Spatial pattern-based elevator car scheduling method, device and scheduling system

Abstract

The invention relates to an elevator car scheduling method, device and system based on a spatial mode. The elevator car scheduling method based on the spatial pattern comprises the steps that target detection is conducted on an image when a car stops so as to extract object information, and the occupied capacity is estimated based on an object mobility rule; whether a historical scene is matched or not is judged in combination with the case library, if yes, a corresponding target function is directly selected, and if not, traffic flow prediction is conducted through object information of the lift car, lift car direction information and historical data, fuzzy recognition is conducted through the fuzzy rule library, and a current traffic mode is determined; then, an objective function is selected according to a use mode or a traffic mode of a historical case, optimization solution is carried out in combination with the current occupied capacity, and an optimal lift car scheduling scheme is generated. According to the elevator car scheduling method based on the spatial pattern, the problems of inaccurate scheduling and low efficiency caused by neglecting the capacity factor are effectively solved.

Description

Technical Field

[0001] The present invention relates to the technical field of elevator control, and particularly to an elevator car scheduling method, an elevator car scheduling device, and a multi-elevator intelligent scheduling system based on a spatial pattern. Background Art

[0002] Elevators are common vertical transportation tools in modern buildings and are widely used in various buildings such as commercial buildings, residences, and hospitals. With the increase in the number of building floors and the diversification of the flow of people at different times, the technology of elevator group control system (EGCS) has gradually become the key to improving the efficiency of elevator systems and reducing passenger waiting time. The traditional single-elevator scheduling method has been difficult to meet the transportation needs of high-rise buildings and diverse traffic, while the elevator group control technology optimizes the overall operation efficiency through the joint scheduling of multiple elevators, reduces the waiting time of passengers, and avoids energy waste.

[0003] The EGCS technology can achieve the coordinated scheduling of multiple elevators by combining certain scheduling algorithms according to the current position, load condition, target floor of the elevator, and the real-time needs of users. In order to optimize the passenger elevator experience, improve transportation efficiency, etc., generally, priority, prediction algorithms, or heuristic algorithms are used as scheduling algorithms to reduce energy consumption and enhance the service capacity of the elevator. Among them, the traditional shortest distance scheduling algorithm usually assigns tasks based on the distance between the elevator and the requested floor, but this method easily ignores the actual needs of passengers, such as the balance of waiting time, resulting in a low passenger elevator experience.

[0004] Based on this, the scheduling algorithms of the prior art generally combine prediction algorithms and heuristic algorithms, analyze and predict the future states of elevators and passengers through data-driven, and design objective functions, such as minimizing waiting time or shortest elevator ride time, etc., and use heuristic algorithms to optimize the scheduling path. However, the scheduling algorithms of the prior art usually do not consider the capacity and occupancy information of the elevator car, often only consider weight as the capacity constraint, and ignore the situation of passengers carrying large items, etc., resulting in an underestimation of the actual number of passengers that the elevator can accommodate, and the situation where new passengers cannot be accommodated, which not only affects the passenger elevator experience but also wastes the operating resources of the elevator.

[0005] Therefore, due to the lack of combination of the capacity and occupancy information of the elevator car in the prior art, there are problems of insufficient scheduling accuracy and low scheduling efficiency. Summary of the Invention

[0006] Based on this, the object of the present invention is to provide an elevator car scheduling method based on a spatial pattern.

[0007] An elevator car scheduling method based on a spatial pattern, comprising the following steps:

[0008] S1. Perform object detection on the image when the current car stops to obtain the object information in the current car; wherein, the object information includes object type, object quantity, object position information, and occupied size;

[0009] S2. Estimate the capacity of the object information in the current car according to the target car information table and the preset object mobility rules to obtain the occupied capacity of the current car;

[0010] S3. Judge whether there is a historical case for the object information and occupied capacity of the current car according to the case library: if so, execute step S6; if not, execute step S4;

[0011] S4. Perform traffic flow prediction on the object information, car direction information, and historical data of the current car to obtain the arrival rates of each traffic component within the future time domain;

[0012] S5. Perform fuzzy recognition on the arrival rates of each traffic component within the future time domain according to the fuzzy rule base to obtain the traffic pattern of the current car;

[0013] S6. Select a strategy according to the usage pattern or traffic pattern of the historical case to determine the objective function of the current car;

[0014] S7. Optimize and solve the corresponding objective function according to the occupied capacity of the current car to obtain the car scheduling scheme.

[0015] The elevator car scheduling method based on a spatial pattern according to the present invention, compared with the prior art, through capacity estimation, uses the preset object mobility rules to model the corresponding space release ability of the object in the elevator car, thereby improving the system's accurate judgment ability of the actual available space in the car, avoiding resource waste and task allocation imbalance.

[0016] Meanwhile, a historical case mechanism is introduced to improve the efficiency of quickly reusing strategies for typical scenarios; when it is a new scenario, the traffic pattern of the current car is identified through traffic flow prediction and fuzzy recognition, and thus, based on the corresponding traffic pattern or the usage pattern of the historical case, the corresponding objective function is selected for optimization to enhance the dynamic adaptability of the scheduling strategy.

[0017] Accordingly, through the preset object mobility rules, case library, traffic pattern recognition, and objective optimization, the space adaptability, strategy accuracy, and scheduling efficiency of elevator scheduling are significantly improved, and the problems of inaccurate scheduling and low efficiency caused by ignoring the capacity factor are effectively solved.

[0018] Further, the capacity estimation is used to quantify the ability of the current car to make room for newly added passengers, and the calculation method of the occupied capacity is as follows:

[0019]

[0020] In the formula, C represents the occupied capacity of the current car; s {cab} represents the area of the empty car; among them, the target car information table includes the standardized dimensions of the elevator car and the mapping relationship between the occupied dimensions of the object and the actual dimensions;

[0021] S i represents the occupied area of the i-th object, which is obtained by mapping the occupied dimensions of the i-th object through the target car information table; classes represents the total number of objects in the current car; L is the space utilization factor, and its specific calculation is as follows:

[0022]

[0023] In the formula, P i is the mobility level of the i-th object.

[0024] Accordingly, by introducing a capacity estimation mechanism with the space utilization factor L as the core, combined with the dimension mapping relationship between the object mobility level P i and the target car information table, the dynamic modeling and fine evaluation of the available space in the elevator car are realized, significantly improving the judgment accuracy and feasibility analysis ability of the dispatching system for the remaining space, and effectively avoiding the task allocation errors and resource waste problems caused by the occupancy of low-mobility objects.

[0025] Further, the traffic components include the upward traffic arrival rate, the downward traffic arrival rate, and the inter-floor traffic arrival rate, and their specific calculations are as follows:

[0026]

[0027] In the formula, λ in is the upward traffic arrival rate, which is used to represent the proportion of the number of people rising from the main floor among the total passengers; λ out is the downward traffic arrival rate, which is used to represent the proportion of the number of people descending to the main floor among the total passengers; λ inc is the inter-floor traffic arrival rate, which is used to represent the proportion of the number of people moving between internal floors among the total number of people; up represents the number of passengers who start from the main floor, enter the car and move upward; down represents the number of passengers in the car descending to the main floor; inter represents the number of passengers moving between non-main floors; all represents the total number of passengers carried in the current car;

[0028] Among them, the traffic flow prediction is modeled and predicted based on the arrival rates of each traffic component of the current car and the arrival rates of each traffic component of historical data, and the arrival rates of each traffic component within the future time domain are obtained. Its specific calculation is expressed as follows:

[0029] F t+1 = α t Y t +(1 - α t )F t

[0030] In the formula, F t represents the predicted arrival rates of each traffic component at the t-th moment, including the upward traffic arrival rate, the downward traffic arrival rate, and the inter-floor traffic arrival rate; Y t represents the actual arrival rates of each traffic component at the t-th moment, including the upward traffic arrival rate, the downward traffic arrival rate, and the inter-floor traffic arrival rate; α t represents the adaptive smoothing factor at the t-th moment, and its specific calculation is expressed as follows:

[0031]

[0032] In the formula, E t represents the smoothing error at the t-th moment, and its specific calculation is expressed as follows:

[0033] E t = βe t +(1 - β)E t-1

[0034] In the formula, β represents the smoothing factor; e t is the current prediction error, and its specific calculation is expressed as follows:

[0035] e t = Y t - F t

[0036] AE t represents the absolute smoothing error at the t-th moment, and its specific calculation is expressed as follows:

[0037] AE t = β|e t |+(1 - β)AE t-1

[0038] In the formula, |e t | represents the absolute value of the current prediction error.

[0039] Accordingly, by introducing a traffic flow prediction method based on an adaptive response rate exponential smoothing mechanism, the weight ratio between the historical trend and the current state is dynamically adjusted, enabling the system to respond in real time to sudden changes and regular variations in traffic conditions, avoiding the failure of traditional fixed-parameter prediction during traffic fluctuation periods (such as morning and evening rush hours), and significantly improving the prediction accuracy of future traffic conditions and the foresight of scheduling strategies.

[0040] Further, the fuzzy recognition includes the following steps:

[0041] S51A. Fuzzify the arrival rates of each traffic component within the future time domain using the traffic component membership function to obtain the membership degrees of the upward traffic component, downward traffic component, and interlayer traffic component;

[0042] S51B. Fuzzify the ratio of the total arrival rate to the processing capacity using the traffic intensity membership to obtain the membership degree of the traffic intensity;

[0043] S52. According to the fuzzy rule base, defuzzify the membership degrees of the upward traffic component, downward traffic component, interlayer traffic component, and the membership degree of the traffic intensity to obtain the traffic mode of the current car.

[0044] Accordingly, the present invention identifies the traffic mode through fuzzy rules and selects the scheduling optimization objective function accordingly, significantly improving the strategy flexibility and adaptability of the system in variable scenarios and effectively avoiding the problem of the failure of static scheduling strategies under peak, congested, or skewed traffic conditions.

[0045] Further, the process of the optimization solution includes the following steps:

[0046] S71. Initialize the search tree and its root node; wherein, the root node is the starting node of the search tree, and its state includes the scheduling information of all elevator cars at the current moment, including the position of the current car, the car direction information, the assigned task allocation plan, the occupied capacity of the current car, the occupancy factor, the currently selected objective function, and the set of hall call requests to be processed in the current round;

[0047] Wherein, the calculation method of the occupancy factor is:

[0048]

[0049] In the formula, O n represents the occupancy factor of the nth node; C n represents the occupied capacity of the car corresponding to the nth node; represents the total area of the car of the nth node;

[0050] S72. Update and expand the root node or the parent node of the current search tree to generate several child nodes;

[0051] S73. Calculate the cost and perform priority sorting for all child nodes of the search tree to obtain the total cost;

[0052] Among them, the cost calculation is to multiply the occupancy factor by the objective function and add it to the estimated cost function to obtain the total cost of the current node. The specific calculation is as follows:

[0053] f(n) = O n ×g(n) + h(n)

[0054] In the formula, f(n) represents the total cost of the current nth child node; g(n) is the objective function filtered by the current nth node through the usage pattern or traffic pattern; h(n) is the estimated cost function of the nth node;

[0055] S74. Determine whether the current search tree meets the termination condition: If so, terminate the search process and return the task allocation plan of the node with the minimum current total cost to obtain the optimal car scheduling plan; if not, take the child node as the parent node according to the order of the total cost and execute step S72.

[0056] Accordingly, the present invention introduces the occupancy factor obtained by calculating the occupancy capacity and uses it in the total cost calculation in the scheduling optimization, enabling the scheduling algorithm to dynamically adapt to the car space state, reducing the empty running rate and scheduling conflicts of the car scheduling plan, thereby improving the operation efficiency and service quality of the overall scheduling system.

[0057] An elevator car scheduling device based on a spatial pattern, comprising a target detection unit, a capacity estimation unit, a historical case judgment unit, a traffic flow prediction unit, a traffic pattern recognition unit, an objective function selection unit, and a car scheduling plan optimization unit;

[0058] The target detection unit is used to perform target detection on the image when the current car stops to obtain the object information in the current car; among them, the object information includes the object type, the number of objects, the object position information, and the occupied size;

[0059] The capacity estimation unit is used to estimate the capacity of the object information in the current car according to the target car information table and the preset object mobility rules to obtain the occupancy capacity of the current car;

[0060] The historical case judgment unit is used to judge whether there is a historical case for the object information and occupancy capacity of the current car according to the case library: If so, call the objective function selection unit; if not, call the traffic flow prediction unit;

[0061] The traffic flow prediction unit is used to perform traffic flow prediction based on the object information, car direction information of the current car and historical data, and obtain the arrival rates of various traffic components within a future time domain;

[0062] The traffic mode recognition unit is used to perform fuzzy recognition on the arrival rates of various traffic components within a future time domain according to the fuzzy rule base, and obtain the traffic mode of the current car;

[0063] The objective function selection unit is used to select a strategy according to the usage mode or traffic mode of historical cases, and determine the objective function of the current car;

[0064] The car scheduling scheme optimization unit is used to optimize and solve the corresponding objective function according to the occupancy capacity of the current car, and obtain a car scheduling scheme.

[0065] A multi - elevator intelligent scheduling system, characterized by comprising an image acquisition device installed in each elevator car, an elevator total control unit, and the above - mentioned elevator car scheduling device based on a spatial pattern;

[0066] The image acquisition device is used to perform real - time image acquisition inside the parked elevator car, and transmit the acquired image when the car is parked to the elevator car scheduling device through wire or wirelessly;

[0067] The elevator total control unit is used to receive hall - floor call requests, the position and car direction information of the elevator car, and transmit them to the elevator car scheduling device through wire or wirelessly; meanwhile, schedule all elevator cars according to the car scheduling scheme generated by the elevator car scheduling device.

[0068] For better understanding and implementation, the present invention will be described in detail below with reference to the accompanying drawings. Brief Description of the Drawings

[0069] Figure 1 It is a simple structural schematic diagram of the elevator car scheduling device based on a spatial pattern;

[0070] Figure 2 It is a simple flow schematic diagram of the elevator car scheduling method based on a spatial pattern;

[0071] Figure 3 It is a simple example schematic diagram of the occupancy capacity of some objects;

[0072] Figure 4 It is a simple example schematic diagram of the static density when a crowd occupies;

[0073] Figure 5 It is a simple schematic diagram of the membership function of traffic components;

[0074] Figure 6 It is a simple schematic diagram of the membership function of traffic intensity;

[0075] Figure 7 It is a simple schematic diagram of the rules stored in the fuzzy rule base. Detailed implementation manners

[0076] To solve the problems of insufficient scheduling accuracy and low scheduling efficiency in the prior art during elevator scheduling, the present invention performs object detection on the real-time input car images to obtain object information in the car; then, according to the target car information table and the preset object mobility rules, capacity estimation is performed on the object information in the car to obtain the occupancy capacity of the car; and it is judged whether there are historical cases for the object information and occupancy capacity of the car according to the case base: if so, the objective function corresponding to the usage pattern of the historical case is optimized and solved according to the occupancy capacity of the car to obtain a car scheduling plan; if not, traffic flow prediction is performed according to the occupancy capacity of the car, the car direction information and historical data to obtain the arrival rate of each traffic component in the future time domain; and fuzzy recognition is performed on the arrival rate of each traffic component in the future time domain according to the fuzzy rule base to obtain the traffic pattern of the car; then, the objective function corresponding to the traffic pattern is optimized and solved according to the occupancy capacity of the car to obtain a car scheduling plan.

[0077] Accordingly, the present invention can effectively improve the response speed of the system and thus enhance the scheduling efficiency by preferentially identifying the usage pattern of historical cases through the case base and selecting the corresponding objective function for optimization. At the same time, traffic pattern recognition is performed on the occupancy information and traffic component arrival rate through fuzzy rules, which can identify traffic flow and emergencies in real time and dynamically, so as to select the most accurate objective function for optimization and improve the scheduling accuracy.

[0078] In addition, the present invention can effectively solve the capacity constraint problem by adopting a dynamic weight for designing the objective function corresponding to the occupancy capacity, and optimize the scheduling decision in real time according to the actual conditions of the elevator, improve the flexibility of scheduling, especially in high-load periods or special demand scenarios, and can perform more accurate scheduling.

[0079] Based on the above design, the present invention proposes an elevator car scheduling method based on a spatial pattern, and proposes an elevator car scheduling device based on the spatial pattern.

[0080] A multi-elevator intelligent scheduling system includes an image acquisition device installed in each elevator car, an elevator master control unit, and the elevator car scheduling device based on the spatial pattern;

[0081] The image acquisition device is used to perform real-time image acquisition inside the parked elevator car and transmit the acquired image of the car during parking to the elevator car scheduling device through wired or wireless means.

[0082] The elevator total control unit is used to receive hall call requests, the position of the elevator car, and car direction information, and transmit them to the elevator car scheduling device through wired or wireless means; at the same time, according to the car scheduling plan generated by the elevator car scheduling device, all elevator cars are scheduled.

[0083] Among them, the hall call request is triggered by a call panel button outside the elevator, indicating that a user on a certain floor requests to take the elevator up or down; the car position information is obtained through an encoder, a displacement sensor, a door area detection module in the elevator's own control system or directly by the EGCS, and is used to identify the floor where the car is located in real time; the car direction information is provided in real time by the elevator's own control system or by the EGCS, and is used to represent the current running trend state of the car, usually including three states: "upward", "downward", and "stationary".

[0084] Please also view Figure 1 and Figure 2 , Figure 1 which is a schematic diagram of the simple structure of the elevator car scheduling device based on the spatial pattern, Figure 2 and which is a schematic diagram of the simple process of the elevator car scheduling method based on the spatial pattern.

[0085] The elevator car scheduling device based on the spatial pattern includes a target detection unit 1, a capacity estimation unit 2, a historical case judgment unit 3, a traffic flow prediction unit 4, a traffic pattern recognition unit 5, a target function selection unit 6, and a car scheduling plan optimization unit 7.

[0086] The target detection unit 1 is used to execute step S1: perform target detection on the image when the current car stops to obtain the object information in the current car.

[0087] Specifically, the target detection is performed using a pre-trained Mask R-CNN model, and the pre-trained Mask R-CNN model includes a backbone network, a feature pyramid, a region of interest pooling layer, and a prediction module.

[0088] The backbone network, using ResNet or VGG, is used to perform multi-scale feature extraction on the current car image to obtain feature maps of different scales.

[0089] It should be noted that since the selection of the backbone network will be adjusted according to the actual application scenario or requirements, the specific selection of the backbone network is not limited in the present invention. For example, ResNet-18, ResNet-50, VIT, etc. can all be candidates, and the present invention defaults to selecting ResNet-50 as the backbone network.

[0090] The Feature Pyramid Network (FPN) is used to perform feature fusion on feature maps of different scales to obtain fused feature maps of different scales, thereby improving the detection ability for objects of different sizes.

[0091] The Region of Interest Pooling layer is used to generate candidate regions (Regions of Interests, ROIs) for the fused feature maps of different scales, and perform region pooling (ROIAlign) on the generated candidate regions to obtain the pooled feature maps.

[0092] Among them, the generation of candidate regions is performed by the Region Proposal Network (RPN), and the RPN is used to perform sliding window operations on the input fused feature maps of different scales.

[0093] The prediction module includes a classification branch and a mask prediction branch. The classification branch is used to classify objects in the pooled feature maps to determine the object types and the number of objects of the object information in the current car.

[0094] The mask prediction branch is used to perform regression prediction on the pooled feature maps to generate the segmentation masks of each object instance, and determine the object position information and the occupied size of the object information in the current car.

[0095] Among them, in order to improve the detection accuracy of the pre-trained Mask R-CNN model for objects in the car, the present invention fine-tunes the Mask R-CNN model in the following manner, which includes the following steps:

[0096] T1: Collect the data set.

[0097] Among them, the data set includes the COCO data set of Microsoft, which specifically contains 320,000 labeled images and more than 90 category labels. At the same time, for common items in the elevator, such as backpacks, bicycles, cats, chairs, dogs, handbags, people, snowboards, balls, suitcases, surfboards, tennis rackets or umbrellas, etc., additional annotations are made and new categories are added. Different data can be added according to the specific scenario. For example, categories specific to the elevator (such as wheelchairs, ambulance beds, etc.) can be added. The present invention does not specifically limit the object types and labels of the data set, but only requires it to contain common elevator items.

[0098] T2: Divide the data set in a ratio of 8:2 to obtain a training set and a test set.

[0099] Among them, the training set is used to train the model, and the test set is used to evaluate the model performance. The ratio of the training set and the test set can be adjusted according to actual needs, or the participation of a validation set can be introduced. The present invention does not specifically limit this.

[0100] T3: Freeze the shallow - layer convolutions of the backbone network of the pre - trained Mask R - CNN model to obtain a frozen Mask R - CNN model.

[0101] It should be noted that freezing the shallow - layer convolutions of the backbone network is to utilize the advantages of transfer learning and avoid retraining the extraction of basic features. At the same time, the shallow - layer convolutions mainly learn the extraction of general features such as edges or textures, which are actually common for different visual tasks and thus do not require retraining. Accordingly, by freezing the shallow - layer convolutions of the backbone network, the training time can be effectively reduced, the convergence can be accelerated, and the need for a large amount of data can be reduced.

[0102] Meanwhile, according to the different specific application scenarios, users can choose to freeze the entire backbone network and only train the subsequent feature pyramid and prediction modules. Therefore, the present invention does not limit the specific selection of the frozen module.

[0103] T4: Input the images in the training set into the frozen Mask R - CNN model for forward propagation to obtain recognition results.

[0104] T5: Use a loss function to calculate the loss between the recognition results and the type labels corresponding to the annotated images to obtain a loss value.

[0105] It should be noted that the loss function can be an optional cross - entropy loss function or a CIOU loss function, etc., and the choice of the loss function should be determined according to the specific task. For example, cross - entropy is used for multi - class classification, and the CIOU loss function is used to optimize the regression accuracy of bounding boxes. The loss function or its weight can be adjusted according to the actual situation. The present invention does not specifically limit the choice of the loss function.

[0106] T6: Calculate the gradient corresponding to the loss value through backpropagation, and according to the gradient, use an optimizer to update the parameters of the frozen Mask R - CNN model to obtain an updated Mask R - CNN model.

[0107] Among them, the optimizer can be Adam or SGD, etc., and the optimizer will adjust the update step of the parameters according to the preset learning rate to avoid unstable training caused by over - updating.

[0108] T7: Input the images in the test set into the updated Mask R - CNN model for forward propagation and calculate the corresponding performance metrics.

[0109] Among them, the performance metrics include accuracy, recall, F1 - Score, etc., to evaluate the performance of the model on unseen data. It can also be combined with mAP (mean Average Precision) as the main metric for instance detection tasks. The present invention does not specifically limit the choice of metrics here.

[0110] T8: Determine whether the performance metrics of the updated Mask R-CNN model meet a performance threshold. If so, complete the training and obtain a Mask R-CNN model with high accuracy. If not, return to step T4 for execution.

[0111] Among them, the performance threshold is used to determine whether the model meets the performance threshold. Common judgment criteria include whether the accuracy rate is higher than 90%, or whether the loss function is lower than a preset threshold, etc., to ensure that the model can be continuously improved during the training process. Therefore, the present invention does not specifically limit the selection of the performance threshold.

[0112] The capacity estimation unit 2 is used to execute step S2: According to the target car information table and the preset object mobility rules, estimate the capacity of the objects in the current car to obtain the occupied capacity of the current car.

[0113] Specifically, the capacity estimation is used to quantify the ability of the current car to make room for newly added passengers, that is, to quantify the space release ability of the car, so as to allocate space for new passengers in actual scheduling, so that the system can optimize the use efficiency of the elevator and ensure that the remaining space is maximally utilized under the condition of existing object occupancy. The specific calculation of its occupied capacity is shown as follows:

[0114]

[0115] In the formula, C represents the occupied capacity of the current car, that is, the remaining available space capacity of the current car, and its unit is square meters; S {cab} represents the area of the empty car, which is obtained from the target car information table and is a known constant, generally defaulting to 4 square meters. Among them, the target car information table includes the standardized size of the elevator car, that is, the area of the empty car, and the mapping relationship between the object occupancy size and the actual size, that is, the conversion between pixels and sizes, so as to provide an accurate occupied area for various types of objects for mapping the detected objects;

[0116] S i represents the occupied area of the i-th object, which is obtained by mapping the occupancy size of the i-th object through the target car information table; classes represents the total number of objects in the current car, that is, the number of objects detected by the target; L is the space utilization factor, which is used to measure the space utilization rate of the current car, so as to reflect the comprehensive influence of each object's occupied space, and can be used to dynamically adjust the effective availability of the remaining area, so as to reflect the influence of object mobility on the space release ability. The specific calculation of its space utilization factor is shown as follows:

[0117]

[0118] wherein, P i is the mobility level of the i-th object, which is used to reflect the mobility of the i-th object. Generally, the mobility level of a high-mobility object is smaller, while that of a low-mobility object is larger. Its default values are 1 (high), 2 (medium), and 3 (low). For details, please refer to Figure 3 , Figure 3 is a schematic diagram of a simple example of the occupancy capacity of some objects, Figure 4 is a schematic diagram of a simple example of the static density during occupancy of a crowd.

[0119] It should be noted that since it is difficult to directly model the relationship between the object position and the capacity, and the camera angle will also affect the positioning accuracy, the present invention introduces the mobility level to reflect the influence of the object position, thereby compensating for the positioning error caused by the camera angle and enabling the system to perform effective scheduling optimization based on the actual space utilization situation. Therefore, the mobility level is essentially an indirect modeling of the object blocking degree and the possibility of movement, playing the role of a weak cost function.

[0120] Accordingly, by modeling the space utilization factor, the occupied area and the mobility level of each object are comprehensively considered to more accurately measure the space usage inside the elevator, ensuring that the remaining car space is maximally utilized under the condition that there are already occupied objects.

[0121] The historical case judgment unit 3 is used to execute step S3: judging whether there is a historical case for the object information and occupancy capacity of the current car according to the case library. If so, the target function selection unit 6 is called; if not, the traffic flow prediction unit 4 is called.

[0122] Specifically, the case library stores multiple historical cases, and each historical case includes the usage mode and the corresponding solution when the elevator is operating. For any historical case, it can be specifically expressed as follows:

[0123]

[0124] wherein, c i represents the characteristic parameters of the usage mode, including boolean variables, positive integers, and areas. The boolean variable is used to indicate whether there are special objects in the current car, such as stretchers and cleaning carts, etc.; the positive integer is used to describe the number of objects in the current car, such as the total number of passengers, the total number of carts, etc.; the area is used to represent the occupancy capacity; n represents the total number of characteristic parameters, and its default value is 4. The corresponding characteristic parameters can be set according to the actual situation. For example, in a medical care scenario, the number of stretchers, wheelchairs, and personnel configurations, etc. can be added as characteristic parameters; ID represents the unique identifier of the case, which is used for quick indexing and efficient retrieval of historical cases in the case library; m represents the usage mode of the current historical case and corresponds to the solution one by one.

[0125] Next, by using the nearest neighbor matching method, calculate the similarity between the object information and occupancy capacity of the current car and the characteristic parameters of historical cases, and determine whether the similarity result meets a case threshold: if so, it is considered that there is a historical case for the current car state, and the target function selection unit 6 needs to be called; if not, it is considered that there is no historical case for the current car state, and the traffic flow prediction unit 4 needs to be called.

[0126] Among them, the calculation of the similarity is expressed as follows:

[0127]

[0128] In the formula, Similar represents the similarity, w i represents the weight factor corresponding to the i-th characteristic parameter, which is a preset value. Different values can be set according to different characteristic parameters. The present invention does not specifically limit the preset method of its weight factor. For example, if the characteristic parameter of a Boolean variable is 1, indicating that there is a stretcher in the current car, the weight is 1.0, thus indicating that the current car is in the emergency use mode and no new task needs to be assigned, and the corresponding solution can be directly matched; represents the i-th characteristic parameter of the current car and the i-th characteristic parameter of a certain historical case in the case library The degree of similarity can be calculated by means of normalization based on the matching of Boolean variables, the absolute value difference of quantities, and the proportion of areas.

[0129] The matching of the case threshold is expressed as follows:

[0130]

[0131] In the formula, th sim is the case threshold, and its specific default value is 0.8.

[0132] It should be noted that traditional elevator dispatching strategies often make decisions only relying on static rules or real-time perception data, and it is difficult to make a quick response to sudden or special scenarios (such as priority for hospital stretchers, occupancy by logistics transportation).

[0133] Therefore, the present invention introduces a "case reasoning mechanism" aiming to quickly make a more adaptable dispatching response through the experience transfer of known scenarios and corresponding dispatching rules, avoid repeated analysis and calculation, and improve the response speed and strategy accuracy of the system.

[0134] In addition, the case structure design is compatible with Boolean quantities, numerical values, and area-type indicators, which helps to cover various car state scenarios and has good scalability.

[0135] Accordingly, through the above judgment mechanism, when the system identifies the current state similar to the historical situation, it can skip the complex traffic prediction modeling process and directly call the corresponding target function configuration to achieve fast response and high adaptability, thereby constructing a scheduling architecture with both stability and scalability.

[0136] The traffic flow prediction unit 4 is used to execute step S4: perform traffic flow prediction based on the object information of the current car, the car direction information, and historical data to obtain the arrival rates of various traffic components within the future time domain.

[0137] Specifically, according to the current car direction information and object information, calculate the arrival rates of the various traffic components of the current car. The specific calculation is as follows:

[0138]

[0139]

[0140] In the formula, λ in 、λ out and λ inc are the upward traffic arrival rate, the downward traffic arrival rate, and the inter-floor traffic arrival rate respectively. The upward traffic arrival rate represents the proportion of the number of people ascending from the main floor to the total passengers; the downward traffic arrival rate represents the proportion of the number of people descending to the main floor to the total passengers; the inter-floor traffic arrival rate represents the proportion of the number of people moving between internal floors to the total number of people.

[0141] up represents the number of passengers who depart from the main floor, enter the car, and move upward. The estimation method is: after the car stops at the main floor, "the difference between the minimum number of passengers during docking and the number of passengers after docking";

[0142] down represents the number of passengers in the car descending to the main floor. The estimation method is by taking "the difference between the number of passengers before docking and the number of passengers after docking";

[0143] inter represents the number of passengers moving between non-main floors. The specific judgment is: when the elevator is running upward and the current stop floor is not the main floor, if there are passengers boarding the elevator, this number is recorded as inter; conversely, when the elevator is running downward and the current stop floor is not the main floor, if there are passengers getting off the elevator, this number is recorded as inter;

[0144] all represents the total number of passengers carried in the current car, which is used as the basis for normalization calculation.

[0145] Among them, the main floor represents the main floor that is common as the starting or ending point daily, and is generally defaulted to the 1st floor or the lobby; the number of passengers before docking, during docking, and after docking is obtained from the object information obtained by target detection, and combined with the car direction information to infer the behavior of passengers getting on or off the elevator; the car direction information represents the state parameter of the current running trend of the car, which is used to distinguish the elevator running state as upward, downward, or stationary, and is generally provided in real time by the elevator control system.

[0146] Next, based on the arrival rates of each traffic component of the current car and the arrival rates of each traffic component of the historical data, a model prediction is made to obtain the arrival rates of each traffic component within the future time domain, and its specific calculation is as follows:

[0147] F t+1 = α t Y t +(1 - α t )F t

[0148] In the formula, F t represents the predicted arrival rates of each traffic component at the t-th moment, including the upward traffic arrival rate, the downward traffic arrival rate, and the inter-floor traffic arrival rate; Y t represents the actual arrival rates of each traffic component at the t-th moment, including the upward traffic arrival rate, the downward traffic arrival rate, and the inter-floor traffic arrival rate, which is obtained from the historical data. The historical data is the measured record of the upward, downward, and inter-floor traffic arrival rates calculated based on the change of object information after each stop during the historical operation of the elevator car;

[0149] α t represents the adaptive smoothing factor at the t-th moment, which is used to dynamically adjust the weight relationship between the historical data and the predicted traffic arrival rates at present. Its specific calculation is as follows:

[0150]

[0151] In the formula, E t represents the smoothing error at the t-th moment, which is used to capture the trend direction. The specific calculation is as follows:

[0152] E t = βe t +(1 - β)E t-1

[0153] In the formula, β represents the smoothing factor, which is used to dynamically adjust the weight factor between the current prediction error e t and the historical smoothing error, and is defaulted to 0.3; e t is the current prediction error, and its specific calculation is as follows:

[0154]

[0155] AE t Represents the absolute smoothing error at the t-th moment, used to estimate the magnitude of the error, and its specific calculation is as follows:

[0156] AE t =β|e t |+(1-β)AE t-1

[0157] In the formula, |e t | represents the absolute value of the current prediction error.

[0158] It should be noted that traffic flow shows obvious characteristics of time period, direction and volatility in elevator dispatching. Traditional static models or linear prediction methods are difficult to effectively handle situations such as mode switching and sudden increase in traffic flow. The present invention adopts Adaptive Response Rate Exponential Smoothing (ARRES) to ensure real-time tracking of trends of different traffic components while maintaining computational efficiency.

[0159] Accordingly, through the above ARRES algorithm, it is ensured that when historical cases are not hit, the future traffic flow is quantitatively estimated based on the current state trend and historical data, thereby improving the forward-looking and robustness of the dispatching strategy.

[0160] The traffic mode recognition unit 5 is used to execute step S5: according to the fuzzy rule base, perform fuzzy recognition on the arrival rates of various traffic components in the future time domain to obtain the traffic mode of the current car.

[0161] Specifically, the traffic component membership function is used to fuzzify the arrival rates of various traffic components in the future time domain to obtain the membership degrees of the upward traffic component, downward traffic component and inter-floor traffic component.

[0162] Among them, the traffic component membership function is an optional trapezoidal membership function, or combined with a triangular membership function, and corresponding intervals are set. For details, refer to Figure 5 , Figure 5 is a simple schematic diagram of the traffic component membership function. For example: if the upward traffic arrival rate is 0.6, and the membership function of the "medium" traffic is set as a triangular function with its domain being [0.4, 0.7], and the domain of the membership function of the "high" traffic is [0.6, 0.9], then:

[0163] μ 低 (0.6)=0, μ 中 (0.6)=0.6, μ 高 (0.6)=0.3

[0164] That is, the input value 0.6 will partially belong to both "medium" and "high" simultaneously, with membership degrees of 0.6 and 0.3 respectively. Through this cross - definition, the recognition flexibility of the system for adjacent states is improved;

[0165] The so - called fuzzification is the first stage of fuzzy logic, which means assigning each component to the meaning of multiple linguistic values (such as "low", "medium", "high"), that is, the process of converting precise numerical values into fuzzy values.

[0166] Meanwhile, the ratio of the total arrival rate to the processing capacity is fuzzified using the membership degree of traffic intensity to obtain the membership degree of traffic intensity.

[0167] Among them, the membership degree function of traffic intensity can be a trapezoidal membership degree function. For details, please refer to Figure 6 , Figure 6 which is a simple schematic diagram of the membership degree function of traffic intensity; the ratio of the total arrival rate to the processing capacity is used to represent the load level of the current car, and its specific representation is as follows:

[0168]

[0169] In the formula, rate represents the ratio of the total arrival rate to the processing capacity; P max is the maximum processing capacity of the car, which can be selected as the maximum number of passengers the car can carry (such as 16 people), or the estimated transportation capacity per unit time (such as 20 people per minute), etc.

[0170] Finally, according to the fuzzy rule base, the membership degrees of the upward traffic component, downward traffic component, inter - floor traffic component, and the membership degree of traffic intensity are defuzzified to obtain the traffic mode of the current car.

[0171] Among them, the fuzzy rule base can be manually set by expert experience or extracted by induction from historical traffic data to form several fuzzy rules that combine knowledge - driven and data - driven methods;

[0172] The input of the fuzzy rule is the membership degree of traffic components and the membership degree of traffic intensity, and its output is the corresponding traffic mode. For example: if the upward traffic arrival rate is "high", the downward traffic arrival rate is "low", and the traffic intensity is "extremely heavy", then the traffic mode is "extremely heavy morning rush hour". For details, please refer to Figure 7 , Figure 7 which is a simple schematic diagram of the rules stored in the fuzzy rule base.

[0173] The defuzzification is used to represent the conversion of fuzzy values into specific and executable traffic patterns, which can be defuzzified by the weighted average method or the maximum membership degree method. The weighted average method is applicable to the case of continuous output results, and the maximum membership degree method is applicable to decision-making scenarios with clear categories, such as "morning rush hour" and "evening rush hour" in traffic pattern recognition.

[0174] Accordingly, the present invention uses the fuzzy logic method, takes the arrival rate and traffic intensity of traffic components as input variables, combines the preset membership function and fuzzy rule base, identifies the traffic pattern of the current car, realizes the intelligent classification of various typical traffic states such as upward peak, downward peak, and inter-floor flow, and can handle the fuzzy transition and complex mixing situations between traffic states, providing a reliable basis for the accurate matching of subsequent scheduling strategies.

[0175] The objective function selection unit 6 is used to execute step S6: select a strategy according to the usage pattern or traffic pattern of historical cases to determine the objective function of the current car.

[0176] Specifically, the objective function is used to measure the resource consumption or operation cost during the elevator's execution of the scheduling task. For example, the time cost, energy consumption cost, or passenger congestion degree generated during the upward or downward transportation of passengers. Therefore, by minimizing the objective function, the optimization goal of reducing the overall scheduling cost can be achieved.

[0177] Among them, according to the currently recognized usage pattern or traffic pattern, the weights of each cost item in the objective function are adjusted to reflect the change of the current scheduling priority, so as to dynamically adapt to different operation requirements. The example representation of the objective function is as follows:

[0178] Target=w1×WaitTime+w2×Energy+w3×Comfort

[0179] In the formula, Target is the objective function; WaitTime represents the passenger waiting time; Energy represents the operation energy consumption; comfort represents the passenger congestion degree; w1, w2, and w3 are the weight factors corresponding to the cost items, which can be dynamically adjusted according to the recognized traffic pattern or usage pattern.

[0180] It should be noted that the objective function described in this article is only an example, and its structure and cost item configuration are not fixed templates. In actual applications, the objective function can be flexibly customized according to the traffic pattern or usage pattern defined by the specific usage scenario (such as hospitals, shopping malls, residential buildings, office buildings, etc.).

[0181] On the one hand, the scheduling requirements vary in different scenarios, and the system can adjust the weight configuration of each cost item in the objective function according to the identified pattern. On the other hand, the cost items of the objective function can also be added or deleted. For example, in some special modes, customized metrics such as "response time limit" and "service priority" can be added, or irrelevant metrics can be deleted.

[0182] For example, when the traffic mode is "upward peak", priority can be given to reducing waiting time, increasing the weight of w1, and at the same time reducing or ignoring the impact of passenger crowding degree w3×Comfort. In some medical scenarios, "first aid response level" may even be introduced as a new task cost item in the objective function.

[0183] Therefore, the objective function should be configured according to the actual scenario, either by only adjusting the weights or by dynamically defining or simplifying the cost structure according to the task requirements, so as to ensure that the scheduling logic best fits the application requirements.

[0184] The car scheduling scheme optimization unit 7 is used to execute step S7: optimize and solve the corresponding objective function according to the occupancy capacity of the current car to obtain the car scheduling scheme.

[0185] Generally speaking, heuristic algorithms are often used for task allocation and path selection in traditional elevator scheduling optimization, such as the rule-based nearest elevator strategy, the nearest request first method based on the greedy principle, or the heuristic search algorithm based on the cost function. Among them, the A*(A-star) search algorithm is a graph search algorithm based on the heuristic function, which has the advantage of balancing path cost estimation and search efficiency. It constructs a search tree and calculates the total cost function of each state node:

[0186] f(node) = g(node) + h(node)

[0187] In the formula, g(node) represents the actual cost from the starting state to the current state, that is, the objective function, and h(node) represents the estimated cost function from the current state to the target state. Thus, during the search process, paths with smaller estimated total costs are preferentially expanded to achieve a fast approximate optimal solution. However, in the elevator scheduling problem, traditional heuristic algorithms often ignore the actual occupancy status of the car space or lack the ability to dynamically evaluate the remaining capacity when dealing with task allocation. They continue to assign tasks to cars with few actual passengers but poor space release ability, resulting in the inability to carry new passengers, causing "idle running" and resource waste, and thus reducing the overall scheduling efficiency and service quality.

[0188] Accordingly, based on the A* search algorithm, the present invention models the current space occupancy capacity of the elevator car as an occupancy factor and dynamically combines this occupancy factor with the objective function, enabling the A* search process to more realistically reflect the actual load state, avoiding unreasonable task allocation, and thus obtaining a better elevator car scheduling scheme. The process of the optimization solution includes the following steps:

[0189] S71. Initialize the search tree and its root node.

[0190] Specifically, the search tree is a multi-way tree structure containing all potential scheduling allocation states of the current round, used to represent the search scope of the entire task allocation space; the root node is the starting node of the search tree, and its state includes the scheduling information of all elevator cars at the current moment, including the position of the current car, the car direction information, the already allocated task allocation scheme, the occupancy capacity of the current car, the occupancy factor, the currently selected objective function, and the set of hall call requests to be processed in the current round.

[0191] Among them, the calculation method of the occupancy factor is:

[0192]

[0193] In the formula, O n represents the occupancy factor of the nth node, used to quantify the influence weight of this space on the scheduling cost; C n represents the occupancy capacity of the car corresponding to the nth node, used to represent the space actually occupied by passengers and carried objects in the car; represents the total area of the car of the nth node.

[0194] Accordingly, through the introduction and calculation of the occupancy factor, the search tree can consider the space availability factor in the initialization stage, thus providing a more realistic cost constraint basis for subsequent path expansion.

[0195] Furthermore, the "already allocated task allocation scheme" in the root node is set according to different recognition paths, specifically:

[0196] If the current scheduling process is based on the usage mode, the root node can directly reference the task allocation scheme matching the current usage mode m in the case library as the initial task configuration;

[0197] If the current scheduling process is based on the traffic mode, the task allocation scheme of the root node defaults to an empty task allocation table or an initial allocation scheme randomly generated based on heuristic rules.

[0198] Accordingly, the present invention improves the scheduling flexibility and response accuracy of the A* algorithm in the search process through the above-mentioned initialization methods according to different usage modes or traffic modes.

[0199] S72. Update and expand the nodes of the current search tree to generate several child nodes.

[0200] Specifically, for the root node or the parent node, unassigned hall call requests are sequentially assigned to each car according to the consistent car direction information to form several child nodes.

[0201] S73. Calculate the cost and sort the priorities of all child nodes of the search tree to obtain the total cost.

[0202] Specifically, the cost calculation is to multiply the occupancy factor by the objective function and add it to the estimated cost function to obtain the total cost of the current node. The specific calculation is as follows:

[0203] f(n) = O n × g(n) + h(n)

[0204] In the formula, f(n) represents the total cost of the current nth child node, which is used to reflect the comprehensive cost of the current task allocation scheme in terms of space load, task priority, and future cost; g(n) is the objective function filtered by the current nth node through the usage pattern or traffic pattern; h(n) is the estimated cost function of the nth node, which is used to evaluate the future feasibility and potential cost of the scheme and can be set based on factors such as the distance from the current position of the passenger to the nearest elevator, the occupancy level, and the direction consistency.

[0205] The priority sorting is to sort all child nodes according to the total cost, so as to preferentially expand the path with the lowest cost during subsequent updates.

[0206] S74. Determine whether the current search tree meets the termination condition: If so, terminate the search process and return the task allocation scheme of the node with the minimum current total cost to obtain the optimal car dispatching scheme; if not, take the child node as the parent node according to the order of the total cost and execute step S72.

[0207] Among them, the termination condition is that all cars are assigned at least one task, or the preset maximum search time is reached.

[0208] It should be noted that in each iteration, the present invention does not force the allocation of all requests, but adopts a rolling allocation strategy, that is, in each round of scheduling process, the system first ensures that each car during docking obtains at least one request task, and the remaining requests will be continuously allocated based on the state update in subsequent rounds to ensure the load balance and processing flexibility of task scheduling.

[0209] Accordingly, by constructing a search tree model that is constrained by the cost of actual space availability and combines a target function for path guidance, the present invention effectively avoids the unreasonable scheduling problems caused by ignoring the actual car load state in traditional algorithms. At the same time, by adopting a rolling allocation strategy, the task allocation process has good stageability and scalability, improving the response ability and resource utilization rate of elevator scheduling under complex traffic patterns.

[0210] Furthermore, after obtaining the optimal car scheduling plan, the case library can be updated based on the key parameters in this round of scheduling process to achieve continuous iterative optimization of the strategy. The update process includes:

[0211] Extract the object category c1, quantity information c2, occupied capacity c3, predicted passenger flow characteristics c4, and / or performance feedback index c5 after the execution of this round of scheduling plan, etc. from the object information of the current car, and combine them with the original usage pattern or the traffic pattern m identified by fuzzy recognition to construct the feature vector of the current scenario;

[0212] Among them, the performance feedback index includes but is not limited to: average passenger waiting time, maximum passenger waiting time, average elevator energy consumption, car utilization rate, request completion rate, and scheduling response time, etc., which are used to characterize the execution effect of the scheduling plan in the current scenario.

[0213] Take the feature vector of the current scenario and its corresponding scheduling strategy as a new set of cases and store them in the case library to complete the update of the case library;

[0214] Accordingly, by extracting scenario features and policy performance indicators after each round of scheduling and updating them to the case library, the dynamic accumulation and rapid adaptation ability of scheduling knowledge are realized, and then the dynamic evolution ability and scenario adaptability of the system in variable traffic scenarios are improved.

[0215] Compared with the prior art, the present invention effectively reflects the space release ability of the car by introducing an image recognition and capacity estimation mechanism and accurately modeling the available space of the car based on preset object mobility rules, avoiding resource waste and allocation imbalance in the scheduling process. At the same time, by introducing a historical case mechanism, the reuse efficiency of scheduling strategies for typical scenarios is improved; for unknown scenarios, the traffic pattern is identified by combining traffic flow prediction with fuzzy rules, and the target function is dynamically selected to match, realizing the accurate adaptation of scheduling strategies to the operation situation.

[0216] Furthermore, in the process of scheduling optimization, the occupancy factor and the target function are integrated, and combined with heuristic search to achieve efficient search and evaluation of the task plan, effectively improving the spatial rationality of task allocation and the global scheduling efficiency.

[0217] Accordingly, the present invention significantly improves the overall performance of the multi-elevator scheduling system in terms of space adaptability, strategy accuracy, and operation efficiency, and effectively solves the problems of inaccurate scheduling and response lag caused by the lack of capacity modeling in the prior art.

[0218] Based on the same inventive concept, the present application also provides an electronic device, which can be a terminal device such as a server, a desktop computing device, or a mobile computing device (e.g., a laptop computing device, a handheld computing device, a tablet computer, a netbook, etc.). The device includes one or more processors and a memory, where the processor is used to execute a program to implement the elevator car scheduling method based on a space pattern according to an embodiment of the present invention; the memory is used to store a computer program executable by the processor.

[0219] Based on the same inventive concept, the present application also provides a computer-readable storage medium, corresponding to an embodiment of the above-mentioned elevator car scheduling method based on a space pattern. The computer-readable storage medium stores a computer program thereon, and when the program is executed by a processor, it implements the steps of the elevator car scheduling method based on a space pattern recorded in any of the above embodiments.

[0220] The present application can be in the form of a computer program product implemented on one or more storage media (including but not limited to disk memories, CD-ROMs, optical memories, etc.) containing program codes. Computer-usable storage media include permanent and non-permanent, removable and non-removable media, and information storage can be implemented by any method or technology. The information can be computer-readable instructions, data structures, program modules, or other data. Examples of computer storage media include but are not limited to: phase change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technologies, compact disc read-only memory (CD-ROM), digital versatile disc (DVD) or other optical storage, magnetic cassette tapes, magnetic disk storage or other magnetic storage devices, or any other non-transmission media that can be used to store information accessible by a computing device.

[0221] The above-described embodiments merely represent several implementation manners of the present invention, and their descriptions are relatively specific and detailed, but should not be construed as limiting the scope of the invention patent. It should be noted that for those of ordinary skill in the art, without departing from the inventive concept of the present invention, several modifications and improvements can still be made, and the present invention also intends to include these modifications and improvements.

Claims

1. An elevator car scheduling method based on a spatial pattern, characterized in that, It includes the following steps: S1. Conduct object detection on the image when the current car stops to obtain the object information in the current car; wherein, the object information includes object type, object quantity, object position information, and occupied size; S2. According to the target car information table and the preset object mobility rules, estimate the capacity of the object information in the current car to obtain the occupied capacity of the current car; S3. Judge whether there is a historical case for the object information and occupied capacity of the current car according to the case library: if so, execute step S6; if not, execute step S4; S4. Conduct traffic flow prediction based on the object information, car direction information of the current car and historical data to obtain the arrival rate of each traffic component in the future time domain; S5. According to the fuzzy rule base, conduct fuzzy recognition on the arrival rate of each traffic component in the future time domain to obtain the traffic mode of the current car; S6. Select a strategy according to the usage mode or traffic mode of the historical case to determine the objective function of the current car; S7. Optimize and solve the corresponding objective function according to the occupied capacity of the current car to obtain the car scheduling plan.

2. The elevator car scheduling method based on a spatial pattern according to claim 1, wherein The capacity estimation is used to quantify the ability of the current car to make room for newly added passengers, and the calculation method of the occupied capacity is as follows: where C represents the occupancy capacity of the current car; S {cab} represents the area of the empty car; wherein, the target car information table includes the standardized dimensions of the elevator car and the mapping relationship between the object occupancy dimensions and the actual dimensions; S i represents the occupied area of the \(i\)th object, which is obtained by mapping the occupied size of the \(i\)th object through the target car information table; classes represents the total number of objects in the current car; L is the space utilization factor, and its specific calculation is shown as follows: where P i is the mobility level of the i-th object.

3. The elevator car scheduling method based on a spatial pattern according to claim 2, characterized in that The traffic components include the upward traffic arrival rate, the downward traffic arrival rate, and the inter-floor traffic arrival rate, and their specific calculations are as follows: where λ in is the upward traffic arrival rate, which is used to represent the proportion of the number of people rising from the main floor to the total number of passengers; λ out is the downward traffic arrival rate, which is used to represent the proportion of the number of people descending to the main floor among the total passengers; λ inc is the inter - floor traffic arrival rate, which is used to represent the proportion of the number of people moving on the internal floors to the total number of people; up represents the number of passengers who depart from the main floor, enter the car and move upward; down represents the number of passengers whose car descends to the main floor; inter represents the number of passengers moving between non - main floors; all represents the total number of passengers carried in the current car; Among them, the traffic flow prediction is modeled and predicted according to the arrival rate of each traffic component of the current car and the arrival rate of each traffic component of historical data to obtain the arrival rate of each traffic component in the future time domain, and its specific calculation is as follows: F t+1 = α t Y t +(1 - α t )F t where, F t represents the arrival rates of each traffic component predicted at the t-th moment, including the arrival rate of the upstream traffic, the arrival rate of the downstream traffic, and the arrival rate of the interlayer traffic; Y t represents the actual arrival rates of each traffic component at the t-th moment, including the arrival rate of the upstream traffic, the arrival rate of the downstream traffic, and the arrival rate of the interlayer traffic; α t represents the adaptive smoothing factor at the t-th moment, and its specific calculation is as follows: where E t represents the smoothing error at the t-th moment, and the specific calculation is as follows: E t = βe t + (1 - β)E t-1 where β represents the smoothing factor; e t is the current prediction error, and its specific calculation is as follows: e t = Y t - F t AE t represents the absolute smoothing error at the t-th moment, and its specific calculation is as follows: AE t = β|e t | + (1 - β)AE t-1 where, |e t | represents the absolute value of the current prediction error.

4. The elevator car scheduling method based on a spatial pattern according to claim 3, wherein, The fuzzy recognition includes the following steps: S51A. Fuzzify the arrival rate of each traffic component in the future time domain by using the traffic component membership function to obtain the membership degrees of the upward traffic component, the downward traffic component, and the inter-floor traffic component; S51B. Fuzzify the ratio of the total arrival rate to the processing capacity by using the traffic intensity membership degree to obtain the membership degree of the traffic intensity; S52. According to the fuzzy rule base, defuzzify the membership degrees of the upward traffic component, the downward traffic component, the inter-floor traffic component, and the traffic intensity to obtain the traffic mode of the current car.

5. The elevator car scheduling method based on spatial patterns according to claim 4, wherein, The process of the optimization solution includes the following steps: S71. Initialize the search tree and its root node; wherein, the root node is the starting node of the search tree, and its state includes the scheduling information of all elevator cars at the current moment, including the position of the current car, the car direction information, the allocated task assignment plan, the occupied capacity of the current car, the occupancy factor, the currently selected objective function, and the set of hall call requests to be processed in the current round; Among them, the calculation method of the occupancy factor is: Where, O n represents the occupancy factor of the nth node; C n represents the occupancy capacity of the car corresponding to the nth node; represents the total car area of the nth node; S72. Update and expand the root node or the parent node of the current search tree to generate several child nodes; S73. Calculate the cost and prioritize the search tree for all child nodes to obtain the total cost; Among them, the cost calculation is to multiply the occupancy factor by the objective function and add it to the estimated cost function to obtain the total cost of the current node, and its specific calculation is as follows: f(n) = O n × g(n) + h(n) Wherein, f(n) represents the total cost of the current nth child node; g(n) is the objective function filtered by the current nth node using the mode or traffic mode; h(n) is the estimated cost function of the nth node; S74. Determine whether the current search tree meets the termination condition: If so, terminate the search process and return the task allocation plan of the node with the minimum current total cost to obtain the optimal car scheduling plan; if not, take the child node as the parent node according to the order of the total cost, and execute step S72.

6. An elevator car scheduling device based on a spatial pattern, characterized in that, It includes an object detection unit, a capacity estimation unit, a historical case judgment unit, a traffic flow prediction unit, a traffic mode recognition unit, an objective function selection unit, and a car scheduling plan optimization unit; The object detection unit is used to perform object detection on the image when the current car stops to obtain the object information in the current car; wherein, the object information includes the object type, the number of objects, the object position information, and the occupied size; The capacity estimation unit is used to estimate the capacity of the object information in the current car according to the target car information table and the preset object mobility rules to obtain the occupied capacity of the current car; The historical case judgment unit is used to judge whether there is a historical case for the object information and the occupied capacity of the current car according to the case library: If so, call the objective function selection unit; if not, call the traffic flow prediction unit; The traffic flow prediction unit is used to perform traffic flow prediction based on the object information, the car direction information of the current car and the historical data to obtain the arrival rate of each traffic component in the future time domain; The traffic mode recognition unit is used to perform fuzzy recognition on the arrival rate of each traffic component in the future time domain according to the fuzzy rule base to obtain the traffic mode of the current car; The objective function selection unit is used to select a strategy according to the usage mode or traffic mode of the historical case to determine the objective function of the current car; The car scheduling plan optimization unit is used to optimize and solve the corresponding objective function according to the occupied capacity of the current car to obtain the car scheduling plan.

7. The elevator car scheduling device based on a spatial pattern according to claim 6, characterized in that The capacity estimation is used to quantify the ability of the current car to make room for newly added passengers, and the calculation method of the occupied capacity is as follows: Wherein, C represents the occupancy capacity of the current car; S {cab} represents the area of the empty car; wherein, the target car information table includes the standardized dimensions of the elevator car and the mapping relationship between the object occupancy dimension and the actual dimension; S i represents the occupied area of the i-th object, which is obtained by mapping the occupied dimensions of the i-th object through the target car information table; classes represents the total number of objects in the current car; L is the space utilization factor, and its specific calculation is shown as follows: where P i is the mobility level of the i-th object.

8. The elevator car scheduling device based on a spatial pattern according to claim 7, characterized in that The traffic components include the upward traffic arrival rate, the downward traffic arrival rate, and the inter-floor traffic arrival rate, and their specific calculations are as follows: where λ in is the upward traffic arrival rate, which is used to represent the proportion of the number of people rising from the main floor to the total number of passengers; λ out is the downward traffic arrival rate, which is used to represent the proportion of the number of people descending to the main floor among the total passengers; λ inc is the inter-floor traffic arrival rate, which is used to represent the proportion of the number of people moving within the building floors to the total number of people; up represents the number of passengers departing from the main floor, entering the car and moving upward; down represents the number of passengers in the car descending to the main floor; inter represents the number of passengers moving between non-main floors; all represents the total number of passengers carried in the current car; Wherein, the traffic flow prediction is modeled and predicted based on the arrival rate of each traffic component of the current car and the arrival rate of each traffic component of the historical data to obtain the arrival rate of each traffic component in the future time domain, and its specific calculation is as follows: F t+1 = α t Y t + (1 - α t )F t Where F t represents the arrival rates of each traffic component predicted at the t-th moment, including the arrival rate of the upward traffic, the arrival rate of the downward traffic, and the arrival rate of the interlayer traffic; Y t represents the actual arrival rates of each traffic component at the t-th moment, including the arrival rate of the upward traffic, the arrival rate of the downward traffic, and the arrival rate of the interlayer traffic; α t represents the adaptive smoothing factor at the t-th moment, and its specific calculation is as follows: where E t represents the smoothing error at the t-th moment, and the specific calculation is as follows: E t = βe t + (1 - β)E t-1 where β represents the smoothing factor; e t is the current prediction error, and its specific calculation is as follows: e t = Y t - F t AE t Represents the absolute smoothing error at the t-th moment, and its specific calculation is shown as follows: AE t = β|e t | + (1 - β)AE t-1 where |e t | represents the absolute value of the current prediction error; Wherein, the fuzzy recognition includes the following steps: S51A. Fuzzify the arrival rate of each traffic component in the future time domain using the traffic component membership function to obtain the membership degrees of the upward traffic component, the downward traffic component, and the inter-floor traffic component; S51B. Fuzzify the ratio of the total arrival rate to the processing capacity using the traffic intensity membership to obtain the membership degree of the traffic intensity; S52. According to the fuzzy rule base, defuzzify the membership degrees of the upward traffic component, downward traffic component, inter - floor traffic component, and the membership degree of the traffic intensity to obtain the traffic mode of the current car.

9. The elevator car scheduling device based on a spatial pattern according to claim 8, wherein, The process of the optimization solution includes the following steps: S71. Initialize the search tree and its root node; wherein, the root node is the starting node of the search tree, and its state includes the scheduling information of all elevator cars at the current moment, which includes the position of the current car, car direction information, the allocated task assignment plan, the occupied capacity of the current car, occupancy factor, the currently selected objective function, and the set of hall - floor call requests to be processed in the current round; Among them, the calculation method of the occupancy factor is: Wherein, O n represents the occupancy factor of the nth node; C n represents the occupancy capacity of the car corresponding to the nth node; represents the total car area of the nth node; S72. Update and expand the root node or the parent node of the current search tree to generate several child nodes; S73. Calculate the cost and perform priority sorting on all child nodes of the search tree to obtain the total cost; Among them, the cost calculation is to multiply the occupancy factor by the objective function and add it to the estimated cost function to obtain the total cost of the current node. The specific calculation is as follows: f(n) = O n × g(n) + h(n) In the formula, f(n) represents the total cost of the nth child node currently; g(n) is the objective function screened by the current nth node through the usage mode or traffic mode; h(n) is the estimated cost function of the nth node; S74. Determine whether the current search tree meets the termination condition: If so, terminate the search process and return the task assignment plan of the node with the minimum current total cost to obtain the optimal car scheduling plan; if not, take the child node as the parent node according to the order of the total cost, and execute step S72.

10. A multi-elevator intelligent scheduling system, characterized in that, Including an image acquisition device installed in each elevator car, an elevator total control unit, and the elevator car scheduling device based on spatial mode according to any one of claims 6 - 9; The image acquisition device is used to perform real - time image acquisition inside the parked elevator car and transmit the acquired image when the car is parked to the elevator car scheduling device through wired or wireless means; The elevator total control unit is used to receive hall - floor call requests, the position and car direction information of the elevator car, and transmit them to the elevator car scheduling device through wired or wireless means; at the same time, schedule all elevator cars according to the car scheduling plan generated by the elevator car scheduling device.

Images