Centralized buffer and station permeate material distribution control method for complex whole machine mixed line assembly

Abstract

The application discloses a centralized cache and work station permeate material distribution control method for complex whole machine mixed line assembly, and relates to the technical field of industrial logistics scheduling and robot control. The method comprises the following steps: calculating material centroid eccentricity according to stress distribution and supporting point coordinates of each supporting unit, driving a variable distance mechanism to translate the supporting unit to compensate the centroid eccentricity; determining the safe allowable speed of each road section according to the residual eccentricity, generating the arrival time interval of each topological node and resetting the interval width at the loading and unloading site to solve the distribution timing; when the attitude leveling control amount is saturated or the downstream predicted delay is out of limit during distribution, the event trigger flag is latched, the timing is reconstructed in the local space-time subgraph and the external is kept unchanged; and the mobile robot is controlled to distribute the material according to the distribution timing. The application is used to solve the problem of channel intersection deadlock caused by material lateral overturning and leveling speed limit.

Description

Technical Field

[0001] This invention relates to the field of industrial logistics scheduling and robot control technology, and more specifically, to a centralized buffer and workstation-penetrating material distribution control method for complex mixed-line assembly of complete machines. Background Technology

[0002] Complex mixed-line assembly manufacturing involves multi-station, high-frequency parallel operations and fast-paced, flexible switching, with a large span of transported materials, including large, irregularly shaped, and heavy-duty components. To ensure the assembly cycle time of mixed lines and avoid collision damage to precision components during transport, the logistics system must take into account both the spatiotemporal conflict-free collaborative scheduling of multiple mobile robot clusters and the stable attitude control during the transportation of large and heavy-duty components.

[0003] To address the aforementioned needs, existing technologies typically rely on a gridded multi-objective layout model of the workshop. At the scheduling layer, swarm intelligence algorithms such as the gold mining optimization algorithm are used to solve the discrete conflict-free paths and time window timing of multiple robots. At the execution layer, the vehicle pitch-changing mechanism is combined with an incremental PID control algorithm to achieve attitude correction and smooth control during transportation.

[0004] In the aforementioned technical system, the discrete path planning layer and the continuous attitude control layer adopt a unidirectional serial control architecture. Due to the uneven mass distribution of irregularly shaped components, the center of mass of the material deviates from the support center plane of the carrier, and the superimposed ground undulation disturbance easily generates lateral overturning moments during transportation. The deceleration generated by the bottom attitude leveling mechanism when performing suppression and protection actions causes the robot's actual speed to deviate from the preset speed profile, resulting in the accumulation of nonlinear delays and the destruction of the discrete time window mutual exclusion benchmark, causing channel intersection conflicts and deadlocks. If the continuous dynamic differential equations and discrete path planning are solved together, the problem of a surge in high-dimensional non-convex optimization computation is faced, making it difficult to meet the millisecond-level real-time response requirements of industrial sites. If only global conservative speed limiting or statically increasing the time safety margin is adopted, it will directly cause the workstation to stop waiting for materials, reducing the overall assembly cycle time and the level of material supply. Summary of the Invention

[0005] To overcome the aforementioned deficiencies of the prior art, this invention provides a centralized buffer and station-penetrating material delivery control method for complex mixed-line assembly. This method compensates for centroid eccentricity by using translational support units, generates arrival time intervals based on residual eccentricity and resets the solution timing at loading and unloading stations, and locally reconstructs the timing when attitude leveling saturation or predicted delay exceeds limits. This solves the technical problems of material lateral overturning and channel deadlock induced by leveling speed limits.

[0006] To achieve the above objectives, the present invention provides the following technical solution: A centralized buffer and workstation-penetrating material delivery control method for complex mixed-line assembly includes the following steps: calculating the eccentricity of the material's center of mass relative to the longitudinal symmetry plane of the vehicle body based on the force distribution of each support unit and the coordinates of the support points; driving the variable pitch mechanism to translate the support units to compensate for the eccentricity of the center of mass; determining the safe allowable speed of each segment of the planned path based on the compensated residual eccentricity; generating the arrival time interval of each topological node; and resetting the width of the time interval at the loading and unloading station to solve for the delivery sequence that satisfies the spatiotemporal mutual exclusion constraint; during the delivery process, when the attitude leveling control quantity is saturated or the downstream prediction delay exceeds the limit, latching the event trigger flag and reconstructing the delivery sequence of the associated robots within the local spatiotemporal subgraph, while keeping the external sequence of the subgraph unchanged; and controlling each mobile robot to deliver materials according to the delivery sequence.

[0007] Preferably, the step of calculating the eccentricity of the material's center of mass relative to the longitudinal symmetry plane of the vehicle body and driving the pitch mechanism to translate the support unit to compensate for the eccentricity includes: using the initial support spacing determined by the material's geometric envelope as a reference, calculating the planar projection coordinates of the material's center of mass on the bearing surface using the static moment balance equation based on the vertical reaction force collected by each support unit and its support point coordinates; driving the pitch mechanism to translate the support unit laterally within the maximum stroke, so that the projection of the material's center of mass on the bearing surface approximates the longitudinal symmetry plane of the vehicle body.

[0008] In summary, the present invention has the following beneficial effects: This invention calculates the material's centroid eccentricity based on the force distribution and support point coordinates of each support unit and drives a pitch-changing mechanism to translate the support units to compensate for the centroid eccentricity. It then determines the safe allowable speed for each road segment based on the compensated residual eccentricity to generate the arrival time interval for each topological node. The invention also resets the time interval width at loading and unloading stations to solve for the delivery sequence. Furthermore, during delivery, it reconstructs the delivery sequence of associated robots within a local spatiotemporal subgraph based on the saturation of the attitude leveling control quantity or the trigger flag of the downstream predicted delay exceeding the latch limit, while maintaining the external time sequence of the subgraph unchanged. This solves the technical problems of lateral overturning moments caused by the centroid deviating from the vehicle's symmetry plane of irregularly shaped heavy-load materials, and the accumulation of delays caused by the bottom-level attitude leveling speed limit, which damages the spatiotemporal mutual exclusion benchmark and induces deadlock at channel intersections. By pre-reducing the lateral overturning moment during the loading stage to reduce the online speed limit amplitude, it prevents the time interval from monotonically diverging along the path and limits the resolution of timing conflicts within the local topology to avoid global replanning calculation delays. This ensures the transport stability and spatiotemporally conflict-free collaborative delivery of multiple mobile robots in a mixed-line assembly environment. Attached Figure Description

[0009] Figure 1 A schematic diagram of a centralized buffer and workstation-penetrating material distribution control method for complex mixed-line assembly of complete machines is provided in this embodiment of the invention. Figure 2This is a safety-permissible velocity response surface plot in an embodiment of the present invention; Figure 3 This is a timing diagram of the reaction force difference measurement and centroid height filtering estimation for the turning section in this embodiment of the invention; Figure 4 This is a timing relationship diagram of the saturation of the leveling control quantity, speed reduction, prediction delay integral, and event trigger flag state in an embodiment of the present invention; Figure 5 This is a comparison diagram of the resolution of spatiotemporal trajectory and velocity profile of a local spatiotemporal subgraph in an embodiment of the present invention. Detailed Implementation

[0010] 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 of ordinary skill in the art without creative effort are within the scope of protection of the present invention.

[0011] Figure 1 This invention presents a centralized buffering and workstation-penetrating material delivery control method for complex mixed-assembly lines, comprising the following steps: S1, calculate the eccentricity of the material's center of mass relative to the longitudinal symmetry plane of the vehicle body based on the force distribution of each support unit and the coordinates of the support points, and drive the variable pitch mechanism to translate the support unit to compensate for the eccentricity of the center of mass.

[0012] In this embodiment, the runtime environment is first handled and discretized, including the following steps: S101, a two-dimensional gridded topology model is established based on the physical boundaries of the workshop, workstation coordinates, and the distribution of static obstacles, with the grid resolution set to [value missing]. For large, irregularly shaped materials, perform minimum bounding rectangle fitting on the 3D point cloud data to extract the length, width, height, and overall dimensions of the material's geometric envelope; edge computing nodes combine the material demand sequence of each workstation on the production line to complete the allocation of delivery tasks and initial path planning, determining the planned path of each mobile robot and its included road segments and topology node sequences, the first... The planned path of the mobile robot. The above-mentioned gridded topology modeling, geometric envelope fitting and initial path planning are basic technologies already existing in this field. In this embodiment, loading period eccentricity identification and geometric compensation are performed based on the output data structure. The eccentricity calculation and support unit translation compensation are completed before the material is placed and the mobile robot leaves the loading station, without occupying the production line's online operation cycle.

[0013] S102, using the initial support spacing determined by the material's geometric envelope as a reference, and based on the vertical reaction forces and support point coordinates collected from each support unit, the planar projection coordinates of the material's center of mass on the bearing surface are calculated using the static moment equilibrium equation. Specifically, the flexible carrier is configured according to the fitted material geometric envelope dimensions. The initial support spacing of each variable-pitch support unit is used to establish a vehicle coordinate system: the origin of the coordinate system is set at the geometric center of the vehicle body. The axle points directly forward along the longitudinal direction of the vehicle body. The axle points laterally to the left along the vehicle body, and the planar coordinates of each support point in the vehicle body coordinate system are configured as follows: Collect the vertical reaction force output by the force measuring elements configured on each support unit. The planar projection coordinates of the material's center of mass on the bearing surface are calculated using the planar static moment equilibrium equation. The total mass of the material is determined by the algebraic sum of the vertical reaction forces. : (1) In the formula, For the first Vertical reaction force collected by each support unit ; For the first Planar coordinates of each support unit in the vehicle coordinate system ; It is the acceleration due to gravity; The eccentric component of the material's center of mass along the longitudinal direction ; The eccentricity of the material's center of mass relative to the vehicle's center plane in the lateral direction. ; Total mass of materials The transverse support span of the variable-pitch support unit before spanning is... The height of the support surface from the ground is .

[0014] Substitute the two sets of measured load data into the analysis. Under load condition A, the vertical reaction forces at four points are measured as follows: ,have to , Substituting into the static moment equilibrium equation, we get

[0015] The vertical reaction forces measured at four points under working condition B are: The total mass of the materials is the same Substituting into the lateral moment balance equation, we get

[0016] The longitudinal eccentricity obtained by solution Used to verify the vertical load distribution and adhesion margin of the front and rear steering wheel axles of a mobile robot, ensuring that slippage does not occur during acceleration and deceleration.

[0017] S103, the drive torque converter translates the support unit laterally within its maximum stroke, causing the projection of the material's center of mass onto the bearing surface to approximate the longitudinal center symmetry plane of the vehicle body. Specifically, the vehicle body center plane is the longitudinal center symmetry plane of the vehicle body. The onboard motion controller determines the lateral eccentricity based on this. Output variable pitch execution command, translational compensation displacement. With the signed residual lateral eccentricity Determined by the following formula: (2) In the formula, The maximum permissible translational stroke in one direction for the pitch mechanism ; Signed translation displacement ; The remaining lateral eccentricity after translation compensation This value, indicated by a positive or negative sign, indicates the direction of residual eccentricity and is used in subsequent dynamic calculations to determine the rollover hazard condition. Condition A ,have to , The lateral eccentricity moment elimination rate is Under operating condition B Because the mechanical travel limit has been reached, it is still taken Residual eccentricity The eccentric moment elimination rate is The retained positive sign indicates that the centroid is still biased to the left.

[0018] S104, further, the maximum allowable span is determined based on the aisle width, vehicle width, and safety clearance. Under the premise of meeting the aisle clearance requirements, the left and right support units are symmetrically widened to increase the support span, and the overall width of the extended machine is included in the clearance check. Specifically, the variable-pitch support unit is mounted on the chassis and is not grounded. The symmetrical widening action only changes the material support span on the carrier's bearing surface, and the nominal half-track of the mobile robot chassis... Since the position of the rollover line on the ground remains constant, the lateral anti-rollover boundary of the entire aircraft does not change with the span, and the span is not introduced into the roll constraint equation. By channel width Vehicle body width Operating safety clearance The maximum lateral profile of the entire machine after extension and translation must meet the clearance constraints: (3) In the formula, Single-sided span ; The lateral half-width of a single pitch support unit ; The geometric envelope width of the material ; The absolute value of the translational compensation displacement in step S103 ; Support span before extension After the span is extended, the support span is... When both the translated material envelope and the expanded support outline do not exceed the width of the vehicle body, the above formula degenerates into: (4) Substituting the values ​​into the theoretical upper limit of one-sided span is... In this embodiment, After obtaining the span of the support, Verify the support outline after the span. The outer contour of the material envelope after translation None of them exceed the width of the vehicle body, therefore the maximum lateral outline of the entire machine is still determined by the width of the vehicle body, that is... The clearance check of the passageway has been passed.

[0019] S105, when the outer contour of the translated material exceeds the channel clearance, the translation amount is cut off to meet the channel clearance, and the vertical reaction force of each support unit after translation is collected to recalculate the residual eccentricity. Specifically, after the translation and extension actions of the support units are locked, a secondary check of the channel clearance is performed on the overall envelope of the translated material. The available clearance on one side of the vehicle body is...

[0020] The unilateral outward convexity of the material envelope relative to the vehicle body boundary after translation is: (5) Must meet In this embodiment, the material envelope width Translation amount Substituting the values, we get the amount of outward convexity. After translation, the material's outline does not exceed the channel's clearance, maintaining the original translation amount. However, if the material's envelope width is large or the channel width is narrowed, causing... At that time, the translation amount is forcibly truncated to (6) This triggers a downgraded speed limit indicator, and the uncompensated residual eccentricity is handled by the subsequent safe speed envelope. Then, the new fulcrum geometric coordinates after translation are used. (The translational displacement is superimposed from the original pivot coordinates) (The obtained data) is used as input to re-acquire the vertical reaction force of each support unit. Substitute into the static moment equilibrium equation of S102 to recalculate the transverse coordinate of the centroid. And update the residual eccentricity. This constitutes a closed-loop verification process of "identification-translation-re-verification" during the loading period; the residual eccentricity obtained from the re-measurement is consistent with the theoretical residual in step S103 (condition A is...). Geometric compensation is in place.

[0021] It should be noted that the force-measuring element of the support unit is not limited to a resistance strain gauge load cell. In another optional embodiment, it can also be replaced with a piezoelectric force sensor, a hydraulic cylinder pressure sensor, or a load cell mounted on the chassis suspension guide mechanism; the pitch mechanism, in addition to a servo-driven ball screw, can also be replaced with a servo synchronous toothed belt drive mechanism, a precision gear rack pair, or a direct-drive linear motor; the number of support units... It can also be configured as 3 points or 6 points, but when configured as... In statically indeterminate multi-point support configurations, the distribution of reaction forces at each support point is directly affected by the distribution of support stiffness and the flatness tolerance of the material's bottom surface, resulting in "virtual leg" distortion. In this case, it is necessary to connect elastic floating elements (disc spring groups or parallel damping orifice hydraulic floating support pads) in series at the mechanical interfaces of each support unit. This allows each support point to adaptively undergo minute displacements under static load to homogenize stiffness. Alternatively, the onboard motion controller can perform weighted least squares linear regression fitting on the reaction force equations of each support point to calculate the equivalent centroid projection coordinates, ensuring that the reaction force distribution truly reflects the material's centroid. All of the above substitutions do not depart from the present invention's concept of using geometric reconstruction during loading to achieve stability.

[0022] This step directly analyzes the centroid eccentricity during the static period after material loading and before the mobile robot leaves the workstation by utilizing multi-point static reaction forces and algebraic torque balance. This drives a high-rigidity variable-pitch mechanism to perform signed centering displacement compensation and symmetrical widening. A closed-loop geometric reconstruction is formed using the channel clearance as a hard constraint. This solves the problem that lateral overturning moments caused by centroid offset in large, irregularly shaped materials would directly reduce material flow capacity and worsen assembly cycle time if only passively suppressed by reducing the operating speed online. It achieves the elimination of the main eccentric moment in one go during the loading stage (elimination under baseline conditions). ; in the case of eccentricity obeying Under the condition of uniform distribution, the theoretical average elimination rate is calculated according to the expected integral formula to reach... This allows the residual eccentricity to converge to a controlled feasible region, thereby providing a physical safety margin for subsequent operating speeds.

[0023] S2 determines the safe allowable speed of each segment of the planned path based on the compensated residual eccentricity, generates the arrival time interval of each topology node, and resets the time interval width at the loading and unloading station to solve the delivery sequence that satisfies the spatiotemporal mutual exclusion constraint.

[0024] In this embodiment, the material eccentricity geometric parameters and the vehicle dynamics boundary are discretized into an algebraic velocity envelope. Interval arithmetic is used to map the travel time domain of each node piecewise, and the accumulated error is physically truncated based on the resynchronization anchor point, as follows: S201, based on the mobile robot's half-wheelbase, maximum linear velocity, sideslip limit, and the lateral anti-tipping arm determined by residual eccentricity and steering curvature, constructs a lookup table for safe linear velocities regarding path curvature, residual eccentricity, and center of mass height. Specifically, it defines the signed effective lateral anti-tipping arm. Speeds permissible on each road section The algebraic governing equations are: (7) (8) In the formula, The nominal half track width of the vehicle body ; Lower bound of lever arm numerical clamping This is used to prevent the effective lever arm from being non-positive under atypical abnormal working conditions, which would result in no real solution within the square root. For the planned path geometric curvature The symbol is conventionally defined as a left turn. ,Turn right ; The residual lateral eccentricity retaining the algebraic symbol after translation compensation in step S1. ; The height of the material's center of mass above the ground ; The coefficient of lateral adhesion between the tire and the ground (dimensionless). To account for the safety reduction factor (dimensionless) for workshop floor unevenness and suspension dynamic vibration. To drive the mechanical limit speed of the system The three terms in the formula correspond to the actuator limit, tire sideslip limit, and lateral rollover limit, respectively. Because the pitch support unit is mounted on the vehicle frame and does not directly contact the ground, the increased span of the support increases the overall span. Without changing the position of the ground wire on the outer side of the wheel, the lateral rollover boundary of the entire machine is entirely determined by the half-track of the chassis. It was decided that the span would not be introduced into the rollover constraint equations. .

[0025] S202, on each segment of the planned path of each mobile robot, the corresponding safe allowable speed is obtained by retrieving the safe linear velocity lookup table using trilinear interpolation. Specifically, the lookup table is stored in a discrete grid, with a defined domain coverage. (Step length) ), (Step length) ), (Step length) For straight road sections, a straight-line curvature determination threshold is set. When the path geometric curvature satisfies When the condition is determined to be straight-going, a safe permissible speed is directly assigned. Trilinear interpolation is not performed. The mobile robot moves to each planned curve section. At that time, the curvature of the road segment was extracted. Residual eccentricity Compared with the current centroid height estimate The lookup table is searched, and when the parameter goes out of bounds, it is adaptively projected to the nearest boundary to take a conservative limit value.

[0026] Following the definition of the working condition in step S1, taking a typical right-angle turn grid corner in a workshop as an example, the local curvature of the corner is known. (i.e., turning radius) (Right turn lane), the height of the centroid is identified. Choose the direction with the most unfavorable dynamics, i.e., turn right. The superimposed centroid is biased to the left. Operating Condition A ( )Down The sideslip limit is The rollover limit is Take the smallest value Compare with nominal operating speed ,because The mobile robot can pass through the corner at its nominal speed, with a deceleration rate of [missing value]. Operating Condition B ( )Down Maintaining the sideslip limit The rollover limit is reduced to Take the smallest value ;because The driver was forced to reduce speed at this corner section, with the reduction amount being [missing information]. The above numerical comparisons demonstrate that the active centering during loading to eliminate main eccentricity makes a direct technical contribution to maintaining online traffic speed.

[0027] The numerical correspondence between safe permissible speed, path curvature, and residual eccentricity is as follows: Figure 2 As shown. Figure 2 (a) in the figure represents the height of the centroid. The safe allowable velocity response surface plot at that time. Where: Axis represents the geometric curvature of the path (Scope) ), The axis represents the residual lateral eccentricity. (Scope) ), The axis indicates the safe permissible speed. (Scope) The surface is generated by taking the minimum value of the algebraic control equation for the safe allowable speed in step S201: In the low curvature region, the speed is taken as the maximum linear velocity of the driving system. In regions where the steering curvature increases or the anti-rollover lever arm shortens, the curved surface decreases due to the constraints of sideslip and rollover limits; the plane... The nominal operating speed is the reference; such as Figure 2 As shown in (b), the calculation point for working condition A is ( The corresponding safe and permissible speed is Located above the reference plane; Calculation point for working condition B ( The corresponding safe and permissible speed is Located below the reference plane; the contour projection at the bottom of the surface represents the gradient distribution of the allowable steering speed under different eccentricities.

[0028] S203, determine the total mass of the material based on the sum of the vertical reaction forces of each support unit; when the mobile robot passes through a turning section with known curvature and the lateral acceleration reaches a preset threshold, collect the difference in vertical reaction forces between the left and right support units; construct an observation equation using the total mass of the material, the height of the support surface from the ground, the support span, and the aforementioned difference in vertical reaction forces, using the estimated material density as the initial value, and update the estimated material centroid height online using Kalman filtering. Specifically, to address the uncertainty in centroid height caused by the uneven distribution of internal material density in irregularly shaped workpieces from different assembly batches, construct a one-dimensional random walk state equation for centroid height and a lateral load transfer observation equation for the curve: (9) (10) In the formula, For the first Estimated height of centroid above ground ; The process excitation noise is denoted as follows: ; For the observation of the centroid height based on dynamic inverse calculation ; Height of the support surface from the ground ; The difference between the vertical reaction forces of the outer and inner support units during steady-state cornering. ; and The measured reaction force values ​​of the outer and inner support units are respectively. ; The span of the support after the span is extended. ; Total mass of materials; The measured steady-state operating speed on the curve section ; The known geometric curvature of the curve segment. .

[0029] The specific filter parameters are configured as follows: the standard deviation of the force sensor's measurement noise is statically calibrated to... (i.e., variance) Based on the error propagation of the first-order Taylor partial derivative of the observation equation, the standard deviation of the equivalent height observation noise is: Substitute the numerical values ​​into the identified road segments. The equivalent height observation noise variance is calculated as follows: The initial covariance of the state estimate is set to... The initial value of the prior height based on the nominal density of the workpiece is set as follows: .

[0030] Mobile robots Entering a known curvature For a smooth curve, a lookup table is used to verify the anti-rollover and sideslip safety speed limits under this curvature. ,satisfy The safe operating limits. Steering generates lateral acceleration. It meets the set observable threshold. Activate Kalman observation update.

[0031] The measured reaction force difference of the force sensing element during steady-state operation of the turn. For example, substituting the values ​​into the observation equation to solve for the corresponding instantaneous physical observation is: The recursive steps of the filtering algorithm are as follows: The first step is to predict the prior state. The prior error covariance extrapolated is ; The second step is to calculate the Kalman gain. ; The third step is to perform a posterior update based on the observations to obtain the state estimate. Updated covariance ; In subsequent filtering cycles, with continuous filter updates, the posterior variance monotonically converges and remains at the preset convergence threshold. The height estimate eventually converges smoothly to... This process clearly demonstrates the recursive behavior of the Kalman filter, which rapidly converges from the initial a priori density estimate to the physical truth. When the mobile robot is traveling straight or turning, the acceleration... When the lateral load transfer is deemed unobservable, the Kalman gain is frozen and the previous height estimate is kept unchanged.

[0032] The timing sequence for reaction force difference measurement and centroid height filtering estimation of the turning section is as follows: Figure 3 As shown. Figure 3 Includes two corresponding waveforms: Figure 3 In the diagram, (a) represents time. [ The vertical axis represents the measured lateral acceleration. [ Difference between vertical reaction force and support unit [ ];exist Upon entering the curve, the lateral acceleration naturally decreases. Increase to Exceeding the threshold value reaction force difference The mean is formed Furthermore, the measured signal is superimposed with measurement noise; Figure 3 In (b), the horizontal axis is aligned with time in (a), and the vertical axis on the left represents the estimated centroid height. [ The logarithmic vertical axis on the right represents the covariance of the estimation error. [ The estimated value is derived from the initial value. Begin the recursion, updating to step 1. And after going through the filtering window, it tends to converge to the value. Estimation error covariance From initial value Continues to decay, falling below the convergence threshold .

[0033] S204. Considering the internal coupling effect in mechanics caused by widening the support span, further compensation is implemented to improve the observation sensitivity after widening. Specifically, the support span before widening... From the torque balance equation The theoretical support reaction force difference under the above turning condition is calculated to be: ; Symmetrical span to Subsequently, the same lateral moment is distributed over a larger physical span, and the difference in reaction force decreases to... In other words, while increasing the load-bearing capacity and overturning margin, the span reduction also weakens the signal-to-noise ratio of the reaction force sensor. Therefore, the edge computing nodes update the geometric span after the span reduction action is locked. Rigidly written into the observation matrix parameters; when the measured reaction force difference is lower than the set measurement lower limit threshold Temporarily suspend observation updates; or set up identification and guidance sections at the beginning of the path, until the centroid height estimation variance converges to the preset threshold ( Previously, the support span widening was temporarily suspended, and the widening was carried out after the centroid parameters were identified and converged.

[0034] S205 defines the shortest operating time for each road segment as the speed profile operating time that takes into account the maximum acceleration and jerk limits of the drive system. Specifically, the onboard controller bases this on the drive system's dynamic limits—maximum acceleration... Maximum jerk Generate a trapezoidal or S-shaped continuous velocity profile, and extract the shortest travel time of the road segment using the following formula. : (11) In the formula, For road section Physical centerline length ; and Acceleration and deceleration time ; and To accelerate / decelerate the sliding distance Taking the section of road after a right-angle turn with reduced speed in condition B as an example, let's take... Nominal speed of entrance connection Limited export speed ,according to Braking distance of deceleration Deceleration duration The remaining sections traveled at a constant speed for a period of time. Therefore, the shortest travel time for this section of the road is... .

[0035] S206, the time to reach the topology node is represented as an arrival time interval formed by adding the shortest running time and the time corresponding to the preset minimum speed. Specifically, the arrival time interval refers to a continuous closed interval generated by progressively adding segments using the shortest theoretical running time of the mobile robot along the road segment as the lower bound and the running time corresponding to the minimum safe speed for preventing obstruction as the upper bound, which characterizes the range of physically feasible times to reach the target topology node. It is determined by the following formula: (12) In the formula, For continuous interval addition operators, satisfying ; To prevent stalling, minimum speed Following the previous example, the upper bound of the interval time is... Determined, if the preceding time reference The cumulative arrival and arrival time intervals generated for this road segment are: The width of a single segment reaches Without a baseline repair mechanism, simply accumulating five similar road sections consecutively will cause the interval width to monotonically diverge to approximately [value missing]. This will severely strain spatiotemporal graph resources and cause multi-machine spatiotemporal mutual exclusion constraints to degenerate into infeasible solutions. Therefore, in the topology design, the maximum number of consecutive anchorless road segments between any two adjacent resynchronization anchor points is limited to no more than [a certain number]. indivual.

[0036] S207: When a mobile robot enters a centralized buffer area or lingers at an assembly station, the upper and lower limits of the arrival time interval are reset to the actual departure time, and the loading and unloading time interval is included in the node occupancy time. Specifically, in the workshop topology, each centralized buffer area and assembly station is defined as a physical resynchronization anchor point. When the... Taiwanese mobile robot enters anchorage site During stationary operations, the system obtains the physical time when the operator actually leaves the station. And force the upper and lower bounds of the node's departure time interval to align and reset: (13) The reset action compresses the interval width to zero, effectively cutting off the time uncertainty drift accumulated in the upstream path. Based on this, the assembly operation time of the mobile robot at the site is determined using a bounded time interval. (Nominal working hours) This means that the time interval is directly included in the time and space occupancy constraint equation of the site node, avoiding the compression of process cycle time caused by using fixed nominal point values.

[0037] S208 introduces Boolean order variables and constructs disjunctive inequality constraints in the form of the Big M method on the shared topology nodes and shared road segments of any two mobile robots; it restricts the departure time interval of the preceding robot at the shared topology node, plus the node occupancy time and safety time interval, to be earlier than the arrival time interval of the following robot, where the departure time interval is the sum of the arrival time interval and the node occupancy time interval. Specifically, a departure time interval is introduced. For shared topology nodes Introduce a Boolean variable indicating the order in which items are passed. ( Indicates mobile robot Prior to mobile robots Occupying nodes , Conversely). Define the collision-free isolation time algebraic term for nodes. and And construct the disjunctive inequality of the Big M method: (14) (15) (16) In the formula, Safe time interval for topology nodes to prevent rear-end collisions and traffic disturbances ; To be a sufficiently large positive constant, plan the time domain ; For mobile robots Assembly time at this node (0 for ordinary nodes, nominal value for loading / unloading stations). ); For mobile robots Reaching the node upper bound of time interval ; To leave the node upper bound of time interval ; For subsequent mobile robots Lower bound of the time interval for reaching this node .

[0038] Using the upper bound of the reach interval obtained from example S206 The nominal value is used for site assembly time. Leaving the upper bound of the interval is Safe distance When discrete decision settings are used for mobile robots Priority passage, i.e. When, the inequality degenerates into:

[0039] The solution yields the subsequent mobile robot. The lower bound of the physical time allowed to reach this node is a hard constraint. The reverse inequality contains terms. And natural relaxation holds. For solving the Boolean order variable, the edge computing nodes first generate an initial topology sort based on a heuristic rule of the urgency of assembly materials and the remaining running time of each machine. Under a fixed Boolean order, the problem is transformed into a convex continuous velocity profile programming problem, which can be solved quickly within the control cycle. When the first-level solution of the bottom-level feedback fails to find a feasible solution, the Boolean value of the conflict node is flipped (setting...). This enables efficient reconstruction of discrete conflict topologies.

[0040] S209 stipulates that on bidirectional single-track sections, two mobile robots are prohibited from simultaneously occupying opposite lanes. Specifically, mobile robots must leave the shared lane. The time interval is determined by the sum of the entry interval and the running time interval, i.e. For shared one-way road segments, the same-direction rear-end collision avoidance constraint requires that the entry time of the subsequent mobile robot must be later than the departure time of the preceding mobile robot plus a safety time interval, as expressed below: (17) For a set of bidirectional single-track operating road segments Define the direction of movement of the mobile robot on this road segment as an identifier variable. (in This indicates that the operation will proceed in the set positive direction. Indicates running in the opposite direction. Indicates in Unoccupied road sections The Boolean proposition strictly prohibits occupancy in the opposite direction at any consecutive time intervals, as expressed below: (18) It should be noted that the safety linear velocity lookup table is not limited to a three-dimensional offline mesh table combined with trilinear interpolation. In another optional embodiment, it can also be equivalently replaced by an offline pre-calibrated ternary polynomial response surface regression algebraic equation, or by using a compact radial basis function (RBF) surrogate model for analytical expression, where the dimensions of the input independent variables (curvature, residual eccentricity, and centroid height) are consistent with the dimensions of the output physically permissible velocity; the replacement range of the underlying actuators and sensors on which this step depends inherits the mechanical and sensing boundaries described in step S1. All the above replacements do not depart from the concept of discretizing the continuous dynamic boundary into an algebraic velocity envelope fed forward to the discrete planning layer.

[0041] This step transforms the physical boundaries of continuous rollovers and skids into an algebraic velocity lookup table and extreme values ​​of road segment operation times. Bounded arrival time intervals are generated using interval arithmetic. At loading and unloading stations, resynchronization benchmarks are reset using physically measured times, resolving computational delays and fixed-time-window divergence deadlocks caused by online nonlinear dynamics solutions. Furthermore, a spatiotemporal mutual exclusion model of nodes and road segments is established using the Big M method to unify disjunctive logic. This achieves complete decoupling of continuous dynamic constraints to the discrete programming layer and enables deadlock-free, collision-free global collaborative delivery within deterministic time domain boundaries.

[0042] S3, during the delivery process, when the attitude leveling control quantity is saturated or the downstream prediction delay exceeds the limit, the latch event trigger flag is activated and the delivery sequence of the associated robot is reconstructed within the local spatiotemporal subgraph, while keeping the external sequence of the subgraph unchanged.

[0043] In this embodiment, a causal chain is constructed from the saturation of the underlying physical execution mechanism to the discrete-time planning layer, and a state machine is used to achieve reliable latching and hysteresis reset of disturbance events: S301, when the attitude leveling control quantity reaches the saturation threshold, outputs a deceleration protection command to reduce the target speed in steps, thus controlling the decrease in the measured speed of the mobile robot. Specifically, while the mobile robot is carrying large materials along the planned path, the vehicle body is excited by ground unevenness, start-stop inertia, or centrifugal tilting torque. The underlying attitude leveling controller outputs a leveling actuator drive voltage command. The driving leveling actuator is used for correction, and the Boolean function for saturation judgment of the control quantity and the speed target instruction for deceleration protection are defined as follows: (19) In the formula, For the first The attitude leveling control quantity of the control cycle (the analog drive control voltage of the leveling actuator). ; To level the saturation voltage limit of the actuator ; The safety ratio coefficient for saturation discrimination (dimensionless); The deceleration protection ratio reduction factor (dimensionless) is used when the attitude is disturbed. Nominal operating speed ; The safe permissible speed for the current road segment is calculated from S202. ; For the controlled deceleration target command output to the drive motor The voltage saturation threshold is When passing through a corner section under operating condition B. ,when hour The system outputs a speed reduction protection command. Speed ​​setpoint from Step-by-step reduction to The underlying drive motor accelerates. Deceleration, duration A new controlled steady-state velocity is then reached. This leads to the construction of a "saturation leveling control quantity". Output speed reduction protection command Actual operating speed was controlled to decrease The clear physical causal chain of "spatiotemporal drift occurs when downstream nodes arrive" is established.

[0044] S302, calculate the downstream predicted delay by integrating the difference between the measured filtered speed and the nominal speed within the range of downstream monitoring nodes. Specifically, the downstream predicted delay refers to the absolute delay deviation in arrival time estimated by integrating and accumulating the measured filtered speed and the lower limit of the safe allowable speed for each road segment along the planned topology path forward in the direction of travel to the nearest downstream monitoring node when the current actual travel position lags behind the nominal trajectory due to a speed reduction caused by a disturbance in the underlying actuator. It is determined by the following formula: (20) In the formula, For downstream monitoring nodes, a shared topology node is defined as the nearest node along the current direction of travel whose arrival time interval overlaps with that of other mobile robots. This is considered within the planning time domain. If there are no conflicting nodes, then select the next node and synchronize the anchor point. The current node To downstream monitoring nodes The set of planned road segments between; and These are the nominal arc length and actual arc length along the path planning line at the current moment, respectively. ; The cumulative amount of historical delays that have occurred so far. ; For road section Physical centerline length ; The nominal design speed of the road section ; To obtain the smooth steady-state velocity from the measured linear velocity signal by low-pass filtering with a sliding time window The sampling step size is taken from the control period. Slide window to retrieve Shoot (corresponding to physical time window) ).

[0045] Set the downstream monitoring node to be 3 road segments away from the current location, and each segment The nominal speed is Filtered measured steady-state velocity Safe speed limits for subsequent road sections Take the minimum value of the denominator Then the predicted delay increment for a single road segment is The incremental algebraic sum of the three consecutive road segments is The current measured mobile robot lags behind the nominal arc length. Substitute the historical delays that have already occurred into the calculation. The total downstream predicted delay is obtained after superposition. Exceeding the set delay trigger threshold The delay exceeding the limit event has occurred. It should be noted that the delay trigger threshold... Greater than the topology node safe time interval The reason is that the upper bound of the arrival time interval of each topology node has been set according to the minimum speed for preventing obstruction. A conservative envelope was applied, which can smoothly absorb operational fluctuations within a certain range; Set as It is specifically designed to determine whether the delay has substantially exceeded the range defense margin, thus avoiding frequent invalid replanning caused by minor speed disturbances.

[0046] S303, when the attitude leveling control quantity reaches the saturation threshold, the downstream predicted delay exceeds the delay threshold, or the measured speed is lower than the preset lower limit, and the time interval between the last trigger and the current trigger meets the anti-shake cycle, the event trigger flag is set and latched; when the attitude leveling control quantity, the downstream predicted delay, and the measured speed all recover to the corresponding hysteresis interval, the event trigger flag is cleared. Specifically, the onboard motion controller constructs a dual-threshold hysteresis state machine, executing "OR logic set / AND logic reset (OR-Set / AND-Reset)": (twenty one) (twenty two) In the formula, For actual instantaneous velocity ; Debouncing time constant for events ; The time interval since the last event was set. ; To balance the hysteresis constant of the control voltage ; Hysteresis constant for predicting delay When set, it occurs when the "balancing voltage reaches the saturation threshold". "Downstream forecast delay exceeds" "The measured speed dropped below" "If any one of the three conditions is met and the hardware timing meets the anti-shake lockout cycle..." hour, When the transition is set to 1, the event trigger flag is set and latched high by the internal register, and a replanning service request is then sent to the edge computing node; during reset, the event trigger flag is set high only when the leveling voltage drops back to 1. The following downstream forecasts are delayed and rolled back to The following, and the actual measured speed recovered to When all three conditions above are met, If the condition is determined to be 1, the event trigger flag is cleared and the latch is released. The synergy between the dual hysteresis band and the anti-jitter cycle suppresses state repetitive jitter and frequent switching from the control law level.

[0047] The timing relationship between saturation of the leveling control quantity, speed reduction, integral of the prediction delay, and the state of the event trigger flag is as follows: Figure 4 As shown. Figure 4 It consists of four signal channels sharing a common time axis: Figure 4 (a) The first channel in the figure displays the leveling actuator drive voltage. exist When the disturbance reaches the saturation threshold After the disturbance weakens, it falls back to the reset threshold. the following; Figure 4 (b) The second channel displays the speed command and the filtered measured speed, and the speed target command. When the voltage reaches The time ladder was lowered to Actual speed of filtering through After deceleration, the temperature stabilized. ; Figure 4 (c) The third channel shows the predicted latency of the downstream monitoring nodes. As the vehicle speed decreases, the integral is accumulated segment by segment. achieve And exceeding the delay threshold Subsequently, it will gradually return to the reset threshold as the vehicle speed is adjusted. the following; Figure 4 The fourth channel (d) displays the event trigger flag output level when the delay exceeds the limit and the time interval meets the requirement. The timer is set high and latched until the variables of the first to third channels all fall back to the reset threshold and the measured speed is not lower than the reset threshold. It then returns to a low level.

[0048] S304, extract the set of associated robots whose arrival time intervals intersect with the disturbed robot within a preset time domain to construct a local spatiotemporal subgraph. Specifically, the local spatiotemporal subgraph refers to the set of associated robots within a preset look-ahead planning time domain. Within this framework, focusing on disturbed mobile robots experiencing dynamic disturbances or significant delays, a directed acyclic local topology set is constructed by extracting sets of associated robots whose arrival time intervals on shared nodes or road segments have non-empty intersections through topological traversal. The set of associated robots is determined by the following formula: (twenty three) In the formula, The current event trigger time ; and Mobile robots and Reaching the node Arrival time interval ; This represents the complete set of nodes in the workshop road network topology. Edge computing nodes simultaneously perform boundary isolation operations: for mobile robots located outside the local spatiotemporal subgraph, their existing time intervals for reaching each topology node, continuous velocity profiles of road segments, and Boolean order variables are strictly kept unchanged, frozen in place, to suppress the spread of local disturbances to the entire field.

[0049] S305 performs a four-level progressive resolution within the local spatiotemporal subgraph: The first-stage digestion optimizes the continuous velocity profile under a fixed order; When the first-level resolution fails to resolve the conflict within the preset time limit, the second-level resolution reverses the order variable of the conflict node and applies a follow-up speed limit to the subsequent robot. When the second-level solution has no solution, the third-level solution searches for local bypass branches within the local spatiotemporal subgraph and verifies the clearance of the new path. When the third-level solution has no solution, the fourth-level solution keeps the time interval of the robots outside the subgraph unchanged and performs full-graph constrained replanning on the set of disturbed robots.

[0050] Specifically, the first-level unfreezing process removes the original Boolean order variables of each shared node within the frozen subgraph. Only the continuous velocity profile of each associated robot To optimize the decision variables, a convex quadratic programming algorithm is used, with a set pure algorithmic solution time limit of [time value missing]. After accounting for the overhead of subgraph extraction, data serialization, and communication packetization, the end-to-end control delay of the first-level resolution does not exceed [a certain value]. If the first-level resolution fails to converge to a feasible solution within the time limit, it indicates that the original passage order is no longer feasible under the current delay. The system then flips the Boolean order variable of the conflicting nodes (let...). Simultaneously, a speed limit command is issued to the obstructed mobile robot that is adjusted to pass later, increasing the spatial physical distance between it and the preceding mobile robot. If the second-level resolution iteration still yields no solution, the system searches for physical alternative bypass branches within the local spatiotemporal subgraph, and immediately calls the channel clearance geometry verification steps of S104 and S105 after determining the rerouting path, based on the physical channel width of the local bypass branch. Substitute into the verification span The clearance for the material outline on the new branch road is verified to meet the requirements. The new path is then introduced into the subgraph reconstruction. If the three-level resolution still cannot eliminate the blockage in the local area, the scope of replanning is expanded, and the time interval occupied by all mobile robots outside the local spatiotemporal subgraph is taken as a fixed and unassailable rigid spatiotemporal boundary condition. The whole graph restricted replanning is performed only with the delivery time sequence of the disturbed robot set as the solution variable to avoid affecting other normally executed transportation tasks in the field.

[0051] S306, in optimizing the continuous velocity profile during the first-level solution, the objective function is to minimize the weighted sum of the normalized arrival time deviation, the deviation from the nominal speed command, and the arrival time variance. This is done while satisfying the spatiotemporal mutual exclusion constraints, the upper and lower bound constraints on velocity, and the constraints on maximum acceleration and maximum jerk. Specifically, the objective function for continuous velocity profile optimization is constructed as follows: (twenty four) (25) In the formula, The nominal workstation cycle time for mixed-line assembly production lines, also serving as a time normalization factor. ; and Mobile robots Optimized estimated arrival time at workstation and planned nominal arrival time ; A dimensionless weighting coefficient assigned based on the urgency of material shortages at the workstation assembly level; To optimize the solved velocity vector function ; The nominal speed vector for each road segment ; The dimensionless penalty coefficient for velocity deviation; The dimensionless balance coefficient for the discrete variance of arrival time; Discrete variance operator for the time when each associated mobile robot in the subgraph reaches its respective target assembly station. It is used to mitigate the backlog of workstations caused by multiple machines delivering concurrently. Optimize the weighted functional (dimensionless) for multiple objectives.

[0052] set up , , , Then the first item is The second item is The third item is The sum of the three items is The temporal reshaping of the aforementioned local spatiotemporal subgraph is entirely driven by event-triggered flags, executed asynchronously by edge computing nodes, and does not occupy the underlying cycle of the mobile robot. Real-time motion control time slots for wheels and tracks.

[0053] Comparison of the spatiotemporal trajectory and velocity curves of the mobile robot before and after local spatiotemporal subgraph reconstruction. Figure 5 As shown. Figure 5 Includes a spacetime diagram (a) and a velocity profile (b): Figure 5 In (a) of the diagram, the horizontal axis represents time. [ [The vertical axis represents the path distance] [ The strip region represents the arrival time interval of each topological node; before reconstruction, the disturbed mobile robot... Due to deceleration The delay causes its spatiotemporal trajectory to shift to the right and merge with the subsequent mobile robot. At the node The arrival intervals at the location overlap; Figure 5 In (b) of the diagram, the horizontal axis is... Figure 5 (a) corresponds to time, and the vertical axis represents velocity. [ The curve represents the continuous velocity command obtained through first-stage elimination and optimization. and ; in acceleration With jerk Under constraints, the order variable is maintained. Subsequent mobile robots After adjusting the velocity profile, it reaches the node. The lower bound of the time interval is shifted to To meet the safe time distance ; Figure 5 In (a), the spatiotemporal trajectories of the remaining mobile robots outside the local spatiotemporal subgraph remain consistent before and after reconstruction.

[0054] It should be noted that the attitude leveling controller that generates the output saturation control quantity is not limited to incremental digital PID control law. In another optional embodiment, it can also be equivalently replaced by a positional digital PID controller, or a proportional-derivative-post-integral (PI-D) control algorithm with an anti-integral saturation element. The above control laws are all feedback controllers with output limited by amplitude physical characteristics and have a clear saturation cutoff threshold. The following speed limit control in the secondary resolution can also be replaced by decelerating and waiting at a fixed distance in front of the conflict topology intersection, or by a longitudinal distance maintaining algorithm based on multi-vehicle virtual formation control law. In the tertiary resolution, the local bypass branch search can be replaced by dynamic window method (DWA) or artificial potential field method (APF) to perform local collision avoidance and detour, in addition to heuristic path retrieval based on topology map. All the above replacements do not depart from the concept of the present invention of repairing the spatiotemporal reference by local back-injection triggered by events.

[0055] This step establishes a physical state transmission chain of "leveling saturation -> speed reduction protection -> measured speed reduction -> look-ahead delay integration -> dual hysteresis latch," precisely isolating the minimum-association spatiotemporal subgraph on the directed graph using interval intersection. A four-level progressive mechanism and continuous speed optimization are employed to reconstruct the timing sequence. This solves the problem of physical delays caused by the saturation or speed reduction protection of the bottom-level leveling mechanism failing to feed forward when encountering uneven road surfaces or large eccentricity instability during material transportation. This leads to the failure of the upper-level spatiotemporal mutual exclusion model and convergence deadlock at intersection nodes, resulting in task disconnection. The goal is to confine disturbances to a topologically associated local subgraph (typically no more than 6 mobile robots) and smoothly resolve them within milliseconds, while maintaining the external timing sequence of the subgraph, thus avoiding computational oscillations caused by global replanning.

[0056] S4, control each mobile robot to deliver materials according to the delivery sequence. In this embodiment, the discrete passage order, node arrival time interval reconstruction solution, and continuous time velocity trajectory obtained from the upper layer solution are distributed to the vehicle control system of each associated mobile robot to coordinate the collaborative action of the drive wheel system, steering mechanism, pitch mechanism, and end-effector positioning tooling, as follows: S401: After the edge computing node completes the local spatiotemporal subgraph reshaping, it packages and sends the calculated control parameters to the onboard motion controllers of each mobile robot. The sent data payload includes a sequence of continuous-time target speed control commands for each mobile robot on each segment of its planned path. Mutual exclusion access Boolean order variables for each shared topology node and shared road segment Determine the value, and the updated departure time of the reference point after resynchronization along the route. The aforementioned control parameters are transmitted and distributed via real-time industrial Ethernet or industrial 5G wireless private network links deployed in the workshop. Since the reshaping and solving of the local spatiotemporal subgraph is completed independently at the edge computing nodes, the wireless communication link only carries the periodic distribution of constant packets of the optimization solution results and the return of underlying state variables in a one-way manner. The physical transmission delay of the link itself is completely independent of the computation time of the timing reconstruction algorithm and is not included in the computational time budget for reshaping. The underlying motion servo closed loop of the mobile robot is calculated and executed entirely locally in real time by the onboard controller, without relying on the high-frequency interaction of the wireless communication link.

[0057] In S402, the control system is strictly divided into two control levels with different scales in the time dimension. In the underlying real-time motion control clock, the servo drives of the mobile robot's underlying drive motors and the chassis steering controller operate strictly according to the control cycle. (Corresponding execution frequency) ) Executes fixed-cycle closed-loop interrupt servo control, based on the issued... The path guidance updates the linear and angular velocity control values ​​of the drive motor in real time; in the upper-level spatiotemporal replanning asynchronous clock, the timing reconstruction of the local spatiotemporal subgraph is a non-periodic asynchronous calculation process driven by the event trigger flags latched by the lower-level state machine. In the state interaction mechanism between the two, the trigger debouncing time constant set by the state machine... Specifically designed to prevent high-frequency replanning oscillations in edge computing nodes caused by concurrency delays from multiple vehicles, the reconstruction algorithm's actual end-to-end solution and interaction time within the local subgraph strictly converges within millisecond time limits (where the total end-to-end latency of the first-level resolution does not exceed...). The time limit of the pure solution algorithm In the system architecture, the two belong to "asynchronous task triggering constraints" and "synchronous continuous servo cycles" respectively, and have clear hierarchical boundaries.

[0058] S403: When the mobile robot arrives at the centralized buffer area of ​​the assembly station or the production line penetration replenishment station on time according to the reconstructed sequence, it enters the terminal physical transfer process. First, it performs pose fine-tuning and alignment: activating the laser displacement sensors installed at the front and sides of the vehicle body, or aligning them with the station's physical reference pins through the vehicle body's mechanical positioning holes, to measure the longitudinal positional deviation between the vehicle body's load-bearing center and the station's material receiving bracket. Lateral position deviation The vehicle-mounted motion controller drives the chassis to perform minor forward, lateral, or small-angle rotation corrections to eliminate alignment deviations; subsequently, it performs vertical load transfer: after posture correction and alignment, it controls the variable-pitch support unit of the flexible carrier to synchronously move the vertical drive pair, smoothly lowering the material at a controlled micro-motion speed. Large, irregularly shaped workpieces are smoothly transferred to the assembly station fixture and the physical load contact is released; finally, the time reference is recorded: after the transfer is completed, the actual physical moment when the mobile robot leaves the assembly station is captured. It also reports to the edge computing node and forces clock resynchronization anchoring according to S207 to cut off the cumulative drift of the machine in subsequent task intervals.

[0059] It should be noted that the mobile robot's chassis is not limited to a two-wheel differential steering wheel system. In another alternative embodiment, it can be equivalently replaced by a multi-steering wheel independent steering chassis, a Mecanum wheel omnidirectional moving chassis, or a heavy-duty dual-track walking mechanism; the material delivery action of the flexible vehicle, in addition to the vertical fine-tuning and lowering of the variable pitch support unit, can also be equivalently replaced by a chassis scissor-type hydraulic lifting mechanism, an active double-row roller conveyor self-driven docking conveyor mechanism, or a multi-degree-of-freedom industrial onboard robotic arm mounted on the vehicle body to perform collaborative grasping and handover. All the above substitutions do not depart from the concept of the present invention.

[0060] This step employs a deterministic industrial communication bus to distribute timing calculation packets, strictly isolating the underlying servo control cycle. ) and upper-level event stabilization cycle ( With the integration of terminal multi-sensor closed-loop fine-tuning docking and real-time anchoring and back-injection, the problem of the discrete mutually exclusive sequence and continuous velocity curve generated by the upper layer calculation being unable to be precisely synchronized and linked with the underlying multi-physical degree of freedom actuators, resulting in the disconnection of control command issuance and excessive deviation in the alignment of end delivery, is solved. This achieves highly consistent and stable execution of the reconstructed delivery sequence at the multi-machine physical layer and smooth penetration and handover of large items at the end.

[0061] The above embodiments can be implemented, in whole or in part, by software, hardware, firmware, or any other combination thereof. When implemented using software, the above embodiments can be implemented, in whole or in part, in the form of a computer program product.

[0062] Those skilled in the art will recognize that the modules and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0063] In addition, the functional modules in the various embodiments of this application can be integrated into one processing module, or each module can exist physically separately, or two or more modules can be integrated into one module.

[0064] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

[0065] In conclusion, the above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A centralized buffering and workstation-penetrating material delivery control method for complex mixed-assembly lines, characterized in that, include: The eccentricity of the material's center of mass relative to the longitudinal symmetry plane of the vehicle body is calculated based on the force distribution of each support unit and the coordinates of the support points. The variable pitch mechanism is then used to translate the support unit to compensate for the eccentricity of the center of mass. The safe allowable speed of each segment of the planned path is determined based on the compensated residual eccentricity, the arrival time interval of each topology node is generated, and the width of the time interval is reset at the loading and unloading station to solve the delivery sequence that satisfies the spatiotemporal mutual exclusion constraint. During the delivery process, when the attitude leveling control quantity is saturated or the downstream predicted delay exceeds the limit, the latch event trigger flag is activated and the delivery sequence of the associated robot is reconstructed within the local spatiotemporal subgraph, while keeping the external sequence of the subgraph unchanged. The delivery sequence is used to control each mobile robot to deliver materials.

2. The method according to claim 1, characterized in that, Calculate the eccentricity of the material's center of mass relative to the longitudinal symmetry plane of the vehicle body and drive the translational support unit of the pitch mechanism to compensate for the eccentricity, including: Based on the initial support spacing determined by the geometric envelope of the material, the plane projection coordinates of the material's center of mass on the bearing surface are calculated using the static moment equilibrium equation, according to the vertical reaction force collected by each support unit and the coordinates of its support points. The drive mechanism moves the support unit laterally within its maximum stroke, causing the projection of the material's center of mass onto the bearing surface to approximate the longitudinal symmetry plane of the vehicle body.

3. The method according to claim 2, characterized in that, Also includes: The maximum allowable span is determined based on the channel width, vehicle body width and safety clearance. Under the premise of meeting the channel clearance, the left and right support units are driven to symmetrically expand to increase the support span, and the overall width of the machine after expansion is included in the clearance check. When the outer contour of the translated material exceeds the clearance of the channel, the translation amount is cut off to meet the clearance of the channel, and the vertical reaction force of each support unit after translation is collected to recalculate the residual eccentricity.

4. The method according to claim 1, characterized in that, Determine the safe permissible speed for each segment of the planned route, including: Based on the mobile robot's half wheel track, maximum linear velocity, sideslip limit, and lateral anti-tipping arm determined by residual eccentricity and steering curvature, a safe linear velocity lookup table is constructed for path curvature, residual eccentricity, and center of mass height. On each segment of the planned path of each mobile robot, the corresponding safe allowable speed is obtained by retrieving the safe linear velocity lookup table using trilinear interpolation.

5. The method according to claim 4, characterized in that, The centroid height is identified online through the following steps: The total mass of the material is determined by the sum of the vertical reaction forces of each supporting unit; When the mobile robot passes through a turning section with known curvature and the lateral acceleration reaches a preset threshold, the difference in vertical reaction force between the left and right support units is collected. An observation equation is constructed using the total mass of the material, the height of the support surface above the ground, the support span, and the difference in vertical reaction force. The estimated value of the material density is used as the initial value, and the estimated value of the material centroid height is updated online using Kalman filtering.

6. The method according to claim 1, characterized in that, Generate arrival time intervals for each topology node and reset the time interval width at loading / unloading stations, including: A speed profile is generated based on the maximum acceleration and maximum jerk limit of the drive system, and the running time of the speed profile is determined as the shortest running time for each road segment. Using the shortest running time as the lower bound and the time corresponding to the preset minimum speed as the upper bound, the arrival time intervals of each topological node are generated by accumulating segment by segment through interval addition. When the mobile robot enters the centralized buffer area or stays at the assembly station, the upper and lower limits of the arrival time interval are reset to the actual departure time, and the loading and unloading time interval is included in the node occupation time.

7. The method according to claim 1, characterized in that, The spatiotemporal mutual exclusion constraint includes: In any two mobile robots, Boolean order variables are introduced and disjunctive inequality constraints in the form of the Big M method are constructed on the shared topology nodes and shared road segments. The time interval of the departure of the preceding robot at the shared topology node, plus the node occupation time and safety time interval, is earlier than the time interval of the arrival of the following robot. The departure time interval is the sum of the arrival time interval and the node occupation time interval. Two mobile robots are prohibited from occupying opposite sections of the track simultaneously on bidirectional single-track sections.

8. The method according to claim 1, characterized in that: The attitude leveling control quantity saturation includes: when the attitude leveling control quantity reaches the saturation threshold, outputting a deceleration protection command to reduce the speed target in steps, so that the measured speed of the mobile robot is reduced in a controlled manner. The downstream predicted delay is calculated by integrating the difference between the measured filtering speed and the nominal speed within the range of downstream monitoring nodes. It also includes setting and locking the event trigger flag when the attitude leveling control quantity reaches the saturation threshold, the downstream predicted delay exceeds the delay threshold, or the measured speed is lower than the preset lower limit, and the time interval with the last trigger meets the anti-shake cycle. When the attitude leveling control quantity, downstream predicted delay quantity, and measured speed all recover to the corresponding hysteresis interval, the event trigger flag is cleared.

9. The method according to claim 1, characterized in that, Reconstruct the delivery sequence of associated robots within the local spatiotemporal subgraph, including: Extract the set of associated robots that intersect with the disturbed robot in the arrival time interval within the preset time domain to construct a local spatiotemporal subgraph; Perform a four-level progressive resolution within the local spatiotemporal subgraph: The first-stage digestion optimizes the continuous velocity profile under a fixed order; When the first-level resolution fails to resolve the conflict within the preset time limit, the second-level resolution reverses the order variable of the conflicting nodes and applies a speed limit to the subsequent robot. When the second-level resolution has no solution, the third-level resolution searches for local bypass branches within the local spatiotemporal subgraph and verifies the clearance of the new path. When the third-level solution has no solution, the fourth-level solution performs full graph replanning on the disturbed robot set, and the full graph replanning uses the delivery time sequence of the disturbed robot set as the solution variable and the time interval of the robots outside the subgraph as the fixed boundary condition.

10. The method according to claim 9, characterized in that, When optimizing the continuous velocity profile in the first-level solution, the objective function is to minimize the weighted sum of the normalized arrival time deviation, the deviation from the nominal speed command, and the arrival time variance. The continuous velocity command of each mobile robot is solved under the constraints of spatiotemporal mutual exclusion, velocity upper and lower bound constraints, and maximum acceleration and maximum jerk constraints.

Images