An elevator intelligent detection and scheduling method for special groups

Abstract

The application discloses a special group-oriented elevator intelligent detection and scheduling method. The method locally processes video data by deploying edge computing nodes, uses an improved YOLOv7-tiny model integrated with an attention mechanism to realize real-time identification of the age, luggage state and auxiliary tool use of passengers, dynamically allocates priority weights according to the identification results, constructs a multi-objective optimization function integrating time cost, energy cost and special service weight, uses a whale optimization algorithm with dynamic parameter self-adaptation to solve the function online, generates an optimal dispatching instruction sequence and sends the sequence to an elevator group control system for execution, and continuously monitors running data and feeds back and closes loop self-adaptively tunes the attribute identification model and optimization algorithm parameters based on historical data. The application realizes precise perception and personalized scheduling under high real-time performance and high privacy protection, and effectively improves the service fairness and overall operation efficiency of the elevator system for special groups.

Description

Technical Field

[0001] This invention belongs to the field of elevator intelligent control technology, specifically relating to an elevator intelligent detection and scheduling method for special groups, which is suitable for providing personalized elevator scheduling services for the elderly, people with mobility impairments, and passengers carrying large luggage. Background Technology

[0002] In modern high-rise buildings, the scheduling efficiency of elevator group control systems directly impacts passenger travel experience and building energy consumption. Traditional elevator scheduling systems often employ strategies based on fixed rules, such as "first-come, first-served" or "minimizing average waiting time." Their optimization objectives are relatively singular and fail to adapt to the diverse personalized needs of passengers. Especially for special groups such as the elderly, wheelchair users, or those carrying large luggage, their elevator usage behaviors exhibit significant differences and priority requirements. Traditional systems lack the ability to perceive passenger identities, resulting in a lack of human-centered scheduling strategies.

[0003] To enhance system perception capabilities, some existing technologies attempt to incorporate computer vision for passenger detection. However, most of these solutions rely on cloud-based video data processing, resulting in issues such as high transmission latency, high bandwidth consumption, and privacy risks, failing to meet the real-time and reliability requirements of elevator scheduling. Regarding optimization algorithms, while some research employs metaheuristic methods like genetic algorithms and particle swarm optimization for multi-objective scheduling, their optimization functions remain limited to traditional metrics such as time and energy consumption, failing to integrate real-time perceived fine-grained passenger attribute information, thus hindering truly personalized dynamic priority scheduling. Furthermore, existing systems generally lack self-learning and adaptive capabilities, unable to optimize their performance based on long-term operational data.

[0004] Therefore, existing technologies have the following shortcomings: lack of fine-grained, real-time perception of passenger attributes; single scheduling objective, failing to consider the priority service needs of special groups; poor system adaptability, unable to dynamically adjust strategies based on operational data; and insufficient real-time performance and privacy risks caused by cloud processing mode. Summary of the Invention

[0005] The purpose of this invention is to overcome the shortcomings of the prior art and provide an intelligent elevator detection and scheduling method for special groups of people. By integrating edge computing, an improved lightweight deep learning model and dynamic multi-objective optimization, an intelligent scheduling system with real-time, accurate, personalized and closed-loop self-optimization capabilities is constructed.

[0006] To achieve the above objectives, the present invention adopts the following technical solution:

[0007] An intelligent elevator detection and scheduling method for special groups includes the following steps:

[0008] S1. By deploying edge computing nodes in the elevator lobby and car, data from multiple video sensors is collected and processed in real time. When passengers waiting for the elevator are detected in the video, the passenger identity attribute recognition process is initiated.

[0009] S2. Based on an improved lightweight deep learning target detection model, passengers are detected in real time and their fine-grained attributes are identified. The identified attributes include at least age group, whether they are carrying large luggage, and whether they are using assistive devices. The improvement is achieved by introducing an attention mechanism to enhance the accuracy of fine-grained attribute identification.

[0010] S3. Based on the identified passenger identity attributes, dynamically assign personalized scheduling priority weights to them. This assignment is adaptively adjusted based on real-time changes in passenger attributes. Construct a multi-objective optimization function that includes total system time cost, total system energy cost, and personalized service weights.

[0011] S4. Based on real-time updated elevator call requests and passenger attribute information, a metaheuristic optimization algorithm is used to solve the multi-objective optimization function online, generating the current optimal elevator dispatch instruction sequence;

[0012] S5. Send the elevator dispatch command sequence to the corresponding elevator group control system to control the elevator operation;

[0013] S6. The system continuously monitors the elevator's operating status and passenger flow data, and periodically provides feedback and adaptive optimization of the parameters of the attribute recognition model and optimization algorithm based on accumulated historical scheduling data; the adaptive optimization includes online learning and optimization of the model and algorithm parameters based on historical data to form a closed-loop control system.

[0014] Preferably, the improved lightweight deep learning object detection model used in step S2 is a model based on the YOLOv7-tiny architecture and integrating an efficient channel attention (ECA) module, wherein the ECA module is embedded in the feature extraction layer of the YOLOv7-tiny backbone network to enhance the weights of key feature channels, thereby improving the accuracy of identifying passenger age groups, luggage carrying conditions, and use of assistive devices.

[0015] Preferably, the implementation of the model includes the following steps: S2.1, constructing a passenger image dataset, the dataset containing images labeled with different age groups, luggage carrying conditions, and use of assistive devices; the assistive devices include wheelchairs, strollers, canes, and walking aids; S2.2, integrating an efficient channel attention (ECA) module into the feature extraction layer of the YOLOv7-tiny backbone network to improve feature representation capabilities through cross-channel interaction; S2.3, using the dataset to train the YOLOv7-tiny network with integrated ECA module for multi-attribute recognition, and using a multi-task learning loss function to simultaneously optimize object detection and attribute classification tasks; S2.4, deploying the trained model on edge computing nodes, and adapting to the limitations of edge computing resources through model compression technology.

[0016] Preferably, the multi-objective optimization function constructed in step S3 is: in, This indicates the total waiting time and travel time for the elevator. This represents the total energy consumption of the system. This represents the priority service score for passengers with special needs; α, β, and γ are weighting coefficients, which are dynamically adjusted according to the number and type of special passengers identified. The adjustment rules are as follows: when a passenger using assistive devices is identified, the weight of γ is increased; when the system energy consumption exceeds the threshold, the weight of β is increased.

[0017] Preferably, the priority service score P special The calculation formula is: , where w i For passenger i, different weight values ​​are preset according to its attribute category, d i The factor characterizing the degree of service delay is defined as the ratio of passenger waiting time to the preset maximum tolerable time.

[0018] Preferably, the metaheuristic optimization algorithm used in step S4 is the whale optimization algorithm. The fitness function of the whale optimization algorithm is set as the multi-objective optimization function F, and dynamically adjusted search parameters are introduced in the solution process, including adaptively adjusting the convergence factor and spiral updating weights of the whale optimization algorithm according to real-time passenger flow.

[0019] Preferably, the feedback and adaptive tuning in step S6 includes: incrementally learning the attribute recognition model based on historical scheduling data to update the model parameters to adapt to new scenarios; and dynamically adjusting the parameters of the whale optimization algorithm and the weight coefficients α, β, γ in the multi-objective optimization function based on scheduling effect evaluation, wherein the scheduling effect evaluation is performed by comparing the deviation between the actual scheduling time and the predicted time.

[0020] Preferably, in step S5, the edge computing node sends the elevator dispatch command to the core controller of the elevator group control system through the industrial bus. The industrial bus adopts the CAN bus or Modbus protocol to ensure the real-time performance and reliability of the command transmission.

[0021] Compared with existing technologies, the advantages of this invention are as follows: By using edge computing nodes and a lightweight model integrating ECA attention mechanism, high-precision, low-latency identification of fine-grained passenger attributes is achieved locally, effectively protecting privacy; innovatively, the identified attributes are transformed into dynamic priority weights and integrated into a multi-objective optimization function, enabling scheduling to significantly improve service fairness for special groups while balancing efficiency and economy; an improved whale optimization algorithm with dynamically adaptive parameters is adopted to quickly generate near-optimal dispatch instructions in highly dynamic environments, ensuring real-time response; and continuous optimization of the model and parameters is achieved through historical data feedback, forming a technical closed loop with self-learning capabilities, improving the system's long-term robustness and overall performance. Attached Figure Description

[0022] Figure 1 This is an overall flowchart of the elevator intelligent scheduling optimization method for special groups of people as described in this invention;

[0023] Figure 2 A schematic diagram of the improved YOLOv7-tiny model structure for integrating ECA attention;

[0024] Figure 3 This is a schematic diagram of the multi-objective ladder decision-making process based on the whale optimization algorithm. Detailed Implementation

[0025] The technical solution of the present invention will now be clearly and completely described with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only for explaining the present invention and are not intended to limit the scope of protection of the present invention.

[0026] like Figure 1 As shown, this invention provides an intelligent elevator detection and scheduling method for special groups of people. Its core lies in constructing a closed-loop intelligent system encompassing perception, decision-making, execution, and optimization. Specific implementation methods are as follows:

[0027] Step S1: Edge computing nodes acquire and process video data in real time.

[0028] Edge computing nodes based on ARM architecture (such as NVIDIA Jetson Nano and TX2 series) or lightweight GPUs are deployed in elevator lobbies and elevator cars. These nodes connect to high-definition wide-angle cameras, run embedded Linux operating systems, and integrate computer vision and inference acceleration libraries such as OpenCV and TensorRT. Each node incorporates motion detection and background subtraction algorithms to perceive passengers appearing in the video stream in real time. Once a target is detected, the subsequent attribute recognition process is immediately triggered. To fundamentally ensure real-time performance and user privacy, all video stream analysis and processing are performed locally on the edge nodes. Raw video data is discarded immediately after processing; only structured recognition results (including attribute labels and their confidence levels for passenger location bounding boxes, age groups, luggage status, and assistive device types) are uploaded to the central dispatcher via the internal network. Each edge node has local caching and disconnection resumption capabilities to cope with temporary network interruptions and ensure that dispatch instructions are not lost.

[0029] Step S2: Passenger fine-grained attribute identification based on an improved lightweight deep learning model

[0030] Step S2 is the foundation for personalized scheduling, and its key lies in using the improved YOLOv7-tiny model for high-precision and high-efficiency attribute identification.

[0031] S2.1 Dataset Construction and Annotation: Collect and construct a passenger image dataset covering diverse scenes (different lighting, viewpoints, occlusions), with a total of no less than 10,000 images. The dataset should specifically include images and video sequences of different age groups (children, youth, middle-aged, elderly), different luggage statuses (none, small items, large items), and the use of various assistive devices (wheelchairs, strollers, canes, walking aids). Use tools such as LabelImg to perform fine-grained annotation on the dataset, including the bounding boxes of the targets and corresponding multi-attribute labels (such as "elderly person," "carrying large luggage," "using a wheelchair," etc.).

[0032] S2.2 Model Structure Improvement: In the feature extraction layer of the standard YOLOv7-tiny backbone network, an efficient channel attention (ECA) module is integrated after the CBS (Conv-BN-SiLU) module. This module performs global average pooling (GAP) on the feature map, followed by one-dimensional convolution (the kernel size k is determined by the number of channels C using a function). Adaptively determined (where γ and b are hyperparameters) and sigmoid activation are used to generate weight vectors for each channel, which are then multiplied channel-by-channel with the original feature map to achieve feature recalibration. This allows the model to focus more on feature channels useful for fine-grained attribute discrimination with limited computational resources, thereby improving recognition accuracy while maintaining lightweight design.

[0033] S2.3 Model Training and Deployment: Using a labeled dataset, perform end-to-end training of the improved model in PyTorch or a similar framework. A composite loss function is employed. ,in For bounding box regression loss, For physical loss, For cross-entropy loss in multi-attribute classification, , , To balance the weights, after training, the model is converted to an efficient inference format such as ONNX or TensorRT, and deployed to edge computing nodes using tools such as OpenVINO or TensorRT. To further adapt to edge resources, INT8 quantization is used to compress the model, ensuring that the inference speed on the target hardware remains stable at ≥15 fps, meeting real-time requirements.

[0034] Step S3: Dynamic Priority Assignment and Construction of Multi-Objective Optimization Function

[0035] The central dispatcher dynamically assigns personalized dispatch priority weights based on passenger attributes identified by S2. i The following is an example of the preset weighting rules: the base weight for regular passengers is 1; the weight for elderly passengers or passengers carrying large luggage is 2; and the weight for passengers using wheelchairs, canes, or walking aids is 3. This weighting allocation rule can be configured according to the building type (such as hospital, office building).

[0036] The multi-objective optimization function for elevator dispatching decisions is constructed as follows:

[0037] in, This represents the total waiting and riding time for all passengers in the system (which can be estimated based on elevator operation curves and floor distances). This represents the total energy consumption of all elevators in the system (which can be estimated based on a model of start-stop and travel distance). This represents the total score for priority service for special passengers. Where d... i Let the service delay factor for passenger i be defined as follows: , This is the passenger's current waiting time. The system sets the maximum tolerable waiting time for each attribute (e.g., 90 seconds for wheelchair users). α, β, and γ are dynamic weighting coefficients, with initial values ​​of [0.5, 0.2, 0.3]. The system dynamically adjusts based on the real-time scenario: when a passenger using assistive devices is detected, γ increases in increments (e.g., 0.1); when the system's instantaneous power or total energy consumption exceeds a preset threshold, β increases; and during off-peak hours, α can be appropriately increased to improve efficiency.

[0038] Step S4: Online solution based on the improved Whale Optimization Algorithm (WOA)

[0039] The Whale Optimization Algorithm (WOA) is used as a metaheuristic optimization algorithm to solve the above multi-objective function F online quickly. Each possible elevator dispatch instruction sequence (e.g., assigning which elevator to which passenger) is encoded as a position vector of an individual whale in the search space.

[0040] The algorithm's fitness function is directly set to F, aiming to find the dispatch scheme that minimizes the F value. Based on the standard WOA algorithm, dynamically adjusted search parameters are introduced to improve performance in highly dynamic elevator dispatching scenarios: the linear decrease of the convergence factor a from 2 to 0 is replaced by adaptively adjusting the deceleration rate based on the real-time monitored total number of waiting passengers. The deceleration is slowed down when there are many passengers to enhance global exploration, and accelerated when there are few passengers to achieve rapid convergence. The spiral update weight b is also randomly adjusted within a certain range according to the dynamic changes in the system state. The algorithm iteratively updates the population position through three stages: simulating prey encirclement, bubble net attack, and random search, searching for an approximate optimal solution within a set maximum number of iterations (e.g., 50-100 times) or a convergence threshold. This solution process is triggered each time a new elevator call is received, a new passenger is identified, or the system state changes significantly, ensuring real-time dispatching responsiveness.

[0041] Step S5: Execute the ladder dispatch command

[0042] The optimal elevator dispatch command sequence obtained from the solution is sent from the central dispatcher to the core controller of the elevator group control system via a highly reliable industrial bus (such as CAN bus or Modbus protocol), thereby controlling the corresponding elevators to perform operating tasks. This communication method ensures the real-time performance and reliability of command transmission.

[0043] Step S6: System monitoring and feedback optimization (closed-loop control)

[0044] This step is crucial for achieving long-term system self-adaptation and continuous optimization. The system continuously monitors and records historical data such as elevator operating status (actual waiting time, riding time, energy consumption), passenger flow data, and the accuracy of attribute recognition, forming a scheduling log database.

[0045] Feedback and adaptive tuning mainly include the following two aspects:

[0046] (1) Online incremental learning of the model: Periodically (e.g., weekly) or when the recognition accuracy is consistently below the threshold (e.g., 92%), the system automatically extracts image data with correct attribute labels from a recent period (e.g., the past month) from historical data (a small portion can be reviewed and confirmed by the administrator) to perform incremental learning on the attribute recognition model in S2, fine-tuning the parameters of the fully connected layers or some convolutional layers of the model so that it can adapt to environmental changes such as new lighting conditions, passenger clothing, and new auxiliary tools, and prevent the model performance from degrading.

[0047] (2) Dynamic Algorithm and Parameter Tuning: Based on the evaluation of historical scheduling effects, system parameters are dynamically adjusted. For example, the mean absolute error (MAE) between the model's predicted passenger waiting time and the actual measured value is calculated weekly. If the MAE continues to increase, the parameters of the whale optimization algorithm in S4 are automatically adjusted (e.g., appropriately increasing the population size or the number of iterations). At the same time, based on the completion of scheduling objectives (such as special passenger satisfaction surveys, total energy consumption bills), the weight coefficients α, β, and γ of the multi-objective optimization function in S3 are dynamically optimized using regression analysis or reinforcement learning, so that the system strategy continuously approaches the optimal balance point of actual operation.

[0048] Through continuous feedback and optimization by S6, the entire system forms a closed-loop control system with self-learning and self-adjustment capabilities, which can maintain efficient, reliable and user-friendly operation for a long time.

[0049] Example:

[0050] This embodiment deploys the invention in a group control system for three elevators in a five-story office building. The system identifies a passenger using a wheelchair (attribute recognition confidence level 96%) through an edge node (hardware: NVIDIA Jetson Xavier NX) on the third floor, and assigns a priority weight w to this passenger. i= 3). When the central dispatcher receives this information, the total system energy consumption is at a normal level. Therefore, the weighting coefficients are dynamically adjusted to [α=0.4, β=0.2, γ=0.4] to prioritize special passengers. When using the improved WOA algorithm (population size 30, maximum iterations 80) to solve the multi-objective function F, due to the high weight of this passenger and the continuously increasing delay factor d_i, the generated elevator dispatching scheme prioritizes assigning elevator A, which is about to arrive near the 3rd floor, to serve the passenger, rather than following the traditional "first-come, first-served" principle. Actual operation results show that the waiting time for this wheelchair passenger has significantly decreased from an average of 65 seconds before system deployment to 22 seconds. After three months of operation and closed-loop optimization in step S6, an incremental learning was performed on the attribute recognition model based on approximately 1000 accumulated labeled data. The model's recognition accuracy for "wheelchair" and "large luggage" increased from the initial 92% and 88% to 96% and 93%, respectively. Meanwhile, by analyzing scheduling deviations, the system adaptively adjusted the WOA population size to 40. A questionnaire survey showed that the satisfaction rate of the special passenger group increased from 70% to 92%. This embodiment fully demonstrates the effectiveness, adaptability, and superiority of the present invention.

[0051] The above description is merely a specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any person skilled in the art can easily conceive of various equivalent modifications or substitutions within the technical scope disclosed in the present invention, and these modifications or substitutions should all be covered within the scope of protection of the present invention.

Claims

1. A method for intelligent elevator detection and scheduling for special populations, characterized in that, Includes the following steps: S1. By deploying edge computing nodes in the elevator lobby and car, data from multiple video sensors is collected and processed in real time. When passengers waiting for the elevator are detected in the video, the passenger identity attribute recognition process is initiated. S2. Based on an improved lightweight deep learning target detection model, passengers are detected in real time and their attributes are identified in fine detail. The identified attributes include at least age group, whether they are carrying large luggage, and whether they are using assistive devices. S3. Based on the identified passenger identity attributes, dynamically assign personalized scheduling priority weights to them, and construct a multi-objective optimization function that includes the total system time cost, the total system energy consumption cost, and the personalized service weights. S4. Based on real-time updated elevator call requests and passenger attribute information, a metaheuristic optimization algorithm is used to solve the multi-objective optimization function online, generating the current optimal elevator dispatch instruction sequence; S5. Send the elevator dispatch command sequence to the corresponding elevator group control system to control the elevator operation; S6. The system continuously monitors the elevator's operating status and passenger flow data, and periodically provides feedback and adaptive optimization of the parameters of the attribute recognition model and optimization algorithm based on the accumulated historical scheduling data. In step S2, the improved lightweight deep learning object detection model enhances the accuracy of fine-grained attribute recognition by introducing an attention mechanism. In step S3, the dynamic priority weight allocation is adaptively adjusted based on real-time changes in passenger attributes. In step S6, the adaptive tuning includes online learning and optimization of model and algorithm parameters based on historical data to form a closed-loop control system.

2. The method according to claim 1, characterized in that, The improved lightweight deep learning object detection model used in step S2 is a model based on the YOLOv7-tiny architecture and integrating the efficient channel attention module ECA. The efficient channel attention module ECA is embedded in the feature extraction layer of the YOLOv7-tiny backbone network to enhance the weight of key feature channels, thereby improving the accuracy of identifying passenger age groups, luggage carrying conditions, and use of assistive devices.

3. The method according to claim 2, characterized in that, The implementation of the model includes the following steps: S2.1 Construct a passenger image dataset, which includes images labeled with different age groups, luggage carrying conditions, and use of assistive devices; the assistive devices include wheelchairs, strollers, canes, and walking aids; S2.2 Integrate the efficient channel attention module ECA into the feature extraction layer of the YOLOv7-tiny backbone network to improve feature representation capability through cross-channel interaction; S2.

3. Use the dataset to train the YOLOv7-tiny network with integrated efficient channel attention module ECA for multi-attribute recognition, and use a multi-task learning loss function to simultaneously optimize object detection and attribute classification tasks; S2.4 Deploy the trained model on edge computing nodes and adapt it to the limitations of edge computing resources through model compression technology.

4. The method according to claim 1, characterized in that, The multi-objective optimization function constructed in step S3 is: in, This indicates the total waiting time and travel time for the elevator. This represents the total energy consumption of the system. This represents the priority service score for passengers with special needs; α, β, and γ are weighting coefficients, which are dynamically adjusted according to the number and type of special passengers identified. The adjustment rules are as follows: when a passenger using assistive devices is identified, the weight of γ is increased; when the system energy consumption exceeds the threshold, the weight of β is increased.

5. The method according to claim 4, characterized in that, The priority service score P special The calculation formula is: , where w i For passenger i, different weight values ​​are preset according to its attribute category, d i The factor characterizing the degree of service delay is defined as the ratio of passenger waiting time to the preset maximum tolerable time.

6. The method according to claim 1, characterized in that, The metaheuristic optimization algorithm used in step S4 is the whale optimization algorithm. The fitness function of the whale optimization algorithm is set as the multi-objective optimization function F, and dynamically adjusted search parameters are introduced in the solution process, including adaptively adjusting the convergence factor of the whale optimization algorithm and spiral updating the weights according to the real-time passenger flow.

7. The method according to claim 6, characterized in that, The feedback and adaptive optimization in step S6 includes: incrementally learning the attribute recognition model based on historical scheduling data to update the model parameters to adapt to new scenarios; and dynamically adjusting the parameters of the whale optimization algorithm and the weight coefficients α, β, γ in the multi-objective optimization function based on scheduling effect evaluation, wherein the scheduling effect evaluation is performed by comparing the deviation between the actual scheduling time and the predicted time.

8. The method according to claim 1, characterized in that, In step S5, the edge computing node sends the elevator dispatch command to the core controller of the elevator group control system through the industrial bus. The industrial bus adopts the CAN bus or Modbus protocol to ensure the real-time performance and reliability of the command transmission.

Images