is not zero, the unconstrained least squares solution is:

[0169]

[0170] In the formula, superscript This indicates the matrix transpose.

[0171] When the positive and negative directions of the unconstrained least squares solution are inconsistent with the predicted contact displacement, the solution is not used directly; first, check whether the actual contact corresponds to the original predicted contact event; if the correspondence cannot be confirmed, mark the local elevation difference feature as an unverified local elevation difference feature.

[0172] Example 8: Inversion of actual contact displacements corresponding to multiple predicted contact events

[0173] When there are n predicted contact events within the same contact time window, the unit contact displacement response vectors corresponding to each predicted contact event are combined to form a contact displacement response matrix:

[0174]

[0175] The actual contact displacement vector is determined by the following formula:

[0176]

[0177] And satisfy:

[0178]

[0179] In the formula, The contact displacement vector to be determined; The first contact displacement vector to be determined One component; For the first The predicted contact displacement for each predicted contact event; This is the actual contact displacement vector.

[0180] When multiple solutions with the same minimum residual exist, first determine the set of solutions that satisfy the minimum measured response residual and the sign constraint. Then determine the unique solution according to the following formula:

[0181]

[0182] In the formula, This represents the expected contact displacement vector; The solution set that satisfies the minimum measured response residual and the sign constraint.

[0183] When the number of predicted contact events is greater than the number of independent components that the measured contact response vector can distinguish, and the above-mentioned solution selection relationship still cannot form a credible unique solution, the corresponding predicted contact events are marked as joint unverified events, and the uncertain contact displacement is not used to implement single-value correction for local elevation features.

[0184] Example 9: Correction of position and height parameters

[0185] The direction of vehicle travel along the predicted wheel track is taken as the positive direction of the position parameter. Let the actual contact time be... The expected contact time is The corresponding directional movement speed of the wheel along the predicted wheel track is Then the contact position deviation is:

[0186]

[0187] In the formula, For contact position deviation; The actual moment of contact; For the expected contact time; This corresponds to the directional movement speed of the wheel along the predicted wheel track.

[0188] The corrected contact change position parameters are:

[0189]

[0190] In the formula, The contact change position parameters before correction; These are the corrected contact change position parameters.

[0191] The displacement deviation between the actual contact displacement and the expected contact displacement is:

[0192]

[0193] In the formula, This refers to displacement deviation; This represents the actual contact displacement. This is the expected contact displacement.

[0194] The height parameter at the actual contact point is based on The correction weights for other height sampling points within the same local elevation difference feature are as follows:

[0195]

[0196] In the formula, For the first Corrected weights for each height sampling point; For the first The lateral distance between each height sampling point and the actual contact position; To correct the effective range; For the first Unit terrain normal at each height sampling point; The unit terrain normal at the actual contact location; the first The corrected height of each height sampling point is:

[0197]

[0198] In the formula, The height parameters before correction; This is the corrected height parameter.

[0199] The correction only applies to continuous areas with the same local elevation difference feature. Areas where the difference in terrain normal exceeds the threshold for determining slope change are not corrected based on the actual contact results.

[0200] Example 10: Calibration of the future wheel contact state response matrix

[0201] Within different load system mass ranges, center of gravity position ranges, travel speed ranges, and fork lifting height ranges, wheel contact displacement and wheel-to-ground normal force response matrices were established through wheel ground support structure loading tests and local elevation difference road travel tests.

[0202] Suppose the forklift has 4 wheels, and the wheel contact displacement increment vector is:

[0203]

[0204] In the formula, each component is respectively the first... Each state update node corresponds to the expected contact displacement of the wheel.

[0205] In a sample response matrix used to illustrate the computational relationship:

[0206]

[0207] The unit of each element in the response matrix is ​​kilonewtons per millimeter. This matrix is ​​only used to illustrate the load redistribution relationship. The specific matrix in the formal application document should be determined by the corresponding prototype test.

[0208] When both front wheels are raised by 30 millimeters simultaneously:

[0209]

[0210] The increment of the wheel-to-ground normal force at the current node is:

[0211] Qianniu

[0212] The calculation results indicate that when the contact positions of the left and right front wheels rise simultaneously, the normal force of the front wheels on the ground increases, while the normal force of the rear wheels on the ground decreases accordingly.

[0213] Example 11: Calibration of dynamic response attenuation relationship

[0214] In a driving test on a road surface with local elevation differences, a single wheel is driven over a calibrated protrusion of known height, and the dynamic deviation of the wheel-ground normal force after the wheel leaves the protrusion is recorded over time.

[0215] The tests were repeated in different total load ranges, fork lifting height ranges, and travel speed ranges, and the response decay time constant was fitted.

[0216] No. The first wheel in the The attenuation coefficient under each calibration condition is:

[0217]

[0218] In the formula, For the first The first wheel in the The dynamic response decay coefficient of each state update node; The time interval between the previous state update node and the current state update node; For the first The first wheel in the The response decay time constant under a specified operating condition.

[0219] In a calculation example, the response decay time constant is 0.28 seconds, and the time interval between adjacent state update nodes is 0.15 seconds. Therefore, the dynamic response decay coefficient is:

[0220]

[0221] The attenuation matrix is ​​composed of the attenuation coefficients of each wheel. The dynamic deviation of the wheel-ground normal force is recursively calculated according to the following formula.

[0222]

[0223] In the formula, The dynamic deviation of the wheel-to-ground normal force of the previous state update node; Update the dynamic deviation of the wheel-to-ground normal force of the node in the current state; The increment of the wheel-to-ground normal force generated by the node in the current state update.

[0224] The expected ground normal force of the current state update node is:

[0225]

[0226] In the formula, Update the node reference wheel-ground normal force corresponding to the current state node; To predict the normal force on the wheel.

[0227] Example 12: Determination of the support force of the swing steering axle after reaching the mechanical limit

[0228] Let the effective wheelbase between the centers of the left and right steering wheels be... The center height of the left steering wheel is The center height of the right steering wheel is Then the expected axle sway angle of the symmetrical swing steering axle is:

[0229]

[0230] In the formula, To predict the axle sway angle; The effective wheelbase between the centers of the two steering wheels; This is the center height of the left steering wheel; This is the center height of the right steering wheel.

[0231] For forklifts where the axle hinge center is not located at the midpoint of the wheel track or where the axle structure is asymmetrical, a calibration function between the axle swing angle and the center height of the left and right steering wheels is established based on the rigid body geometry of the axle, instead of directly using the formula for symmetrical structures mentioned above.

[0232] In a calculation example, the effective wheelbase between the centers of the left and right steering wheels is 1000 mm, and the center of the right steering wheel is 100 mm higher than the center of the left steering wheel. Therefore, the expected axle sway angle is:

[0233]

[0234] When the mechanical limit angle of the axle is 5.0 degrees, it is determined that the swing steering axle has reached the mechanical limit.

[0235] Let the sum of the expected wheel-to-ground normal forces of the two steering wheels be... The combined bearing moment generated by the two steering wheels relative to the axle hinge center is The effective lateral distance between the mechanical limit position and the center of the axle hinge is The effective supporting force borne by the mechanical limit position is The effective support borne by the axle articulation center

[0236] Force ,but:

[0237]

[0238]

[0239] In one calculation example, the expected wheel-to-ground normal force of the left steering wheel is 4.2 kN, and the expected wheel-to-ground normal force of the right steering wheel is 3.1 kN. The combined bearing moment generated by the two steering wheels relative to the axle hinge center is 0.55 kNm, and the effective lateral distance from the mechanical limit position to the axle hinge center is... ,but:

[0240]

[0241]

[0242] Therefore, the 4.24 kN effective support force borne by the axle articulation center and the 3.06 kN effective support force borne by the mechanical limit position are respectively written into the future effective support state.

[0243] Example 13: Calibration of Critical Effective Support Force and Early Warning Threshold

[0244] Slowly increase the tilt angle of the support surface on the forklift tilting platform and record the wheel-to-ground normal force when each wheel changes from a stable grounding state to an intermittent grounding state and a continuous off-ground state.

[0245] With the forklift stationary and all wheels stably grounded, collect wheel-to-ground normal force data for no less than 30 seconds to determine the upper limit of wheel-to-ground normal force detection noise.

[0246] The larger of the ground-leaving determination force threshold corresponding to continuous ground-leaving states and the equivalent support force corresponding to the upper limit of wheel-to-ground normal force detection noise is determined as the critical effective support force.

[0247] In a calculation example, the equivalent support force corresponding to the upper limit of the wheel-to-ground normal force detection noise is 0.08 kN, and the ground clearance threshold corresponding to the continuous ground clearance state is 0.12 kN. Therefore, the critical effective support force is taken as 0.12 kN.

[0248] The first and second warning thresholds were determined through tilt table tests and driving tests on roads with local elevation differences. The first warning threshold corresponds to a state in which the wheel support force has been continuously decreasing but has not yet approached the ground. The second warning threshold corresponds to a state in which the wheel is approaching continuous ground lift.

[0249] In a calculation example, the first warning threshold is 0.35 kN, the second warning threshold is 0.15 kN, and the parallel judgment threshold is 0.05 kN. The above values ​​should be established according to parameter tables based on the forklift load range, fork lifting height range, travel speed range, and steering angle range.

[0250] Example 14: Determination of Support Reference Surface and Candidate Overturning Boundary

[0251] When there are 3 effective support points for the car body that are not collinear and maintain effective support, the support reference plane is uniquely determined based on the 3 effective support points of the car body.

[0252] When there are more than 3 effective support points on the car body that are not collinear and maintain effective support, the support reference plane is fitted using a weighted least squares method:

[0253]

[0254] In the formula, To maintain the number of effective support points for the vehicle body; , and For the first Spatial coordinates of each effective support point of the vehicle body; According to the first The fitting weights are determined by the effective support force of each effective support point of the vehicle body. , and These are the parameters of the support reference surface.

[0255] The effective tilt vector of the vehicle body's predicted attitude relative to the supporting reference plane is:

[0256]

[0257] In the formula, The effective tilt angle vector; Predict the pitch angle for the vehicle body; Predict the roll angle for the vehicle body; The longitudinal inclination angle of the supporting reference plane; The lateral tilt angle of the supporting reference plane.

[0258] No. The unloading margin for each effective support point of the vehicle body is:

[0259]

[0260] In the formula, For the first Support unloading margin at each effective support point of the vehicle body. To provide its effective supporting force; It is its critical effective supporting force.

[0261] Candidate unloading support points should be located on the boundary of the effective support area. When the effective support point of the vehicle body with the smallest support unloading margin is located inside the effective support area, removing this point will not change the outer boundary of the effective support area. This internal support point will not be used as a separate candidate point for the rollover boundary, but its support unloading status will continue to be monitored.

[0262] For a candidate unloading support point located on the boundary of the effective support area, the point is removed from the set of effective support points of the vehicle body that maintain effective support. The remaining support area is constructed based on the remaining effective support points of the vehicle body, and the boundary of the remaining support area outside the candidate unloading support point is determined as the candidate overturning boundary.

[0263] Let the unit direction vector perpendicular to the candidate overturning boundary and pointing from the inside of the remaining support region to the outside be:

[0264]

[0265] In the formula, and These are the components of the unit direction vector in the longitudinal and lateral directions of the vehicle body, respectively.

[0266] The effective dip angle is:

[0267]

[0268] In the formula, The directional effective tilt angle is oriented towards the outside of the candidate overturning boundary.

[0269] Example 15: Complete calculation process of a straight line passing through a transverse protrusion

[0270] This embodiment illustrates the process of correcting other wheels that have not yet passed through the same local height difference feature after a single wheel has made actual contact.

[0271] The forklift is carrying 1600 kg of goods, with a fork lifting height of 1200 mm, and is traveling in a straight line at a speed of 1.00 m / s. The forward-looking 3D sensor detects a lateral local bulge.

[0272] The distance from the left front wheel's entry boundary to the current left front wheel contact position is 1.80 meters, and the distance from the left rear wheel's entry boundary to the current left rear wheel contact position is 3.45 meters. The expected contact displacement for both the left front and left rear wheels is +25 millimeters.

[0273] The expected contact time for the left front wheel is 1.80 seconds after the current moment, and the expected contact time for the left rear wheel is 3.45 seconds after the current moment.

[0274] The measured wheel-to-ground normal force of the left front wheel increases by 1.55 kN within the contact time window. Based on the current vehicle operating data and gantry load data, the reference wheel-to-ground normal force of the left front wheel increases by 0.30 kN according to the segmented reference response relationship. Therefore, the increment of the contact force of the left front wheel is:

[0275] 1.55 − 0.30 = 1.25 kN

[0276] The actual contact time of the left front wheel is 1.86 seconds after the current time, which is 0.06 seconds later than the expected contact time. The average speed of the forklift during this period is 1.00 meters per second, so the contact position deviation is +60 millimeters.

[0277] Based on the unit contact displacement response vector and the measured contact response vector of the left front wheel, the actual contact displacement of the left front wheel is determined to be +29 mm. The displacement deviation between the actual contact displacement and the expected contact displacement is:

[0278] 29 − 25 = 4 mm

[0279] The local bulge is corrected forward by 60 mm along the predicted wheel track direction of the left front wheel, and the height of the local bulge at the predicted wheel track of the left front wheel is increased by 4 mm.

[0280] Based on the intersection of the corrected local protrusion and the predicted wheel track of the left rear wheel, the expected contact time of the left rear wheel is corrected from 3.45 seconds to 3.51 seconds, and the expected contact displacement of the left rear wheel is corrected from ±25 mm to ±29 mm.

[0281] The correction does not apply to the predicted contact event of the left front wheel that has already occurred, nor does it apply to other local height difference features that do not belong to this local bulge.

[0282] Example 16: Calculation process for the overlap of two front wheel contact time windows

[0283] The forklift passes diagonally along the edge of the step, and the contact time windows of the predicted contact events for the left and right front wheels overlap.

[0284] In a calculation example, the normalized unit contact displacement response vectors for the left front wheel and the right front wheel are as follows:

[0285]

[0286]

[0287] Among them, each vector represents the normalized vertical support resultant force, roll support moment and pitch support moment response formed by the corresponding wheel contact displacement in units of 1 mm.

[0288] The contact displacement response matrix is ​​formed by combining the two unit contact displacement response vectors:

[0289]

[0290] The normalized measured contact response vector obtained from actual data collection is:

[0291]

[0292] Based on the least squares relation with signed constraints, the actual contact displacement of the left front wheel is +32 mm, and the actual contact displacement of the right front wheel is +30 mm. Substituting these values ​​into the contact displacement response matrix, we obtain:

[0293]

[0294] This allows for the determination of the actual contact displacement of the two front wheels separately, avoiding the attribution of the same measured contact response entirely to a single wheel.

[0295] Example 17: Early warning process when the swing steering axle reaches the mechanical limit

[0296] The forklift enters the single-sided depression at an angle, with the right steering wheel descending 100 mm relative to the left steering wheel. The effective wheelbase between the centers of the two steering wheels is 1000 mm. Based on the expected relative contact height, the axle swing angle is determined to be 5.74 degrees, which is greater than the mechanical limit angle of the axle of 5.0 degrees. Therefore, it is determined that the right mechanical limit position of the swing steering axle is in contact with the frame.

[0297] The expected wheel-to-ground normal force of the left steering wheel is 4.2 kN, and the expected wheel-to-ground normal force of the right steering wheel is 3.1 kN. The combined bearing moment generated by the two steering wheels relative to the axle hinge center is 0.55 kNm, and the effective lateral distance from the right mechanical limit position to the axle hinge center is 0.18 m.

[0298] The effective supporting force borne by the right mechanical limit position is 3.06 kN, and the effective supporting force borne by the axle hinge center is 4.24 kN.

[0299] The effective support area is reconstructed by taking the contact points of the two drive wheels, the center of the axle hinge, and the right mechanical limit position as the effective support points of the vehicle body.

[0300] In three consecutive future effective support states, the support unloading margins of the candidate unloading support points on the right front side are 0.80 kN, 0.48 kN and 0.28 kN respectively, corresponding to directional effective tilt angles of 0.7 degrees, 1.1 degrees and 1.6 degrees respectively. The first warning threshold is 0.35 kN. Therefore, trend warning information is output in the third future effective support state.

[0301] In the next future effective support state, the support unloading margin is reduced to 0.10 kN, which is lower than the second warning threshold of 0.15 kN, and an emergency warning message is output.

[0302] Example 18: Verification process of separation of reference wheel-ground normal force

[0303] This embodiment is used to illustrate the separation of vehicle operating state changes and local elevation difference contact response.

[0304] When the forklift accelerates while turning on a flat road, the measured wheel-to-ground normal force of the left front wheel increases by 0.80 kN. Based on the current longitudinal acceleration, lateral acceleration, steering angle, yaw rate, mast tilt angle, fork lifting height, and cargo load, the reference wheel-to-ground normal force of the left front wheel should increase by 0.72 kN according to the segmented reference response relationship.

[0305] The increment of the contact force on the left front wheel is:

[0306] 0.80 − 0.72 = 0.08 kN

[0307] The contact force increment is less than the contact force determination threshold of 0.12 kN, therefore, it is not determined that actual contact has occurred due to local height difference.

[0308] If only the measured increase in wheel-to-ground normal force is compared with a fixed threshold, the load transfer caused by steering acceleration may be misjudged as local height difference contact. This implementation method distinguishes between the vehicle's own load transfer and the effect of local height difference contact by deducting the change in the reference wheel-to-ground normal force under the corresponding operating conditions.

[0309] Example 19: Processing of unverified local elevation differences

[0310] When the actual contact time is outside the contact time window, and the correspondence between the actual contact and the original local elevation difference characteristics cannot be confirmed based on the wheel position, the direction of the contact force, and the vehicle motion response, the original associated contact event group will not be used to correct the actual contact.

[0311] When the upper limit of the terrain height measurement error of the current three-dimensional sensor is positive 5 mm and the lower limit of the measurement error is negative 4 mm, the height parameters of the current local height difference feature are increased by 5 mm and decreased by 4 mm respectively to form two sets of expected contact displacements.

[0312] The future wheel contact state is recursively deduced based on the two sets of expected contact displacements, forming a first future effective support state sequence and a second future effective support state sequence.

[0313] If the minimum support unloading margin of the first future effective support state sequence is 0.26 kN and the minimum support unloading margin of the second future effective support state sequence is 0.18 kN, then 0.18 kN is used as the support unloading margin corresponding to the unverified local elevation difference feature.

[0314] Example 20: Abnormal Data and Degradation Handling

[0315] When the forward-looking 3D sensor is obstructed by cargo, resulting in a lower effective sampling point density within the predicted wheel track than required for terrain recognition, it does not form a definite local elevation difference parameter. Instead, it forms the upper and lower boundaries of the terrain height based on the effective measurement points and the sensor's measurement error range.

[0316] When the wheel-to-ground normal force detection device becomes saturated, data is interrupted, or multiple consecutive sampling values ​​remain unchanged, the corresponding wheel-to-ground normal force data is marked as invalid. If the wheel-to-ground normal force cannot be calculated from other support detection quantities, the actual contact displacement inversion is not performed using the wheel data. Instead, an upper and lower boundary state sequence is formed based on the forward-looking terrain data.

[0317] When the vehicle motion sensor detects abnormal data, the actual contact moment cannot be directly determined simply because the increase in contact force reaches the threshold. Instead, the predicted contact event is marked as an event to be verified.

[0318] When the current operating condition exceeds the calibration range of the segmented reference response relationship or the wheel contact displacement response matrix, the boundary calibration parameters closest to the current operating condition are used, and the prediction results are marked as low-confidence results. Low-confidence results are treated as unverified local elevation difference features.

[0319] Example 21: Overall Execution Process

[0320] Once the forklift enters the monitoring state, the data processing device first synchronizes the time of each data point and then forms a group of related contact events based on the forward-looking terrain data and the predicted wheel tracks of each wheel.

[0321] Subsequently, based on vehicle operation data and gantry load data, the reference wheel-to-ground normal force variation of each wheel is determined. The reference wheel-to-ground normal force variation is subtracted from the measured wheel-to-ground normal force variation to obtain the contact force increment corresponding to the local elevation difference contact.

[0322] When the wheels that make contact first meet the contact force condition and the vehicle motion response condition, the actual contact time is determined. The measured contact response vector is formed based on the contact force increment. Based on the number of contact events predicted within the contact time window, the actual contact displacement is determined using a single unit contact displacement response vector or a contact displacement response matrix.

[0323] The deviation between the actual contact time and the expected contact time is only used to correct the position parameters of the local elevation difference characteristics, and the deviation between the actual contact displacement and the expected contact displacement is only used to correct the height parameters of the local elevation difference characteristics.

[0324] The correction results are only passed to other wheel predicted contact events in the same associated contact event group that have not yet passed through the local elevation difference feature. They do not backtrack to modify predicted contact events that have already occurred, nor do they modify other local elevation difference features.

[0325] The future wheel contact state is inferred recursively based on the updated predicted contact events. The future wheel contact state is then converted into the future effective support state based on the axle structure and mechanical limit status.

[0326] Finally, based on the future effective support state, the support reference plane, the predicted attitude of the vehicle body, the support unloading margin, the candidate unloading support point, the candidate overturning boundary, and the directional effective tilt angle are determined, and trend warning information or emergency warning information is output based on the continuous changes of the support unloading margin and the directional effective tilt angle.

[0327] It should be noted that, in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, method, article, or apparatus.

[0328] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A method for real-time monitoring and anti-tipping early warning of forklift tilt angle based on multi-sensor fusion, characterized in that, Anti-tipping warning systems applied to forklifts traversing roads with localized elevation differences include: Acquire terrain data in front of the forklift, vehicle operation data, vehicle body motion data, mast load data, current relative contact height and current wheel-to-ground normal force of each wheel, as well as the mapping relationship between wheel contact position and effective support point of vehicle body and the adjacency relationship between effective support points of vehicle body, all with a unified time reference. Based on the vehicle operation data, the predicted wheel tracks of each wheel are determined, and the contact change positions of the same local elevation difference feature on different predicted wheel tracks are determined as predicted contact events, forming a group of associated contact events arranged according to the expected contact time. Based on the vehicle operation data and gantry load data, determine the reference wheel-to-ground normal force change caused by changes in vehicle acceleration / deceleration, steering, and gantry load state when the ground height relationship remains unchanged. The contact force increment is determined by the difference between the measured wheel-to-ground normal force change and the reference wheel-to-ground normal force change. Based on the contact force increment, vehicle motion data, and pre-calibrated contact displacement response relationship, the actual contact time and actual contact displacement are determined; the position parameters of the local elevation difference feature along the predicted wheel track direction are corrected based on the deviation between the actual contact time and the predicted contact time; the height parameters of the local elevation difference feature are corrected based on the deviation between the actual contact displacement and the predicted contact displacement; and the predicted contact events of other wheels in the same associated contact event group that have not yet passed the local elevation difference feature are updated only. Using the current relative contact height and current wheel-ground normal force of each wheel as the initial state, the future wheel contact state is recursively deduced based on the updated predicted contact events, and then converted into the future effective support state according to the mapping relationship. Based on the future effective support state, determine the support reference plane, the predicted attitude of the vehicle body, the support unloading margin, the candidate unloading support points and the corresponding candidate overturning boundaries. When the unloading margin of any candidate unloading support point decreases sequentially in at least two consecutive future effective support states and is less than or equal to the warning threshold, and the attitude difference between the predicted attitude of the vehicle body and the support reference plane increases sequentially in the direction towards the outside of the corresponding candidate rollover boundary, rollover warning information is output.

2. The method for real-time monitoring and anti-tipping early warning of forklift tilt angle based on multi-sensor fusion according to claim 1, characterized in that: The terrain data in front of the forklift includes the spatial coordinates of terrain sampling points collected by a forward-looking 3D sensor, and the terrain normal obtained by fitting adjacent terrain sampling points. The vehicle operation data includes driving speed, driving direction, longitudinal acceleration, lateral acceleration, steering angle, and yaw rate; The vehicle motion data includes the vehicle's vertical acceleration, roll rate, pitch rate, roll acceleration, pitch acceleration, and vehicle attitude. The mast load data includes mast tilt angle, fork lifting height, total cargo load, and left and right fork loads. The wheel-to-ground normal force is collected by wheel-mounted sensors, or calculated based on at least one of axle strain, steering axle support force, and support cylinder pressure. The current relative contact height is determined based on the current vehicle body posture, wheel installation coordinates, effective rolling radius of the wheel, and deformation of the wheel ground support structure under the current wheel-ground normal force.

3. The method for real-time monitoring and anti-tipping early warning of forklift tilt angle based on multi-sensor fusion according to claim 2, characterized in that: Forming the associated contact event group includes: Based on the installation coordinates of each wheel in the vehicle coordinate system and the expected driving speed, expected driving direction and expected turning state in the warning time domain, the movement trajectory of each wheel contact position is determined, and the movement trajectory is laterally extended according to the grounding width of the corresponding wheel to form a predicted wheel track. The reference terrain surface is fitted based on the terrain sampling points located on the front side of the area to be identified along the driving direction. The local elevation difference features are determined based on the height deviation of the remaining terrain sampling points from the reference terrain surface, the angle between the terrain normal and the reference terrain surface normal, and the adjacent distance between the terrain sampling points. The local elevation difference features along the entry boundary, exit boundary, and internal elevation change location of each predicted wheel track are respectively determined as predicted contact events; The predicted contact time is determined based on the position of the predicted contact event in the corresponding predicted wheel track and the predicted speed of the corresponding wheel; for the predicted contact event corresponding to the entry boundary, the directional height difference between the entry boundary and the reference terrain surface is determined as the predicted contact displacement; for the predicted contact events corresponding to the departure boundary and the internal height change position, the directional height difference between the predicted contact event and its previous predicted contact event is determined as the predicted contact displacement. The predicted contact events corresponding to the same local elevation difference feature are arranged according to the expected contact time to form the associated contact event group.

4. The method for real-time monitoring and anti-tipping early warning of forklift tilt angle based on multi-sensor fusion according to claim 3, characterized in that: Determining the variation of the reference wheel-ground normal force includes: On a known flat road surface, the longitudinal acceleration, lateral acceleration, steering angle, yaw rate, mast tilt angle, fork lifting height, total load and load difference between the left and right forks of the forklift are changed respectively, and the wheel-to-ground normal force of each wheel is collected. The calibration conditions are divided according to the total load range, fork lifting height range, and steering angle range. A segmented reference response relationship is established between the vehicle operation data and mast load data and the change in the wheel-ground normal force of each wheel under each calibration condition. Based on the current total cargo load, fork lifting height, and steering angle, select the corresponding segmented reference response relationship, and input the current vehicle operation data and mast load data into the selected segmented reference response relationship to obtain the reference wheel-to-ground normal force change of each wheel.

5. The method for real-time monitoring and anti-tipping early warning of forklift tilt angle based on multi-sensor fusion according to claim 4, characterized in that: Determining the actual contact time and actual contact displacement includes: When the absolute value of the increase in contact force of the wheel corresponding to the predicted contact event reaches the contact force determination threshold, and at least one of the vehicle body vertical acceleration, roll rate and pitch rate reaches the corresponding motion response determination threshold, and its positive and negative directions are consistent with the expected response direction corresponding to the predicted contact displacement, it is determined that actual contact has occurred, and the moment when the above conditions begin to be met simultaneously is determined as the actual contact moment. Based on the incremental contact force of each wheel and the lateral and longitudinal lever arms of each wheel relative to the center of mass of the load system formed by the forklift and the cargo, the changes in vertical support resultant force, tilt support moment and pitch support moment are determined respectively, and normalized according to the standard deviation of each change in the loading calibration to form the measured contact response vector. When the contact time window corresponding to the actual contact moment includes only one predicted contact event, select the unit contact displacement response vector according to the wheel and gantry load state corresponding to the predicted contact event, and find the contact displacement to be determined that minimizes the squared residual of the product of the unit contact displacement response vector and the contact displacement to be determined relative to the measured contact response vector. When the contact time window includes two or more predicted contact events, the contact displacement response matrix is ​​formed by the unit contact displacement response vectors of the wheels corresponding to each predicted contact event. The contact displacement vector to be determined is then found to minimize the squared residual of the product of the contact displacement response matrix and the contact displacement vector to be determined relative to the measured contact response vector. The contact displacement to be determined or the corresponding component in the contact displacement vector to be determined is determined as the actual contact displacement, and the positive and negative directions of each actual contact displacement are defined to be consistent with the positive and negative directions of the corresponding expected contact displacement.

6. The method for real-time monitoring and anti-tipping early warning of forklift tilt angle based on multi-sensor fusion according to claim 5, characterized in that: Updating the predicted contact event includes: The wheel speed of the corresponding wheel during the time interval between the actual contact time and the corresponding expected contact time is integrated to obtain the contact position deviation, and the position parameters of the local elevation feature along the corresponding predicted wheel track direction are corrected by the contact position deviation. Determine the displacement deviation between the actual contact displacement and the corresponding expected contact displacement, and use the displacement deviation to correct the height parameter of the local elevation difference feature at the corresponding predicted wheel track. Based on the contact change position between the corrected local elevation difference feature and the predicted wheel tracks of other wheels that have not yet passed through the local elevation difference feature, the expected contact time and expected contact displacement of the predicted contact event corresponding to the other wheels are re-determined; When actual contact occurs again in the same associated contact event group, based on the contact position deviation and displacement deviation obtained from the last actual contact, only the predicted contact events of other wheels that have not yet passed through the local elevation difference feature are updated.

7. The method for real-time monitoring and anti-tipping early warning of forklift tilt angle based on multi-sensor fusion according to claim 6, characterized in that: The recursive calculation of the future wheel contact state includes: The mass and center of gravity of the load system consisting of the forklift and the cargo are determined based on the mast tilt angle, fork lifting height, total cargo load, and loads on the left and right forks. Based on the mass, center of gravity position, current travel speed range, and fork lifting height range of the load system, the corresponding response matrix is ​​selected from the segmented wheel contact displacement-wheel-ground normal force response matrix formed by the wheel ground support structure loading test and the local elevation difference road travel test. Predicted contact events with overlapping expected contact time windows are grouped into the same state update node. A contact displacement increment vector is established based on the wheel corresponding to each predicted contact event in the state update node. Each expected contact displacement is set as the corresponding vector component, and the corresponding component of the wheel not included in the state update node is set to zero. By inputting the contact displacement increment vector into the selected response matrix, the wheel-to-ground normal force increment of each wheel is obtained; Based on the time interval between the previous state update node and the current state update node, and according to the dynamic response attenuation relationship formed by the local elevation difference road driving test, the wheel-ground normal force deviation in the future wheel contact state of the previous state is attenuated. Based on the vehicle operation data and gantry load data corresponding to the current state update node, determine the node reference wheel-to-ground normal force of each wheel, add the attenuated wheel-to-ground normal force deviation and the current wheel-to-ground normal force increment to the node reference wheel-to-ground normal force, and accumulate the contact displacement increment vector to the relative contact height of the previous future wheel contact state to obtain the next future wheel contact state.

8. The method for real-time monitoring and anti-tipping early warning of forklift tilt angle based on multi-sensor fusion according to claim 7, characterized in that: Converting the future wheel contact state into a future effective support state includes: For wheels that transmit support forces to the frame through rigid axles or independent support structures, the wheel contact position is determined as the effective support point of the vehicle body, and the expected relative contact height of the wheel and the expected wheel-to-ground normal force are determined as the effective support height and effective support force of the corresponding effective support point of the vehicle body, respectively. For two steering wheels that transmit support force to the frame through a swing steering axle, when the axle swing angle determined based on the expected relative contact height of the two steering wheels does not reach the mechanical limit angle of the axle, the horizontal projection position of the hinge center of the swing steering axle is determined as the planar position of the equivalent support point of the vehicle body. The effective support height of the equivalent support point of the vehicle body is determined based on the expected relative contact height of the two steering wheels, the wheel track, the coordinates of the axle hinge center, and the geometric constraint relationship of the swing steering axle. The sum of the expected wheel-to-ground normal forces of the two steering wheels is determined as the effective support force of the equivalent support point of the vehicle body. When the axle swing angle reaches the axle mechanical limit angle, the hinge center of the swing steering axle and the limit position where the limit contact occurs are respectively determined as the effective support points of the vehicle body. The sum of the effective support forces of the two steering wheels is determined according to the sum of the expected wheel-to-ground normal forces of the two steering wheels. The effective support forces of the hinge center and the limit position are determined according to the support torque generated by the two steering wheels relative to the hinge center and the distance between the hinge center and the limit position. Based on the spatial position of the hinge center and the limiting position, the effective support height and the effective support force, the adjacency relationship between the effective support points of the vehicle body is updated to form the corresponding future effective support state.

9. The method for real-time monitoring and anti-tipping early warning of forklift tilt angle based on multi-sensor fusion according to claim 8, characterized in that: The critical effective support force is the ground clearance threshold determined by the calibration test of the wheel changing from a ground-to-ground state to a ground-to-ground state, and is not less than the equivalent support force corresponding to the upper limit of the wheel-to-ground normal force detection noise. When the effective support force is not less than the critical effective support force and there are no fewer than three non-collinear effective support points on the vehicle body, the support reference plane is determined according to the position and effective support height of each effective support point on the vehicle body; when there are fewer than three effective support points on the vehicle body, the corresponding future effective support state is determined as the critical instability state. Based on the current vehicle body attitude, vehicle body angular velocity, expected longitudinal acceleration, expected lateral acceleration, expected steering state and gantry load state, the predicted vehicle body attitude is determined according to the vehicle body attitude response relationship formed by the smooth road test, and the effective tilt angle is determined according to the lateral attitude difference and longitudinal attitude difference of the predicted vehicle body attitude relative to the support reference plane. The difference between the effective support force and the critical effective support force of each effective support point of the vehicle body is determined as the support unloading margin. One or more effective support points of the vehicle body whose difference from the minimum support unloading margin is not greater than the parallel judgment threshold are determined as candidate unloading support points. Based on the adjacency relationship, the boundary formed by the two effective support points of the vehicle body that are adjacent to the candidate unloading support point and maintain effective support is determined as the candidate overturning boundary, and the projection of the effective tilt angle in the direction perpendicular to the corresponding candidate overturning boundary and pointing from the inside of the effective support area to the outside is determined as the directional effective tilt angle. When the support unloading margin of any candidate unloading support point decreases sequentially in at least two consecutive future effective support states and is less than or equal to the first warning threshold, and the corresponding directional effective tilt angle increases sequentially in at least two consecutive future effective support states, trend warning information is output. When the unloading margin of any support is less than or equal to the second warning threshold, or when the corresponding future effective support state is in a critical instability state, an emergency warning message is output, wherein the second warning threshold is less than the first warning threshold.

10. The method for real-time monitoring and anti-tipping early warning of forklift tilt angle based on multi-sensor fusion according to claim 9, characterized in that: When the actual contact time is outside the contact time window of the corresponding predicted contact event, or when the absolute value of the difference between the actual contact displacement and the corresponding expected contact displacement is greater than the correction deviation threshold, the corresponding local elevation difference feature is determined as an unverified local elevation difference feature. Based on the height parameters of the unverified local elevation difference features, the upper limit of terrain height measurement error, and the lower limit of terrain height measurement error, the first and second expected contact displacements are determined respectively. The corresponding predicted contact events are updated with the first and second predicted contact displacements, respectively, to form the first future effective support state sequence and the second future effective support state sequence. The minimum support unloading margin of all effective support points of the vehicle body in the first future effective support state sequence and the second future effective support state sequence at all state update nodes is determined respectively, and the smaller value of the two minimum support unloading margins is determined as the support unloading margin corresponding to the unverified local elevation difference feature.

Images