A postal logistics warehouse-oriented goods transportation planning method

Abstract

The application relates to the technical field of path planning, in particular to a kind of goods transportation planning method for postal logistics warehousing. First, the speed, motor current, path curve, goods point cloud, centroid, inclination angle and other data of AGV are obtained; second, by fusing the dynamics parameters such as speed deviation accumulation, rudder angle deviation, motor current characteristics and path curvature mutation, the dynamic cost function value of AGV real-time performance is quantified; in the obstacle avoidance scene, combined with the change of goods form (highlight detection), task urgency and steering cost, the penalty influence function value is generated; finally, the candidate path is globally screened through the multi-objective fusion model (cost function + penalty value + path length). The application embodiment can effectively improve the matching degree of dynamic change of individual AGV and global task through the three-level architecture of "data-driven-individual optimization-global collaboration", and ensure the overall transportation efficiency.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of path planning, in particular to a kind of goods transport planning method for postal logistics warehouse. BACKGROUND

[0002] In the field of logistics warehouse automation, the path planning efficiency and safety of automatic guided vehicle (AGV) directly affect the throughput and operating cost of the overall warehouse system. With the upgrading of logistics warehouse automation demand, the scheduling and path planning technology of AGV (automatic guided vehicle) has experienced an evolution process from static rules to dynamic optimization, which has to some extent solved the real-time disturbance problems such as order insertion, equipment failure or traffic congestion in dynamic environment.

[0003] However, the existing technology often uses the shortest path to update when performing upper path dynamic planning, but in the actual process of AGV running goods, the physical state of goods (center of mass offset, inclination angle) will affect the stability of AGV driving, and AGV will produce speed fluctuation, path mutation induced steering lag and other dynamic problems due to load, these factors will cause AGV's chain path adjustment when obstacle avoidance action occurs, therefore, if the matching degree of individual AGV's dynamic change and global task is ignored, it will easily fall into local optimal path, affecting the overall transportation efficiency. SUMMARY

[0004] In order to solve the technical problem that in the actual process of AGV running goods, the physical state of goods (center of mass offset, inclination angle) will affect the stability of AGV driving, and AGV will produce speed fluctuation, path mutation induced steering lag and other dynamic problems due to load, these factors will cause AGV's chain path adjustment when obstacle avoidance action occurs, therefore, if the matching degree of individual AGV's dynamic change and global task is ignored, it will easily fall into local optimal path, affecting the overall transportation efficiency, the purpose of the present application is to provide a kind of goods transport planning method for postal logistics warehouse, the technical scheme adopted is as follows:

[0005] In the preset period corresponding to the current time, the mechanical parameters, speed time series data, motor current time series data, path curve, actual rudder angle, point cloud data, the state data of goods carried at the current time and the remaining time of task execution of each AGV are obtained;

[0006] The deviation accumulation between the preset speed and the actual speed time series data of each AGV and the motor current time series data features are fused to determine a mechanical execution conflict factor of each AGV to describe the collaborative conflict between AGV driving planning and mechanical execution; the AGV dynamic curvature margin is obtained according to the mechanical parameters of each AGV and the deviation between the actual rudder angle and the preset rudder angle, and the trajectory replication factor of each AGV is determined to describe the deviation degree of the theoretical turnable margin of the AGV and its actual execution capability in the path turning by combining the bending change of the path curve; the dynamic cost function value of each AGV is determined by fusing the mechanical execution conflict factor and the trajectory replication factor;

[0007] When obstacle avoidance is needed, the updated path and the updated rudder angle of each AGV are obtained, the actual speed time series data, the change of the point cloud data and the cargo state data of each AGV are fused and analyzed, and the brake misfit degree of the AGV in the updated path is calculated, which represents the additional braking distance of the AGV caused by the cargo sway; the obstacle avoidance behavior cost of each AGV in obstacle avoidance is determined according to the brake misfit degree, the length of the updated path, the updated rudder angle and the remaining time of task execution.

[0008] For each possible path update result, the transportation path planning is performed for all AGVs based on the dynamic cost function value of the AGV, the obstacle avoidance behavior cost and the length of the updated path.

[0009] Further, the method for obtaining the mechanical execution conflict factor comprises:

[0010] In the preset time period, in the preset speed time series data and the actual speed time series data of each AGV, the absolute value of the difference between the actual speed and the preset speed at the same time is calculated, and the absolute values at all times are curve fitted to obtain a speed deviation curve;

[0011] The result of the definite integral calculation of the speed deviation curve in the preset time period is normalized to obtain a speed cumulative deviation index of each AGV in the preset time period;

[0012] The ratio of the mean value of all current values in the motor current time series data of each AGV to the preset rated current value is normalized to obtain a load fluctuation index of each AGV in the preset time period;

[0013] The product of the speed cumulative deviation index of each AGV and a preset first weight is taken as a speed imbalance factor, the product of the load fluctuation index of each AGV and a preset second weight is taken as a load imbalance factor, and the normalized value of the sum of the speed imbalance factor and the load imbalance factor is taken as a mechanical execution conflict factor of each AGV, wherein the sum of the preset first weight and the preset second weight is 1.

[0014] Further, the method for obtaining the trajectory replication factor comprises:

[0015] According to the AGV mechanical parameters, a limit curvature value is obtained, which is affected by the AGV wheelbase and the mechanical parameter of the limit rudder angle;

[0016] In a preset time period corresponding to the current time, the curvature value of the position of each AGV at each time is calculated on the path curve of each AGV.

[0017] The difference between the limit curvature and the curvature value of the position of each AGV at the current time is taken as a safety margin coefficient of each AGV.

[0018] The absolute value of the difference between the actual rudder angle of each AGV at the current time and the preset rudder angle is negatively correlated and normalized to obtain an execution reliability coefficient.

[0019] The product of the safety margin coefficient and the execution reliability coefficient of each AGV is negatively correlated and normalized to obtain a mismatch factor of each AGV at the current time.

[0020] The absolute value of the difference between the curvature value of the position of each AGV at the current time and the curvature value of the position at the previous time is taken as a real-time curvature change degree, and based on the real-time curvature change degree, a curvature jump penalty factor of each AGV is determined, and the curvature jump penalty factor is positively correlated with the real-time curvature change degree.

[0021] The product of the mismatch factor and the curvature jump penalty factor of each AGV is normalized to obtain a trajectory replication factor of each AGV.

[0022] Further, the method for obtaining the dynamic cost function value comprises:

[0023] The arithmetic square root of the mechanical execution conflict factor and the trajectory replication factor of each AGV is normalized to obtain a dynamic cost function value of each AGV.

[0024] Further, the method for obtaining the obstacle avoidance behavior cost comprises:

[0025] The cargo state data of each AGV includes a cargo centroid offset, a cargo weight, and a cargo inclination angle.

[0026] According to the cargo state data of each AGV, braking kinetic energy of each AGV is determined, and based on the braking kinetic energy and the friction of each AGV, an additional braking distance is determined, which is used to calculate a braking mismatch degree of each AGV;

[0027] The difference between the length of the updated path of each AGV and the length of the preset path is fused, and the difference between the rudder angle and the preset rudder angle is updated, to determine an obstacle avoidance energy consumption factor of each AGV;

[0028] The sum of the braking mismatch degree of each AGV and the obstacle avoidance energy consumption factor is normalized, and the normalized value is taken as an obstacle avoidance cost index;

[0029] The change of the point cloud data of the AGV at adjacent time instants is analyzed, and the path priority weight of each AGV is determined in combination with the remaining time of task execution;

[0030] The product of the path priority weight of each AGV after negative correlation mapping and the obstacle avoidance cost index of each AGV is normalized to obtain an obstacle avoidance behavior cost of each AGV when avoiding obstacles.

[0031] Further, according to the cargo state data of each AGV, the braking kinetic energy of each AGV is determined, and based on the braking kinetic energy and the friction of each AGV, an additional braking distance is determined, which is used to calculate a braking mismatch degree of each AGV, including:

[0032] The centroid offset of each AGV at the current time instant is taken as a cargo offset speed, the swing kinetic energy of the cargo is calculated according to the weight of each AGV and the cargo offset speed, and based on the swing kinetic energy and the inclination angle of the cargo carried by the AGV, the component of the swing kinetic energy of the AGV in the running direction is calculated, so as to obtain the braking kinetic energy of the AGV;

[0033] The friction is determined based on the weight of the cargo carried by the AGV;

[0034] The ratio of the braking kinetic energy of each AGV to the friction is taken as the additional braking distance of each AGV;

[0035] The ratio of the additional braking distance to the preset braking distance is taken as the braking mismatch degree of each AGV.

[0036] Further, the method for obtaining the obstacle avoidance energy consumption factor includes:

[0037] The product of the ratio of the length of the updated path of each AGV to the length of the preset path and the preset first coefficient is taken as a first penalty coefficient;

[0038] The product of the ratio of the updated turning rudder angle of each AGV to the preset turning rudder angle and the preset second coefficient is taken as a second penalty coefficient;

[0039] The normalized value of the sum of the first penalty coefficient and the second penalty coefficient is taken as an obstacle avoidance energy consumption factor.

[0040] Further, the method for obtaining the path priority weight comprises:

[0041] The envelope volume of the trolley is obtained according to the point cloud data of each AGV at each time, and the difference between the envelope volume at the current time and the envelope volume at the previous time is processed by negative correlation mapping to obtain a priority factor;

[0042] The value of the negative correlation mapping of the task execution remaining time of each AGV is taken as the urgency of each AGV;

[0043] The proportion of the urgency of each AGV in the sum of the urgencies of all AGVs is multiplied by the priority factor, and the normalized value of the obtained product is taken as the path priority weight of each AGV.

[0044] Further, the method for obtaining the path priority weight comprises:

[0045] When obstacle avoidance is needed, the updated cost value of each AGV is determined based on the numerical characteristics of the length of the updated path of each AGV, the dynamic cost function value and the obstacle avoidance behavior cost under each possible path update result;

[0046] The normalized value of the sum of the updated cost values of all AGVs is taken as an obstacle avoidance cost index under each possible path update result;

[0047] The path update result corresponding to the minimum obstacle avoidance cost index is taken as the optimal dynamic path, so as to dynamically plan the transportation path for all AGVs.

[0048] Further, the method for obtaining the path priority weight comprises:

[0049] When obstacle avoidance is needed, the ratio of the length of the updated path of each AGV to the shortest path length under all path update results is taken as a detour index under each possible path update result;

[0050] The sum of the detour index, the dynamic cost function value and the obstacle avoidance behavior cost of each AGV is taken as the updated cost value of each AGV.

[0051] The present application has the following advantages:

[0052] By synchronously acquiring the speed, motor current, point cloud data, task progress and the state of the carried goods (centroid offset, tilt angle) of each AGV, global state perception is realized. In the hierarchical architecture, the upper path planning generates the shortest path based on the ideal environment, while the lower execution mechanism is limited by physical characteristics such as friction, inertia, etc. in actual operation, resulting in a mismatch between path planning and mechanical execution. Therefore, the AGV's speed deviation accumulation, the deviation of the steering angle, the motor current characteristics and the sudden change of the path curvature are fused to quantify the real-time dynamics performance of the AGV, and the dynamic cost function value is obtained. When obstacle avoidance is needed, the centroid offset of the carried goods, the tilt angle of the AGV and the change of the AGV point cloud data are introduced into the obstacle avoidance strategy to identify abnormal shapes such as goods protruding from the vehicle body, dynamically evaluate the actual pose of the goods, and incorporate the task execution progress of the AGV and the updated steering angle into the obstacle avoidance decision, realizing the joint optimization of global path planning and individual safety, and determining the obstacle avoidance behavior cost of each AGV when avoiding obstacles. Finally, for each possible path update result, the dynamic cost function value of the AGV, the obstacle avoidance behavior cost and the length of the updated path are fused to plan the transportation path for all AGVs. Through the three-level architecture of "data-driven-individual optimization-global coordination", the matching degree of the dynamic changes of individual AGVs and the global task can be effectively improved, and the overall transportation efficiency is guaranteed. BRIEF DESCRIPTION OF DRAWINGS

[0053] In order to more clearly illustrate the technical solutions and advantages of the embodiments of the present application or the prior art, the drawings needed in the embodiments or prior art description will be briefly introduced. Obviously, the drawings in the following description are only some embodiments of the present application, and other drawings can be obtained by those skilled in the art without creative labor.

[0054] Figure 1 A method flowchart of a postal logistics warehouse-oriented goods transportation planning method provided by an embodiment of the present application;

[0055] Figure 2 A method flowchart of a dynamic cost function value acquisition method provided by an embodiment of the present application;

[0056] Figure 3 A method flowchart of an obstacle avoidance behavior cost acquisition method provided by an embodiment of the present application. DETAILED DESCRIPTION

[0057] In order to further illustrate the technical means and effects taken by the present application to achieve the predetermined inventive purpose, the following describes in detail the specific implementation, structure, features and effects of a postal logistics warehouse-oriented goods transportation planning method according to the present application, in combination with the accompanying drawings and preferred embodiments. In the following description, different "one embodiment" or "another embodiment" do not necessarily refer to the same embodiment. In addition, the specific features, structures or characteristics in one or more embodiments can be combined in any suitable form.

[0058] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this application belongs.

[0059] The following describes in detail a specific scheme of a postal logistics warehouse-oriented goods transportation planning method provided by the present application.

[0060] Referring to Figure 1 , a method flowchart of a postal logistics warehouse-oriented goods transportation planning method provided by one embodiment of the present application is shown, which includes the following steps:

[0061] Step S1: In a preset time period corresponding to the current time, the mechanical parameters, speed time series data, motor current time series data, path curve, actual steering angle, point cloud data, cargo state data at the current time, and task execution remaining time of each AGV are acquired.

[0062] For example, a large e-commerce warehouse center can process more than 500,000 orders per day, including a two-way channel (width 3.5m), a sorting workstation (including a 360° rotating table), a packaging area, a multi-layer vertical storage area (height 6-12m), and a charging station (deployed at the edge of the sorting area). The edge computing node is deployed in each subarea of the warehouse to process local data (such as AGV obstacle avoidance instruction generation) in real time, reducing the load of the central system. 150 differential drive automatic guided vehicles (AGVs, load capacity 800kg per vehicle, speed adjustable 0-2.5m / s) are provided, which are equipped with a multi-modal fusion positioning and navigation system (laser SLAM + binocular camera + IMU + RFID reader / writer), a double-wheel differential motor, a three-dimensional laser obstacle avoidance, a body sensor, a communication protocol, etc.; the body sensor system (including AGV chassis encoder, Hall sensor, six-axis force sensor, etc.) can read the dynamic parameters of the AGV itself in real time, such as speed, steering angle, motor current (load monitoring), weight of the loaded cargo, centroid offset of the cargo, and inclination angle, etc.

[0063] The environmental perception system comprises a ceiling type 3D laser radar and an infrared thermal imaging camera; through RFID and laser SLAM fusion, a map can be refreshed once every 200 ms, temporary obstacles (such as fallen goods) are marked, and when an obstacle is found, the updated path of each AGV is simulated and refreshed synchronously when the AGV needs to avoid the obstacle, and the moving position of each AGV is marked to obtain a path curve, and the point cloud data of each AGV is obtained.

[0064] In this embodiment of the application, a preset time period corresponding to the current time is constructed, that is, the current time is taken as a starting point, and a length of 1 second is pushed forward in time sequence, as the preset time period corresponding to the current time, then actual speed time sequence data, motor current time sequence data, a path curve, and point cloud data at each time of each AGV trolley in the preset time period are obtained; actual rudder angle of each AGV trolley at the current time, weight of the carried goods, goods centroid offset (Euclidean distance between the centroid at the current time and the centroid at the last time), and goods inclination angle (specifically, the included angle between the goods inclination direction and the AGV running direction) are obtained; at the same time, the remaining time of task execution of each AGV (total time - time spent) is obtained at the current time.

[0065] It should be noted that in this embodiment of the application, the data acquisition frequency is set to 10 ms once, that is, the time sequence data is collected at an interval of 10 ms.

[0066] Step S2: the deviation accumulation between the preset speed and the actual speed time sequence data of each AGV and the motor current time sequence data characteristics are fused to determine the mechanical execution conflict factor of each AGV to describe the collaborative conflict of AGV driving planning and mechanical execution; the AGV dynamic curvature margin is obtained according to the mechanical parameters of each AGV and the deviation between the actual rudder angle and the preset rudder angle, and the trajectory copying factor of each AGV is determined to describe the theoretical turnable margin of the AGV and the deviation degree of its actual execution ability in combination with the bending change of the path curve; the mechanical execution conflict factor and the trajectory copying factor are fused to determine the dynamic cost function value of each AGV.

[0067] Due to the difference in mathematical description dimension between the path planning layer (discrete space-time decision of the central system) and the mechanical execution layer (continuous physical motion of the AGV), there may be a deviation between the theoretical path and the actual trajectory, so in this step, the main collaborative conflict between the upper decision layer and the lower execution layer can be analyzed.

[0068] Under the hierarchical architecture, the upper path planning layer generates a global path based on an ideal environment, and makes the running speed, steering angle, and planned preset path of the AGV; while the lower execution mechanism is limited by physical characteristics in actual operation, such as friction, inertia, load, etc. Moreover, the upper planned path may be a smoothed path, and the AGV torque limit and insufficient encoder accuracy may cause the actual speed and trajectory to deviate from the preset condition; and even if the AGV trolley can pass through in theory when turning during the planning of the preset path, there may be many factors interfering with the turning in the actual execution process, such as discontinuity or even mutation of the curvature at the path connection (such as a right-angle polyline splice), so in this step, the speed, steering angle, motor current, and bending change of the path curve of the AGV are analyzed to determine the dynamic cost function value of each AGV, which is used to describe the contradiction between the path planning of the upper decision layer and the actual turning cost of the lower execution layer.

[0069] Preferably, in an embodiment of the present application, the method for obtaining the dynamic cost function value comprises:

[0070] Referring to Figure 2 , a method flowchart of a method for obtaining a dynamic cost function value in an embodiment of the present application is shown, and the method comprises the following steps:

[0071] Step S201: The deviation accumulation between the preset speed time series data and the actual speed time series data of each AGV is fused with the numerical characteristics of the motor current time series data according to a preset proportion weight, so as to determine the mechanical execution conflict factor of each AGV.

[0072] In a preset period, the absolute value of the difference between the actual speed and the preset speed of each AGV at the same time is calculated in the preset speed time series data and the actual speed time series data of each AGV, the absolute value of the difference is used to reflect the deviation degree between the actual speed of the AGV (the actual running condition of the lower execution layer) and the preset speed (the ideal condition of the upper decision layer), the greater the value, the greater the deviation degree, and then the absolute values of the differences at all times are curve fitted to obtain a speed deviation curve, and the curve fitting here can adopt the least square method, which is a known technology and the process is not described in detail.

[0073] The result of the integral calculation of the speed deviation curve in the preset time period is normalized to obtain a speed cumulative deviation index of each AGV in the preset time period, and the discrete speed deviation is converted into a continuous cumulative amount for capturing the overall speed deviation of the AGV in the preset time period. The greater the speed cumulative deviation index, the greater the deviation of the AGV from the ideal state in the actual running process due to the influence of its own physical characteristics. The normalization is a technical means familiar to those skilled in the art, and the selection of the normalization function can be linear normalization or standard normalization, and the specific normalization method is not limited herein.

[0074] The ratio of the mean value of all current values in the motor current time series data of each AGV to the preset rated current value is normalized to obtain a load fluctuation index of each AGV in the preset time period. The mean value of the current value can be used to reflect the average load of the AGV. The greater the load fluctuation index, the greater the load pressure of the AGV, which is in a high load state and is most likely to have a load imbalance phenomenon.

[0075] Finally, the speed cumulative deviation index and the load fluctuation index of the AGV are fused through weight allocation to balance the proportion of the mechanical characteristics and the energy consumption of the AGV: the product of the speed cumulative deviation index of each AGV and the preset first weight is taken as a speed imbalance factor, the product of the load fluctuation index of each AGV and the preset second weight is taken as a load imbalance factor, and the value of the sum of the speed imbalance factor and the load imbalance factor after normalization is taken as the mechanical execution conflict factor of each AGV. As can be easily understood, the greater the mechanical execution conflict factor of the AGV, the higher the conflict between the driving planning and the mechanical execution. The sum of the preset first weight and the preset second weight is 1, and in this embodiment of the application, the preset first weight is 0.3 and the preset second weight is 0.7. The specific proportion can be adjusted according to the implementation scene, which is not limited herein. The normalization is a technical means familiar to those skilled in the art, and the selection of the normalization function can be linear normalization or standard normalization, and the specific normalization method is not limited herein.

[0076] Step S202: According to the deviation between the actual rudder angle of each AGV and the preset rudder angle, and in combination with the bending change of the path curve, a trajectory reproduction factor of each AGV is determined.

[0077] In the preset time period corresponding to the current time, the curvature value of the AGV at each time on the path curve of each AGV is calculated (which can be calculated by using the three-point geometric method according to the position coordinates of the path points on the path curve).

[0078] The AGV-based mechanical parameters (such as wheelbase, steering rudder angle limit) are used to calculate the theoretically allowed limit curvature value, according to Ackerman steering, the ratio of the wheelbase to the tangent function of the steering rudder angle limit value is the maximum steering angle of the AGV, which is referred to as the limit curvature value of the AGV;

[0079] The difference between the limit curvature and the curvature value at the position of each AGV at the current time and the ratio of the limit curvature are used as the safety margin coefficient of each AGV. The difference in the numerator can be used to represent the safety margin of the curvature value at the position of each AGV at the current time compared with the theoretical limit. The greater the value, the greater the safety margin, and therefore the greater the safety margin coefficient, the greater the curvature steering safety degree of the AGV at the current time.

[0080] The absolute value of the difference between the actual steering rudder angle of each AGV at the current time and the preset steering rudder angle is negatively correlated and normalized, and the value is used as the execution reliability coefficient. The smaller the absolute value of the difference, the closer the actual steering condition at the position of the AGV at the current time to the preset steering condition, and the smaller the execution error and the higher the safety coefficient. Therefore, the absolute value of the difference is negatively correlated and normalized to correct the logical relationship and obtain the execution reliability coefficient. As can be easily understood, the greater the value, the smaller the execution error and the smaller the steering cost. The negatively correlated and normalized processing can be performed by the formula wherein, represents an exponential function with a natural constant e as the base, and x represents the independent variable.

[0081] The product of the safety margin coefficient and the execution reliability coefficient corresponding to each AGV is calculated. Based on the foregoing analysis process, the greater the product, the greater the steering margin of the AGV. Therefore, the negatively correlated and normalized value of the product is used as the mismatch factor of each AGV at the current time. The greater the mismatch factor, the higher the comprehensive pressure of the position of the AGV at the current time on steering. The negatively correlated and normalized processing can be performed by the formula wherein, represents an exponential function with a natural constant e as the base, and x represents the independent variable.

[0082] Then, the change of the curvature value at the position of the AGV at the adjacent time can be compared to analyze whether the curvature will change suddenly, so as to analyze whether the scene of sudden bending, direct splicing and the like will occur.

[0083] The absolute value of the difference between the curvature value of the position of each AGV at the current time and the curvature value of the position at the previous time is taken as the real-time curvature change degree, and the greater the real-time curvature change degree, the greater the mutation of the curvature value, which may increase the energy consumption and mechanical wear of the AGV. Therefore, based on the real-time curvature change degree, the curvature jump penalty factor of each AGV is determined, and the curvature jump penalty factor is positively correlated with the real-time curvature change degree. In this embodiment of the application, the natural constant e is taken as the base, and the real-time curvature change degree is taken as the index, so as to construct the curvature jump penalty factor and make it grow exponentially.

[0084] Finally, the product of the normalized value of the mismatch factor corresponding to each AGV and the curvature jump penalty factor is taken as the trajectory replication factor of each AGV. Based on the foregoing analysis process, the greater the trajectory replication factor, the more factors that may interfere with the turning of the AGV in the actual execution process.

[0085] Step S203: Fusion of the mechanical execution conflict factor corresponding to each AGV and the trajectory replication factor to determine the dynamic cost function value of each AGV.

[0086] The greater the mechanical execution conflict factor of each AGV, the higher the coordination conflict between the driving planning and the mechanical execution; the greater the trajectory replication factor, the more factors that may interfere with the turning of the AGV in the turning process.

[0087] Therefore, the normalized value of the arithmetic square root of the mechanical execution conflict factor and the trajectory replication factor of each AGV is taken as the dynamic cost function value of each AGV, wherein the operation of the arithmetic square root is to amplify the influence and sensitivity of extreme values, and the normalization is a technical means familiar to those skilled in the art. The selection of the normalization function can be linear normalization or standard normalization, and the specific normalization method is not limited herein.

[0088] It should be noted that in this embodiment of the application, the preset speed and the preset steering angle of each AGV can be obtained according to the upper path planning layer, that is, under the hierarchical architecture, the upper path planning layer is generated based on the ideal environment; the limit curvature can be calculated according to the wheelbase of the AGV and the steering angle limit, representing the maximum curvature value allowed in theory.

[0089] Step S3: When obstacle avoidance is needed, the updated path and updated rudder angle of each AGV are obtained, the actual speed time sequence data, the change of point cloud data, and the cargo state data of each AGV are analyzed and fused, the brake mismatch degree of the AGV in the updated path is calculated, and the brake mismatch degree represents that the AGV generates an additional brake distance due to cargo shaking; according to the brake mismatch degree, the length of the updated path, the updated rudder angle, and the remaining time of task execution, the obstacle avoidance behavior cost of each AGV in obstacle avoidance is determined.

[0090] When encountering an obstacle to trigger emergency braking, the path planning layer only uses the ideal kinematic model to obtain the updated path and updated rudder angle of each AGV, ignores the inertia and friction time-varying induced centroid shift of the cargo carried on the trolley, the path planning layer does not obtain sensor data, and continues to generate a path according to the rigid body model, causing spatial misalignment between the theoretical trajectory and the actual pose; if the path planning layer still generates a stop or bypass instruction according to the theoretical brake distance, the theoretical brake distance will be less than the actual physical demand. In addition, single-AGV emergency avoidance causes trajectory mutation (such as right-angle bypass), which causes a sharp increase in energy consumption, forcing the scheduling system to reassign priorities, resulting in the forced extension of the paths of other AGVs; and the change in the envelope volume of the cargo carried by the AGV, even if an edge is extended, will also expand its interference range, which will also enlarge the bypass path length when other AGVs avoid obstacles.

[0091] Therefore, when the AGV needs to avoid obstacles, the updated path and updated rudder angle of each AGV can be obtained first, then the change of the point cloud data of each AGV, the cargo weight, the cargo centroid shift amount, and the cargo inclination angle are analyzed and fused, and the length of the updated path, the updated rudder angle, and the remaining time of task execution are combined to calculate the influence of local path update on the overall path when each AGV encounters a sudden event such as an obstacle or vehicle intersection, and determine the obstacle avoidance behavior cost of each AGV when avoiding obstacles.

[0092] Preferably, in an embodiment of the present application, the method for obtaining the obstacle avoidance behavior cost comprises:

[0093] Please refer to Figure 3 which shows a method flowchart of the method for obtaining the obstacle avoidance behavior cost in an embodiment of the present application, and the method comprises the following steps:

[0094] Step S301: According to the cargo centroid shift amount, the cargo weight, and the cargo inclination angle of each AGV, the brake kinetic energy of each AGV is determined, and based on the brake kinetic energy and the friction of each AGV, the additional brake distance is determined, which is used to calculate the brake mismatch degree of each AGV.

[0095] The centroid offset of the goods carried by each AGV at the current moment can be regarded as the moving distance of the goods in a unit time, so the centroid offset of each AGV at the current moment can be regarded as the goods offset speed, and then the swing kinetic energy of the goods is calculated according to the weight of each AGV and the goods offset speed. Specifically, the swing kinetic energy formula is wherein, is the swing kinetic energy; is the mass of the carried goods; is the goods offset speed.

[0096] Since the swing of the goods has a direction under the action of inertia, the component of the aforementioned calculated swing kinetic energy in the running direction of the AGV needs to be obtained as the braking kinetic energy, so the component of the swing kinetic energy of the AGV in the running direction is calculated based on the swing kinetic energy and the inclination angle of the goods carried by the AGV, thereby obtaining the braking kinetic energy of the AGV. The formula of the braking kinetic energy is wherein, represents the braking kinetic energy; represents the swing kinetic energy; represents the inclination angle of the goods.

[0097] The friction is determined based on the weight of the goods carried by the AGV. Specifically, the formula of the friction is wherein, represents the friction; represents the friction coefficient; represents the mass of the carried goods; represents the acceleration of gravity.

[0098] When the AGV needs to avoid obstacles and emergency braking occurs, the preset braking distance may be insufficient in the case of goods oscillation offset, so the ratio of the braking kinetic energy of each AGV to the friction is taken as the additional braking distance of each AGV, and the ratio of the additional braking distance to the preset braking distance is taken as the braking mismatch degree of each AGV. The greater the braking mismatch degree, the greater the degree that the stop or detour instruction generated by the path planning layer according to the theoretical braking distance will cause the theoretical braking distance to be smaller than the actual physical demand when emergency braking occurs.

[0099] It should be noted that the preset braking distance can be calculated and obtained according to the preset speed of the AGV.

[0100] Step S302: Fusion of the difference between the length of the updated path of each AGV and the preset path length, and the difference between the rudder angle and the preset rudder angle, to determine the obstacle avoidance energy consumption factor of each AGV.

[0101] When the AGV needs to avoid obstacles, trajectory mutation will occur, which will cause the energy consumption to surge, forcing the scheduling system to reassign priorities, resulting in the paths of other AGVs being forced to be lengthened.

[0102] Therefore, the ratio of the length of the updated path of each AGV to the length of the preset path is taken as the path energy consumption increment, and the greater the value, the greater the obstacle avoidance energy consumption brought by path updating.

[0103] The ratio of the updated rudder angle of each AGV to the preset rudder angle is taken as the steering energy consumption increment, and the rudder angle can be used to reflect the wear cost of the motor and tire, and the greater the steering energy consumption increment value, the greater the obstacle avoidance energy consumption brought by path updating.

[0104] Finally, the product of the path energy consumption increment and the preset first coefficient is taken as the first penalty coefficient, the product of the steering energy consumption increment and the preset second coefficient is taken as the second penalty coefficient, and the normalized value of the sum of the first penalty coefficient and the second penalty coefficient is taken as the obstacle avoidance energy consumption factor. Based on the foregoing analysis, the greater the obstacle avoidance energy consumption factor, the greater the cost consumption brought by path updating each time the obstacle is avoided.

[0105] It should be noted that in this embodiment of the application, the preset first coefficient is set to 1.5, and the preset second coefficient is set to 2. The specific values can be adjusted according to the implementation scene, which is not limited here.

[0106] Step S303: determining an obstacle avoidance cost index according to the braking mismatch degree of each AGV and the obstacle avoidance energy consumption factor.

[0107] The normalized value of the sum of the braking mismatch degree of each AGV and the obstacle avoidance energy consumption factor is taken as the obstacle avoidance cost index. It is easy to understand that the greater the obstacle avoidance cost index here, the greater the obstacle avoidance cost generated by the AGV path updating planning, AGV emergency braking, etc. when the obstacle needs to be avoided.

[0108] Step S304: analyzing the change of the point cloud data of the AGV at adjacent time, and combining the remaining time of task execution to determine the path priority weight of each AGV.

[0109] When emergency braking occurs due to the need to avoid obstacles, the goods may be offset due to the influence of inertia factors, so the envelope volume of the goods will change, and even if a corner of the goods extends, it will also expand the interference range, and other AGVs will also amplify the path length when avoiding obstacles, so the path priority weight of the AGV with a larger envelope volume change needs to be reduced to avoid affecting more AGVs.

[0110] Therefore, the overall envelope volume of the vehicle and the goods carried is obtained according to the point cloud data of each AGV at each time, and the difference between the envelope volume at the current time and the envelope volume at the previous time is calculated. The smaller the difference, the smaller the change in the envelope volume, that is, the smaller the volume of the goods extending out of the vehicle, and the smaller the impact on path planning. Therefore, the priority can be appropriately increased, so the difference is negatively correlated and mapped to obtain a priority factor. The negative correlation mapping here can use the formula wherein, represents an exponential function with a natural constant e as the base, and x represents the independent variable.

[0111] For AGVs with a large remaining time for task execution, in order to ensure that the task can be successfully executed on time, the priority needs to be increased. Therefore, the value of the negative correlation mapping of the remaining time for task execution of each AGV is used as the urgency of each AGV. The smaller the remaining time for task execution, the greater the urgency. Similarly, the negative correlation mapping here can use the formula wherein, represents an exponential function with a natural constant e as the base, and x represents the independent variable.

[0112] Finally, the proportion of the urgency of each AGV in the urgency of all AGVs is multiplied by the priority factor, and the normalized value of the product is used as the path priority weight of each AGV. At this time, the greater the path priority weight of a certain AGV, the higher the priority of the AGV. Normalization is a well-known technical means to those skilled in the art, and the selection of the normalization function can be linear normalization or standard normalization, etc. The specific normalization method is not limited here.

[0113] Step S305: Combine the path priority weight of each AGV with the obstacle avoidance cost index to obtain the obstacle avoidance behavior cost of each AGV when avoiding obstacles.

[0114] The greater the path priority weight of the AGV is, the higher the priority of the AGV is, and the greater the obstacle avoidance cost index of the AGV is, indicating that the greater the obstacle avoidance cost generated by the AGV path update planning, AGV emergency braking and the like is when obstacle avoidance is needed. Since the greater the priority weight of a certain AGV is, the smaller the influence of the AGV on path planning is, the obstacle avoidance cost of the AGV can be appropriately reduced when calculating the penalty influence degree, and therefore the product of the value of the path priority weight of each AGV after negative correlation mapping and the obstacle avoidance cost index of each AGV is normalized to obtain the obstacle avoidance behavior cost of each AGV when obstacle avoidance is needed. At this time, the greater the obstacle avoidance behavior cost is, the greater the cost generated by the AGV in the obstacle avoidance process is when obstacle avoidance is needed. Normalization is a technical means familiar to those skilled in the art, and the selection of the normalization function can be linear normalization or standard normalization, and the specific normalization method is not limited herein.

[0115] Step S4: for each possible path update result, based on the dynamic cost function value of the AGV, the obstacle avoidance behavior cost and the length of the updated path, the transportation path of all AGVs is planned.

[0116] In this step, a multi-objective optimization model can be constructed to comprehensively consider the dynamic health state (dynamic cost function value) of the AGV, the obstacle avoidance risk (obstacle avoidance behavior cost) and the path efficiency (length of the updated path) to realize intelligent planning of the global transportation path.

[0117] Preferably, in an embodiment of the present application, for each possible path update result, based on the dynamic cost function value of the AGV, the obstacle avoidance behavior cost and the length of the updated path, the transportation path of all AGVs is planned, including:

[0118] The map is refreshed every 200 ms, and when the AGV needs to avoid obstacles, the updated path of each AGV is simulated and refreshed synchronously during map updating, and then based on the numerical characteristics of the length of the updated path of each AGV, the dynamic cost function value and the obstacle avoidance behavior cost, the update cost value of each AGV is determined under each possible path update result: when obstacle avoidance is needed, the ratio of the length of the updated path of each AGV to the shortest path length under all path update results is taken as a detour index, and the sum of the detour index, the dynamic cost function value and the obstacle avoidance behavior cost of each AGV is taken as the update cost value of each AGV. The smaller the update cost value is, the smaller the detour distance of the updated path of each AGV car can be under this path update result, and the smaller the steering cost of the updated path at the current time is, and the smaller the global penalty influence is.

[0119] Therefore, the sum of the updated generation values of all AGVs is normalized as an obstacle avoidance cost index of each possible path update result, and the path update result corresponding to the minimum obstacle avoidance cost index is taken as the optimal dynamic path, so as to dynamically plan the transportation path for all AGVs. Normalization is a technical means familiar to those skilled in the art, and the selection of the normalization function can be linear normalization or standard normalization, and the specific normalization method is not limited here.

[0120] It should be noted that all the values participating in the calculation in the embodiment of the application are pre-processed, and the dimensional influence is cancelled, and the dimensional influence cancelling means is a technical means familiar to those skilled in the art, which is not limited and described here.

[0121] In summary, the embodiment of the application realizes global state perception by synchronously acquiring the speed, motor current, point cloud data, task progress and state (centroid offset, inclination angle) of the carried goods of each AGV. In the hierarchical architecture, the upper path planning generates the shortest path based on the ideal environment, while the lower execution mechanism is limited by physical characteristics such as friction, inertia, etc. in actual operation, resulting in a mismatch between path planning and mechanical execution. Therefore, the speed deviation accumulation of the AGV, the deviation of the rudder angle, the motor current characteristics and the sudden change of the path curvature are fused to quantify the real-time dynamics performance of the AGV, and the dynamic cost function value is obtained. When obstacle avoidance is needed, the centroid offset, inclination angle and point cloud data change of the carried goods of the AGV are introduced into the obstacle avoidance strategy to identify abnormal shapes such as goods protruding from the vehicle body, dynamically evaluate the actual pose of the goods, and incorporate the task execution progress and updated rudder angle of the AGV into the obstacle avoidance decision, realize the joint optimization of global path planning and individual safety, and determine the obstacle avoidance behavior cost of each AGV when avoiding obstacles. Finally, for each possible path update result, the dynamic cost function value, obstacle avoidance behavior cost and length of the updated path of the AGV are fused to plan the transportation path for all AGVs. The embodiment of the application can effectively improve the matching degree of the dynamic change of individual AGVs and the global task through the "data-driven-individual optimization-global coordination" three-level architecture, and ensures the overall transportation efficiency.

[0122] It should be noted that the above-mentioned embodiment of the application is only for description, and does not represent the advantages and disadvantages of the embodiment. The processes depicted in the drawings do not necessarily require the specific order or continuous order shown to achieve the desired results. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0123] Each embodiment in the specification is described in a progressive manner, and the same or similar parts between each embodiment can be referred to each other, and each embodiment mainly describes the differences from other embodiments.

Claims

1. A method for postal logistics warehouse-oriented cargo transportation planning, characterized in that, The method comprises: In a preset time period corresponding to the current time, the mechanical parameters, speed time series data, motor current time series data, path curve, actual steering angle, point cloud data, cargo state data at the current time, and task execution remaining time of each AGV are acquired; The deviation accumulation between the preset speed and the actual speed time series data of each AGV and the motor current time series data characteristics are fused to determine the mechanical execution conflict factor of each AGV to describe the collaborative conflict between AGV driving planning and mechanical execution; the AGV dynamic curvature margin is obtained according to the mechanical parameters of each AGV and the deviation between the actual steering angle and the preset steering angle, and the trajectory replication factor of each AGV is determined to describe the theoretical turnable margin of the AGV and the deviation degree of the actual execution capability during path turning by combining the bending change of the path curve; the mechanical execution conflict factor and the trajectory replication factor are fused to determine the dynamic cost function value of each AGV; When obstacle avoidance is needed, the updated path and the updated steering angle of each AGV are acquired, the actual speed time series data, the change of the point cloud data, and the cargo state data of each AGV are fused and analyzed, and the brake misallocation degree of the AGV in the updated path is calculated, which represents the additional brake distance caused by the cargo shaking; the obstacle avoidance behavior cost of each AGV during obstacle avoidance is determined according to the brake misallocation degree, the length of the updated path, the updated steering angle, and the task execution remaining time. For each possible path update result, the dynamic cost function value, the obstacle avoidance behavior cost, and the length of the updated path of the AGV are used to plan the transportation path for all AGVs. 2.The method of claim 1, wherein, The method for obtaining the mechanical execution conflict factor comprises: In the preset time period, in the preset speed time series data and the actual speed time series data of each AGV, the absolute value of the difference between the actual speed and the preset speed at the same time is calculated, and the absolute values at all times are curve-fitted to obtain a speed deviation curve; The result of the integral calculation of the speed deviation curve in the preset time period is normalized to obtain the speed cumulative deviation index of each AGV in the preset time period; The ratio of the mean value of all current values in the motor current time series data of each AGV to the preset rated current value is normalized to obtain the load fluctuation index of each AGV in the preset time period; The product of the speed cumulative deviation index of each AGV and a preset first weight is taken as a speed imbalance factor, the product of the load fluctuation index of each AGV and a preset second weight is taken as a load imbalance factor, and the normalized value of the sum of the speed imbalance factor and the load imbalance factor is taken as the mechanical execution conflict factor of each AGV, wherein the sum of the preset first weight and the preset second weight is 1. 3.The method of claim 1, wherein, The method for obtaining the trajectory replication factor comprises: The limit curvature value of the AGV is obtained according to the AGV mechanical parameters, which is affected by the mechanical parameters of the AGV wheelbase and the steering angle limit; In a preset time period corresponding to the current time, on the path curve of each AGV, the curvature value at the position of each AGV at each time is calculated; The difference between the limit curvature and the curvature value at the position of each AGV at the current time is negatively correlated with the limit curvature, and the ratio is taken as the safety margin coefficient of each AGV; The absolute value of the difference between the actual rudder angle of each AGV at the current time and the preset rudder angle is negatively correlated and normalized, and the value after the mapping is taken as the execution reliability coefficient of each AGV; The product of the safety margin coefficient and the execution reliability coefficient of each AGV is negatively correlated and normalized, and the value after the mapping is taken as the mismatch factor of each AGV at the current time; The absolute value of the difference between the curvature value of each AGV at the current time and the curvature value at the previous time is taken as the real-time curvature change degree, and based on the real-time curvature change degree, the curvature jump penalty factor of each AGV is determined, and the curvature jump penalty factor is positively correlated with the real-time curvature change degree; The product of the mismatch factor and the curvature jump penalty factor of each AGV is normalized, and the value after the normalization is taken as the trajectory reproduction factor of each AGV. 4.The method of claim 1, wherein, The method for obtaining the dynamic cost function value comprises: The arithmetic square root of the mechanical execution conflict factor and the trajectory reproduction factor of each AGV is normalized, and the value after the normalization is taken as the dynamic cost function value of each AGV.

5. The method of claim 1, wherein, The method for obtaining the obstacle avoidance behavior cost comprises: The cargo state data of each AGV comprises a cargo centroid offset, a cargo weight, and a cargo inclination angle; According to the cargo state data of each AGV, the braking kinetic energy of each AGV is determined, and based on the braking kinetic energy and the friction force of each AGV, the additional braking distance is determined, which is used to calculate the braking mismatch degree of each AGV; The difference between the length of the updated path of each AGV and the length of the preset path is fused, and the difference between the rudder angle and the preset rudder angle is updated, to determine the obstacle avoidance energy consumption factor of each AGV; The sum of the braking mismatch degree and the obstacle avoidance energy consumption factor of each AGV is normalized, and the value after the normalization is taken as the obstacle avoidance cost index; The change of the point cloud data of the AGV at the adjacent time is analyzed, and the task execution remaining time is combined to determine the path priority weight of each AGV; The product of the path priority weight of each AGV after negative correlation mapping and the obstacle avoidance cost index of each AGV is normalized to obtain the obstacle avoidance behavior cost of each AGV when avoiding obstacles.

6. The goods transportation planning method for postal logistics warehousing according to claim 5, characterized in that, According to the cargo state data of each AGV, the braking kinetic energy of each AGV is determined, and based on the braking kinetic energy and the friction force of each AGV, the additional braking distance is determined, which is used to calculate the braking mismatch degree of each AGV, comprising: The centroid offset of each AGV at the current time is taken as the cargo offset speed, the swing kinetic energy of the cargo is calculated according to the weight and the cargo offset speed of each AGV, and the component of the swing kinetic energy of the AGV in the running direction is calculated based on the swing kinetic energy and the inclination angle of the cargo carried by the AGV, so as to obtain the braking kinetic energy of the AGV; The weight of the cargo carried by the AGV is determined to determine the friction force. A ratio of the braking kinetic energy and the friction force of each AGV is taken as an additional braking distance of each AGV. A ratio of the additional braking distance and a preset braking distance is taken as a braking mismatch degree of each AGV.

7. The method of claim 5, wherein, The method for obtaining the obstacle avoidance energy consumption factor comprises: A product of a ratio of a length of the updated path of each AGV and a length of the preset path and a preset first coefficient is taken as a first penalty coefficient. A product of a ratio of an updated rudder angle of each AGV and a preset rudder angle and a preset second coefficient is taken as a second penalty coefficient. A normalized value of a sum of the first penalty coefficient and the second penalty coefficient is taken as the obstacle avoidance energy consumption factor. 8.The method of claim 5, wherein, The method for obtaining the path priority weight comprises: An envelope volume of each AGV and the carried goods is obtained according to the point cloud data of each AGV at each time, and a difference between the envelope volume at the current time and the envelope volume at the previous time is processed by negative correlation mapping to obtain a priority factor. A value of a negative correlation mapping of a task execution remaining time of each AGV is taken as an urgency of each AGV. A ratio of the urgency of each AGV in a sum of the urgency of all AGVs and a product of the priority factor is taken as a path priority weight of each AGV. 9.The method of claim 1, wherein, The method for planning the transportation path of all AGVs based on the dynamic cost function value of the AGV, the obstacle avoidance behavior cost and the length of the updated path comprises: When obstacle avoidance is needed, an updated cost value of each AGV is determined based on a numerical characteristic of the length of the updated path of each AGV, the dynamic cost function value and the obstacle avoidance behavior cost under each possible path update result. A normalized value of a sum of the updated cost value of all AGVs is taken as an obstacle avoidance cost index under each possible path update result. A path update result corresponding to the minimum obstacle avoidance cost index is taken as an optimal dynamic path, thereby planning the transportation path of all AGVs.

10. The method according to claim 9, wherein, The method for obtaining the updated cost value comprises: When obstacle avoidance is needed, a ratio of the length of the updated path of each AGV and the shortest path length under all path update results is taken as a detour index under each possible path update result. A sum of the detour index, the dynamic cost function value and the obstacle avoidance behavior cost of each AGV is taken as the updated cost value of each AGV.

Images