Elevator control method and elevator control system

Abstract

The invention discloses an elevator control method and an elevator control system.The elevator control method comprises the steps that behavior data of passengers inside and outside an elevator are collected in real time through a camera, a radar and a TOF sensor; based on the behavior data, passenger intentions are judged through a data fusion algorithm, and the passenger intentions comprise the elevator entering intention and the elevator leaving intention; and when the confidence score exceeds a preset threshold value, a corresponding elevator door control signal is generated, and the opening or closing action of the elevator door is dynamically adjusted. Through multi-mode sensor fusion and an intelligent algorithm, the transportation efficiency of the elevator can be greatly improved, the waiting time of passengers is shortened, the energy consumption is optimized, and the operation efficiency of the elevator can be effectively improved especially in the high-flow or peak period.

Description

Technical Field

[0001] The present invention relates to the technical field of elevator control, and more specifically, to an elevator control method and an elevator control system. Background Art

[0002] With the high intelligence of modern buildings, elevator systems are increasingly used in daily life, especially in high-rise buildings, office buildings, commercial centers and other places, where efficient dispatching and intelligent control of elevators are becoming particularly important. Traditional elevator control systems mostly rely on preset door opening and closing times and floor requests, which results in the inability to quickly adjust the door opening and closing timing of elevators according to actual conditions when the passenger flow is large, greatly affecting the passenger experience and elevator operation efficiency.

[0003] Door opening and closing control in traditional elevator systems is usually based on a simple timing mechanism, ignoring the specific intentions and actual behaviors of passengers. For example, elevator doors are usually closed based on a time delay, regardless of whether the passengers are ready to enter or leave the elevator. This fixed control method often leads to the following problems:

[0004] Long passenger waiting times: When elevator doors close too early or without extended door opening time, passengers may not be able to fully enter or exit the elevator before the doors close.

[0005] Energy waste: Traditional systems cannot distinguish between passenger intentions and obstacles, and elevator doors frequently open and close, resulting in increased energy consumption. Especially during peak hours when elevators frequently stop and open doors, energy utilization efficiency is low.

[0006] In order to solve these problems, the elevator control system urgently needs to adopt a more intelligent door opening and closing control solution that can dynamically adjust the timing of door opening and closing according to the passengers' behavior, thereby improving the elevator's operating efficiency, passenger experience, and effectively reducing energy consumption. Summary of the invention

[0007] The elevator control method and system of the present invention, based on the technology of multi-sensor data fusion and passenger intention judgment, provides a more accurate and intelligent elevator door control solution, solving the above-mentioned problems in traditional elevator control systems.

[0008] A first aspect of the present invention provides an elevator control method, comprising:

[0009] Through cameras, radars and TOF sensors, the behavior data of passengers inside and outside the elevator is collected in real time;

[0010] Based on the behavior data, the passenger intention is judged through the data fusion algorithm, and the passenger intention includes the intention to enter the elevator and the intention to leave the elevator. The data fusion algorithm includes:

[0011] a. Obtain the passenger's facial orientation, hand movements and body orientation information through the camera;

[0012] b. Obtain the distance and moving speed between the passenger and the elevator door through radar or TOF sensor;

[0013] c. Generate a confidence score for the passenger’s intention by combining facial orientation, hand movements, body orientation, distance to the elevator door, and movement speed;

[0014] When the confidence score exceeds the preset threshold, a corresponding elevator door control signal is generated to dynamically adjust the opening or closing action of the elevator door.

[0015] Optionally, the passenger intent also includes denial of entry to the elevator when the passenger is outside the elevator.

[0016] Optionally, the judgment conditions for the intention to enter the elevator include: the passenger's face is facing the elevator door and the duration exceeds a first threshold; the passenger's hand movements conform to a preset trajectory close to the edge of the elevator door; the distance between the passenger and the elevator door is less than a second threshold and the moving speed is close to zero.

[0017] Optionally, the judgment conditions for the intention to enter the elevator include: recognizing through the camera that the passenger's face is facing the elevator door and the focus of vision is on the door frame area; detecting through the TOF sensor that the distance between the hand and the door edge is less than 20 cm and lasts for more than 0.5 seconds; generating a door opening command when the above conditions are met at the same time.

[0018] Optionally, the judgment conditions for the intention to leave the elevator include: the passenger's body is facing outside the elevator door and the duration exceeds a third threshold; the passenger's moving speed is greater than a fourth threshold; the distance between the passenger and the elevator door increases over time.

[0019] Optionally, determining the passenger's intention through a data fusion algorithm includes: extracting skeleton key points from the camera image to obtain coordinates of multiple body joint points; aligning radar point cloud data with the joint point coordinates in time and space; and using a Kalman filter to predict the passenger's motion trajectory to generate an intention confidence score.

[0020] Optionally, the spatiotemporal alignment includes: establishing a three-dimensional coordinate system with the center of the elevator door as the origin; clustering the radar point cloud data to extract passenger contour feature points; and aligning the feature point cloud with the image joint point cloud through an ICP algorithm.

[0021] Optionally, the spatiotemporal alignment includes: analyzing action sequences of more than 5 consecutive frames through an LSTM network to identify typical entry and exit modes; the typical entry and exit modes include three preset postures: holding the door with one hand, avoiding sideways, and dragging luggage.

[0022] Optionally, the dynamic adjustment includes: when more than 3 passengers are detected entering at the same time, extending the door opening time reference value T0 to 1.5T0; if the width of the items carried by any passenger exceeds 60% of the door width, an additional delay time of 0.2T0 to 0.4T0 is added.

[0023] Optionally, an exception handling mechanism is also included:

[0024] When the radar detects a stationary obstacle in the door area for more than 10 seconds, the TOF sensor high-precision scanning mode is activated; if it is confirmed that non-passenger items are left behind, a voice prompt is triggered and the door control operation is suspended.

[0025] Optionally, determining the passenger's intention through a data fusion algorithm includes: determining whether the passenger has a sudden intention to enter or leave the elevator by detecting the passenger's proximity to the elevator door and combining the passenger's body orientation.

[0026] Optionally, the elevator control method combines sensor data and historical data of passenger intention judgment, and uses a reinforcement learning method to optimize the timing of elevator door opening to enhance passengers' waiting experience.

[0027] A second aspect of the present invention discloses an elevator control system, the system comprising:

[0028] Multiple sensor modules are used to obtain passenger data inside and outside the elevator; a data processing module is used to analyze and fuse the sensor data to determine the passenger's intention to enter or leave; and a control module is used to generate an elevator door control signal based on the judgment result.

[0029] Optionally, the elevator control system also includes a feedback learning module: recording sensor data and result evaluation of each door control operation; when misjudgment of the same scene occurs five times in a row, the model incremental training process is automatically triggered.

[0030] According to the technical content disclosed in the present invention, the following beneficial effects are achieved:

[0031] The present invention uses cameras, radars and TOF sensors to collect the behavioral data of passengers inside and outside the elevator in real time; and based on obtaining at least two types of behavioral data, the passenger's intention is judged through a data fusion algorithm. When the passenger intends to enter or leave the elevator, the elevator door opening time is extended to avoid repeated opening and closing of the elevator door; the present invention can greatly improve the transportation efficiency of the elevator, reduce the waiting time of passengers, and optimize energy consumption through multi-modal sensor fusion and intelligent algorithms, especially during high traffic or peak hours, and can effectively improve the operating efficiency of the elevator.

[0032] Further features and advantages of the present invention will become apparent from the following detailed description of exemplary embodiments of the present invention with reference to the attached drawings. BRIEF DESCRIPTION OF THE DRAWINGS

[0033] The accompanying drawings, which are incorporated in and constitute a part of the specification, illustrate embodiments of the invention and, together with the description, serve to explain the principles of the invention.

[0034] Figure 1 This is a diagram of the elevator control system of the present invention.

[0035] Figure 2 This is a flow chart of the elevator control method of the present invention. DETAILED DESCRIPTION

[0036] Various exemplary embodiments of the present invention will now be described in detail with reference to the accompanying drawings. It should be noted that the relative arrangement of components and steps, numerical expressions and numerical values ​​set forth in these embodiments do not limit the scope of the present invention unless otherwise specifically stated.

[0037] The following description of at least one exemplary embodiment is merely illustrative in nature and is in no way intended to limit the invention, its application, or uses.

[0038] Technologies, methods, and equipment known to ordinary technicians in the relevant art may not be discussed in detail, but where appropriate, the technologies, methods, and equipment should be considered part of the specification.

[0039] In all examples shown and discussed herein, any specific values ​​should be interpreted as merely exemplary and not limiting. Therefore, other examples of the exemplary embodiments may have different values.

[0040] It should be noted that like reference numerals and letters refer to similar items in the following figures, and therefore, once an item is defined in one figure, it need not be further discussed in subsequent figures.

[0041] See also Figure 1 The first aspect of the present invention discloses an elevator control system for identifying the intention of passengers who are about to enter the elevator and those who are already inside the elevator, and formulating an elevator operation plan in a targeted manner according to the passengers' intention of entering the elevator, thereby improving the elevator's operating efficiency and reducing the elevator's energy consumption.

[0042] The elevator control system includes multiple sensor modules for acquiring passenger data inside and outside the elevator; specifically, they include cameras, radars, pressure distribution sensors, and TOF sensors, wherein the cameras can be used in combination with the radars and TOF sensors separately or the three can be used in combination.

[0043] The camera is mainly used to obtain images of the user's facial orientation, facial expressions, hand movements, etc., for analyzing the user's behavior. In actual applications, the camera uses a binocular camera: installed on the top of the elevator door frame to capture facial orientation and hand movements;

[0044] The radar mainly detects the distance and movement speed of passengers; the radar is a millimeter-wave radar: deployed on the door side (1.2m from the ground), it detects the movement trajectory of passengers (working frequency 60GHz, detection distance 0.1-10m); it can penetrate obstructions to detect the speed and direction of passengers.

[0045] The TOF sensor generates 3D point cloud to assist positioning. The TOF sensor is integrated into the four corners of the inner wall of the car to generate 3D point cloud data (accuracy ±1cm) to locate the distribution of passengers.

[0046] The pressure distribution sensor is installed under the car floor with a measuring range of 0-2000kg and an accuracy of ±0.5%. It is used to detect overload and pressure concentration areas.

[0047] The data processing module is used to analyze and fuse sensor data to determine the passenger's intention to take the elevator, including the intention to continue waiting, enter or leave;

[0048] The data processing module includes a data synchronization unit and a feature extraction unit, wherein:

[0049] The data synchronization unit is used to align multi-sensor timestamps (error < 10ms);

[0050] The feature extraction unit includes:

[0051] Visual features: OpenPose extracts 17 skeleton key points;

[0052] Radar features: Kalman filter predicts the position in the next 0.5 seconds;

[0053] TOF features: point cloud clustering to generate passenger profiles.

[0054] The passenger intention judgment method of the present invention is based on the data fusion of multimodal sensors, and the specific steps include:

[0055] 1. Data input: The camera captures 5 consecutive frames of images, the radar provides the passenger's movement trajectory, and the TOF sensor provides three-dimensional spatial data.

[0056] 2. Intent Classification:

[0057] 1. Entry intention: face facing the elevator door and hands close to the door edge.

[0058] 2. Intention to leave: The body turns toward the elevator door and the moving speed exceeds 0.8m / s.

[0059] Intention judgment conditions:

[0060] Entry intention (must be met at the same time):

[0061] a. The face is facing the elevator door and the sight focus is on the door frame area (lasting ≥ 0.5 seconds);

[0062] b. The distance between the hand and the door edge is ≤ 20cm (TOF sensor detection);

[0063] c. The moving speed approaches zero (verified by radar data).

[0064] Intention to leave (any two of them):

[0065] a. The body faces the elevator door (lasting ≥ 0.3 seconds);

[0066] b. Moving speed ≥ 0.8m / s;

[0067] c. The distance from the elevator door increases.

[0068] Dynamic adjustment strategy:

[0069] Multiple-passenger entry scenario: When more than 3 passengers enter at the same time, the door opening time is extended to 1.5T0;

[0070] Large item scenario: When the width of the item is greater than 60% of the door width, an additional 0.3T0 delay will be added;

[0071] Abnormal handling: If a stationary obstacle is detected for more than 10 seconds, a TOF high-precision scan is triggered and a voice alarm is sounded.

[0072] On this basis, the time series data is analyzed based on the LSTM network, and finally the decision on the opening and closing time of the door is generated.

[0073] The control module is used to generate an elevator door control signal according to the judgment result.

[0074] Specifically, the control module includes:

[0075] Dynamic time calculator: adjusts the base time based on the number of passengers, item volume, and historical efficiency;

[0076] Abnormal response unit: hierarchical triggering of voice prompts, door locks, and remote alarms;

[0077] Group control dispatch interface: Exchange passenger flow forecast data with the building IoT system.

[0078] In some embodiments, it is possible to determine whether a passenger intends to take the elevator by identifying the facial expressions and eyes of the passengers. For example, Xiao A stands at the elevator door. When the elevator door opens, Xiao A finds a dog in the elevator. Since Xiao A is naturally afraid of dogs, he does not want to enter the elevator. His facial expression is fearful or disgusted, and his body posture is backward or ready to retreat. At this time, Xiao A refuses to enter even if the elevator door opens. When the system recognizes that Xiao A does not want to enter the elevator, it determines that Xiao A's intention is to continue waiting. At this time, the control module closes the elevator door, and the elevator continues to run without waiting for Xiao A to enter.

[0079] In order to prevent system errors, the elevator control system also includes a feedback learning module, which is used to record the sensor data and result evaluation of each door control operation; when there are five consecutive misjudgments of the same scene, the model incremental training process is automatically triggered.

[0080] See also Figure 2 The second aspect of the present invention discloses an elevator control method based on the elevator control system of the first aspect of the present invention, comprising:

[0081] Collecting the behavior data of passengers inside and outside the elevator in real time through at least one of a camera, a radar and a TOF sensor;

[0082] Preferably, in this embodiment, a camera and a radar are used together with a TOF sensor and a pressure distribution sensor;

[0083] Based on the behavior data, the passenger intention is judged through the data fusion algorithm, and the passenger intention includes the intention to enter the elevator and the intention to leave the elevator. The data fusion algorithm includes:

[0084] a. Obtain the passenger's facial orientation, hand movements and body orientation information through the camera;

[0085] b. Obtain the distance and moving speed between the passenger and the elevator door through radar or TOF sensor;

[0086] c. Generate a confidence score for the passenger’s intention by combining facial orientation, hand gestures, body orientation, distance, and movement speed;

[0087] When the confidence score exceeds the preset threshold, a corresponding elevator door control signal is generated to dynamically adjust the opening or closing action of the elevator door.

[0088] Further, the passenger intention also includes being denied entry into the elevator when the passenger is outside the elevator.

[0089] Furthermore, the judgment conditions for the intention to enter the elevator include: the passenger's face is facing the elevator door and the duration exceeds a first threshold; the passenger's hand movements conform to a preset trajectory close to the edge of the elevator door; the distance between the passenger and the elevator door is less than a second threshold and the moving speed is close to zero.

[0090] Furthermore, the judgment conditions for the intention to enter the elevator include: recognizing through the camera that the passenger's face is facing the elevator door and the focus of vision is located in the door frame area; detecting through the TOF sensor that the distance between the hand and the door edge is less than 20cm and lasts for more than 0.5 seconds; generating a door opening command when the above conditions are met at the same time. More preferably, through the 3D point cloud data of the TOF sensor, combined with the trajectory prediction model of the key points of the hand skeleton, the distance threshold (range 15-25cm) and duration (0.3-0.8 seconds) between the hand and the door edge are dynamically adjusted to adapt to the intention judgment of passengers of different heights.

[0091] Furthermore, the judgment conditions for the intention to leave the elevator include: the passenger's body is facing outside the elevator door and the duration exceeds a third threshold; the passenger's moving speed is greater than a fourth threshold; the distance between the passenger and the elevator door increases over time.

[0092] Furthermore, judging the passenger's intention through the data fusion algorithm includes: extracting skeleton key points from the camera image to obtain the coordinates of multiple body joints; aligning the radar point cloud data with the joint coordinates in time and space; and using the Kalman filter to predict the passenger's motion trajectory and generate an intention confidence score.

[0093] Furthermore, the spatiotemporal alignment includes: establishing a three-dimensional coordinate system with the center of the elevator door as the origin; clustering the radar point cloud data to extract the passenger contour feature points; and aligning the feature point cloud with the image joint point cloud through the ICP algorithm.

[0094] Furthermore, spatiotemporal alignment includes: analyzing action sequences of more than 5 consecutive frames through the LSTM network to identify typical entry and exit modes; typical entry and exit modes include three preset postures: holding the door with one hand, avoiding sideways, and dragging luggage.

[0095] Furthermore, the dynamic adjustment includes: when more than 3 passengers are detected entering at the same time, the door opening time base value T0 is extended to 1.5T0; if the width of the items carried by any passenger exceeds 60% of the door width, an additional delay time of 0.2T0 to 0.4T0 is added.

[0096] Furthermore, it also includes exception handling mechanisms:

[0097] When the radar detects a stationary obstacle in the door area for more than 10 seconds, the TOF sensor high-precision scanning mode is activated; if it is confirmed that non-passenger items are left behind, a voice prompt is triggered and the door control operation is suspended.

[0098] Furthermore, judging the passenger's intention through the data fusion algorithm includes: judging whether the passenger has a sudden intention to enter or leave the elevator by detecting the passenger's proximity to the elevator door and combining the passenger's body orientation.

[0099] Furthermore, the elevator control method combines sensor data and historical data of passenger intention judgment, and uses reinforcement learning methods to optimize the timing of elevator door opening to enhance passengers' waiting experience.

[0100] Furthermore, when it is detected that a passenger is carrying a pet or a stroller, the door opening time is automatically extended to 2T0 through TOF sensor contour matching and pressure distribution feature analysis, and the safety warning light in the car is activated.

[0101] Furthermore, “the feedback learning module generates incremental training data sets based on the spatiotemporal characteristics of the misjudged scenarios and updates the cloud-based LSTM model parameters through a federated learning framework while ensuring local data privacy.

[0102] Furthermore, when multiple passengers are detected approaching the elevator door at the same time, the TOF sensor and the camera will work together to determine the priority (for example, the passenger closest to the door will be given priority). If multiple passengers meet the entry intention conditions at the same time, the door opening time will be extended to 2T0 to avoid misjudgment and conflict when multiple people enter.

[0103] Specifically:

[0104] 1. Methods for determining passenger intentions:

[0105] 1. Intent classification and judgment conditions:

[0106]

[0107]

[0108] 2. Confidence scoring model:

[0109] Score=α□Sface+β□Shand+γ□Sspeed

[0110] in,

[0111] Sface: facial orientation matching (0-1);

[0112] Shand: the probability of the hand approaching the door edge;

[0113] Sspeed: similarity of speed to the ideal entry / exit curve;

[0114] α, β, γ: dynamic weights (adjusted in real time through LSTM network);

[0115] When the passenger density is high, α (facial orientation weight) is reduced and γ (movement speed weight) is increased to adapt to the scenario of fast passing. When large items are detected, β (hand movement weight) is increased to more accurately judge the entry intention.

[0116] Of course, in some embodiments, an "environmental context factor" (Senv) can be added to evaluate the impact of the environment inside and outside the elevator on the passenger's intention. For example, when a queue is detected outside the elevator, the weight of Senv is increased, and the intention to enter is prioritized. When the elevator is detected to be fully loaded, the weight of Senv is reduced, and the external intention to enter is blocked, so as to improve the accuracy of the model's judgment of complex scenes.

[0117] 3. Logic of forced closure when full

[0118]

[0119] 2. Dynamic Control Strategy

[0120] 1. Dynamic adjustment of door opening time

[0121]

[0122]

[0123] 2. Exception handling mechanism

[0124]

[0125] It should be noted that the working principles of the camera, radar, pressure distribution sensor and TOF sensor involved in the present invention, as well as LSTM network analysis, ICP algorithm, Kalman filter prediction of passenger movement trajectory, etc. are all existing technologies and can be found on the Internet and textbooks. The focus of the present invention is on its application in the special scenario of elevators, and the coordinated and accurate judgment of passengers' intention to take the elevator by multiple sensors, and the intelligent and efficient operation of the elevator based on the intention to take the elevator.

[0126] It should also be noted that in order to prevent user privacy leakage, the technical solution of the present invention also includes privacy and compliance optimization.

[0127] Data anonymization: All sensor data is anonymized, and only key features (such as skeletal key points and movement trajectories) are extracted. The original images or videos are not stored. This is more in line with privacy protection regulations and reduces legal risks.

[0128] Encrypted transmission and storage: A new "data encryption module" uses the AES-256 encryption algorithm to encrypt, transmit and store sensor data in real time, ensuring data security and preventing information leakage.

[0129] Example 1: Peak Scenario in a Commercial Building

[0130] Scene characteristics: From 8:00 to 9:30 in the morning, a large number of passengers are waiting in line on the lower floors, and the elevator is prone to overload.

[0131] Control Flow:

[0132] 1. Data collection:

[0133] The millimeter-wave radar detected five passengers approaching the elevator door at 1.2 m / s;

[0134] The camera recognized that the faces of the four people were facing the door frame and their hands were close to the edge of the door.

[0135] 2. Intention judgment:

[0136] Entry intent confidence score 92% (exceeding threshold 85%);

[0137] The pressure sensor shows that the current load is 80%, allowing entry.

[0138] 3. Dynamic adjustment:

[0139] The door opening time is extended to 1.5×5 seconds=7.5 seconds;

[0140] The group control system dispatches adjacent elevators to the target floor in advance.

[0141] Beneficial effects:

[0142] Improved efficiency: Extend the door opening time when multiple people enter at the same time, reducing the loss of repeated door opening and closing;

[0143] Safety assurance: Early warning before overloading to avoid emergency braking.

[0144] In actual tests, after adopting the present invention, the average waiting time for elevators during the morning peak period was reduced from 120 seconds to 85 seconds, the number of door openings and closings was reduced by 40%, and energy consumption was reduced by 18%.

[0145] Example 2: Hospital emergency scene

[0146] Scenario characteristics: Emergency carts need to move in and out quickly, and traditional systems are delayed due to frequent triggering of safety barriers.

[0147] Control Flow:

[0148] 1. Special identification:

[0149] The TOF sensor detects the cart outline (length > 1.5m, width > 0.8m);

[0150] The pressure distribution shows concentrated forces (which is significantly different from the dispersed forces of standing passengers).

[0151] 2. Strategy switching:

[0152] Disable the safety shutter and extend the door opening time to 10 seconds;

[0153] Voice prompt: "The emergency channel is open, please stand to the side."

[0154] 3. Abnormal monitoring:

[0155] After the cart is fully in, the TOF scan confirms there is no risk of collision;

[0156] Pressure sensors monitor load changes in real time.

[0157] Beneficial effects:

[0158] Rapid response: Accurately identify emergency equipment through contour and pressure characteristics;

[0159] Security first: Disable the baffle but increase the scanning frequency to balance efficiency and risk.

[0160] In actual tests, after adopting the present invention, the passing time of the emergency cart was shortened from 12 seconds of the traditional system to 8 seconds, and the probability of false triggering of the safety baffle was reduced from 35% to 5%.

[0161] Example 3: Collaborative dispatching scenario at a subway station

[0162] Scenario characteristics: Elevators and escalators are adjacent to each other and need to be coordinated and diverted during peak hours.

[0163] Control Flow:

[0164] 1. Data interoperability:

[0165] The escalator control system receives an escalator fault signal;

[0166] The camera detects passengers gathering at the escalator entrance (>10 people).

[0167] 2. Mode switching:

[0168] The elevator switches to "high capacity mode" (door opening time is extended by 50%);

[0169] Electronic signs guide passengers to the elevators.

[0170] 3. Energy efficiency optimization:

[0171] At night, two elevators are combined into a "single elevator loop";

[0172] Idle elevators go into deep sleep (power consumption drops to 5%).

[0173] Beneficial effects:

[0174] Resource integration: Dynamically adjust transport capacity in case of failure to avoid congestion;

[0175] Significant energy saving: Sleep strategy reduces standby energy consumption.

[0176] Example 4: Residential Elderly Care Scenario

[0177] The camera recognizes the slow gait characteristics of the elderly (moving speed <0.3m / s), automatically extends the door opening time to 2.5T0, and triggers a voice reminder "Pay attention to your feet"; after the pressure sensor detects the weight distribution of the wheelchair, the door closing resistance detection function is disabled to avoid false triggering.

[0178] In summary, the system and method provided by the present invention:

[0179] Multimodal sensing layer: cameras, radars, TOF sensors, etc. collaborate to collect passenger behavior data; intention decision layer: generate confidence scores by integrating skeleton key points, movement trajectories, and environmental context; dynamic control layer: adjust door opening and closing strategies based on intention scores and scene features; feedback optimization layer: continuously optimize control logic through reinforcement learning. Core innovations: Rejection of entry intention recognition: combine the passenger's turning action and distance change to determine the behavior of giving up the elevator; full-load forced door closing mechanism: block external entry requests when overloaded, giving priority to safety; multi-level exception handling: progressive response from voice prompts to system cascade control.

[0180] Please note that the technical features of the above embodiments can be combined arbitrarily. In order to make the description concise, all possible combinations of the technical features in the above embodiments are not described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification. The above embodiments only express several implementation methods of the present application, and their descriptions are relatively specific and detailed, but they cannot be understood as limiting the scope of the invention patent. It should be pointed out that for ordinary technicians in this field, without departing from the concept of the present application, several variations and improvements can be made, which all belong to the scope of protection of the present application. Therefore, the scope of protection of the patent in this application shall be based on the attached claims.

[0181] The above are preferred embodiments of the present invention. It should be pointed out that, for ordinary technicians in this technical field, several improvements and modifications can be made without departing from the principles of the present invention. These improvements and modifications should also be regarded as within the scope of protection of the present invention.

Claims

1. An elevator control method, characterized in that: include: Through cameras, radars and TOF sensors, the behavior data of passengers inside and outside the elevator is collected in real time; Based on the behavior data, the passenger intention is judged by a data fusion algorithm, wherein the passenger intention includes the intention to enter the elevator and the intention to leave the elevator; wherein the data fusion algorithm includes: a. Obtain the passenger's facial orientation, hand movements and body orientation information through the camera; b. Obtain the distance and moving speed between the passenger and the elevator door through radar or TOF sensor; c. generating a confidence score of the passenger's intention by combining the facial orientation, the hand movement, the body orientation, the distance from the elevator door, and the moving speed; When the confidence score exceeds a preset threshold, a corresponding elevator door control signal is generated to dynamically adjust the opening or closing action of the elevator door.

2. The elevator control method according to claim 1, characterized in that: The judgment conditions for the intention to enter the elevator include: The passenger faces the elevator door and the duration exceeds the first threshold; The passenger's hand movement conforms to the preset trajectory approaching the edge of the elevator door; The distance between the passenger and the elevator door is less than a second threshold and the moving speed approaches zero.

3. The elevator control method according to claim 2, characterized in that: The judgment conditions for the intention to enter the elevator include: The camera recognizes that the passenger's face is facing the elevator door and the focus of sight is in the door frame area; The TOF sensor detects that the distance between the hand and the door edge is less than the preset distance and lasts for more than the preset time; When the above conditions are met at the same time, a door opening command is generated.

4. The elevator control method according to claim 2, characterized in that: The judgment conditions for the intention to leave the elevator include: The passenger's body is facing out of the elevator door and the duration exceeds the third threshold; The passenger moving speed is greater than a fourth threshold; The distance between the passenger and the elevator door increases over time.

5. The elevator control method according to claim 1, characterized in that: The determining of the passenger intention by using the data fusion algorithm includes: Extract skeleton key points from camera images to obtain coordinates of multiple body joint points; Align the radar point cloud data with the joint point coordinates in time and space; A Kalman filter is used to predict passenger movement trajectories and generate intention confidence scores.

6. The elevator control method according to claim 5, characterized in that: The spatiotemporal alignment includes: Establish a three-dimensional coordinate system with the center of the elevator door as the origin; Perform clustering on radar point cloud data to extract passenger profile feature points; The feature point cloud is registered with the image joint point cloud using the ICP algorithm.

7. The elevator control method according to claim 5, characterized in that: The spatiotemporal alignment includes: Analyze action sequences of more than 5 consecutive frames through the LSTM network to identify typical entry and exit patterns; The typical entry and exit modes include three preset postures: holding the door with one hand, sideways avoidance, and dragging luggage.

8. The elevator control method according to claim 1, characterized in that: It also includes exception handling mechanisms: When the radar detects a stationary obstacle in the door area for more than 10 seconds, the TOF sensor high-precision scanning mode is activated; If it is confirmed that non-passenger items are left behind, a voice prompt will be triggered and the door control operation will be suspended.

9. The elevator control method according to claim 1, characterized in that: The elevator control method combines sensor data and historical data of passenger intention judgment, and uses a reinforcement learning method to optimize the elevator door opening timing.

10. An elevator control system, characterized in that: The system comprises: Multiple sensor modules for acquiring passenger data inside and outside the elevator; A data processing module for analyzing and fusing sensor data to determine a passenger's intention to enter or leave; and, The control module is used to generate an elevator door control signal according to the judgment result.

Images