A control method and device for an embodied humanoid robot for an elevator

Abstract

A control method and apparatus for an embodied humanoid robot used in an elevator. The method involves: constructing a permission access vector; generating a target floor scheduling command based on the permission access vector; determining the three-dimensional physical coordinates of the target floor call control according to the target floor scheduling command; generating a safety release signal based on real-time elevator operating parameters and environmental status monitoring data; generating a pose approximation trajectory based on the safety release signal and the three-dimensional physical coordinates; generating physical contact feedback torque data based on the pose approximation trajectory; determining the pressing state based on the physical contact feedback torque data; and reversing the driving current waveform to generate a reverse driving current waveform, causing the end effector to move to the initial standby position and generating an end effector disengagement and reset trajectory. Implementing the technical solution provided in this application enhances the multi-dimensional safety perception and handling capabilities of the elevator's autonomous operation.

Description

Technical Field

[0001] This application relates to the field of embodied intelligent robot technology, specifically to a control method and device for an embodied humanoid robot used in an elevator. Background Technology

[0002] Construction hoists, as core vertical transportation equipment on construction sites, have traditionally relied on certified operators for manual operation. With the development of automation technology, electrical control retrofitting and remote control solutions have emerged in the industry. These solutions attempt to reduce reliance on human operators by modifying the original control circuits of the hoist or adding remote control modules. Some solutions also introduce simple automatic leveling functions, achieving basic floor stopping through preset limit switches.

[0003] However, existing technical solutions for achieving autonomous operation of construction hoists generally suffer from insufficient safety monitoring dimensions and inadequate control precision. Electrical control retrofitting solutions are difficult to integrate with different models of older equipment and lack real-time sensing capabilities for multi-dimensional safety conditions related to personnel, load, and environment. Remote control solutions rely on communication links, which cannot guarantee equipment safety when signals are interrupted. Simple automatic leveling solutions have low stopping accuracy and cannot effectively identify and handle abnormal situations such as overloading, overcrowding, and unauthorized operations. These shortcomings result in significant safety hazards remaining during the autonomous operation of construction hoists. Summary of the Invention

[0004] To address the aforementioned problems, this application provides a control method and apparatus for a humanoid robot used in an elevator.

[0005] The first aspect of this application provides a control method for a humanoid robot used in an elevator, employing the following technical solution: Acquire occupant feature data and layer call configuration data, and construct a permission access vector based on the occupant feature data and a preset authorization map; Based on the permitted access vector and the layer call configuration data, a security conflict detection calculation is performed to generate a target floor scheduling instruction; According to the target floor dispatch instruction, obtain the control panel image data, and perform image feature extraction based on the control panel image data to determine the three-dimensional spatial physical coordinates of the target floor call control. The system acquires real-time operating parameters of the elevator and environmental status monitoring data, performs a fusion threshold comparison based on the real-time operating parameters of the elevator and the environmental status monitoring data, and generates a safe release signal when the comparison result meets the preset safety conditions. The reverse motion joint angle is calculated based on the safety release signal and the three-dimensional spatial physical coordinates, and the pose approximation trajectory of the end effector of the robot's bionic arm is generated based on the reverse motion joint angle. Based on the pose approximation trajectory, a drive current waveform is output, and the transmission component is adjusted based on the drive current waveform to make the execution end press the target layer call control, thereby generating physical contact feedback torque data. The pressing state is determined based on the physical contact feedback torque data. When the pressing state is determined to be successful, the driving current waveform is reverse modulated to generate a reverse driving current waveform. Based on the reverse driving current waveform, the transmission component is driven to move the execution end to the preset initial standby position, generating the end to disengage from the reset trajectory.

[0006] By adopting the above technical solutions, the target control is located by extracting image features and comparing the operation parameters and environmental data in real time, which improves the stopping accuracy and anomaly recognition capability. Based on the reverse motion student to generate the bionic arm trajectory, combined with physical feedback, precise pressing and reset is achieved, which eliminates communication dependence and enhances compatibility and control accuracy, thereby improving the multi-dimensional safety perception and handling capability of the elevator for autonomous operation.

[0007] Optionally, the step of performing security conflict detection calculations based on the permitted access vector and the layer call configuration data to generate a target floor scheduling instruction includes: The floor stop record matrix contained in the floor call configuration data is logically matched with the permitted access vector to extract a conflict-free candidate floor sequence. The conflict-free candidate floor sequence is traversed and calculated based on a preset scheduling distance evaluation function to determine the minimum distance cost feature; Based on the minimum distance cost feature, the corresponding target layer height distribution parameters are extracted, and the target layer height distribution parameters are encapsulated to generate the target layer scheduling instruction.

[0008] By adopting the above technical solutions, access permission conflicts and the risk of misoperation are effectively avoided; by combining the scheduling distance evaluation function to traverse and calculate the minimum distance cost characteristics, the path optimization and precise positioning of the target floor are realized, improving scheduling efficiency and docking accuracy; by encapsulating the target floor height distribution parameters to generate scheduling instructions, the compatibility and execution reliability of the instructions are enhanced, further ensuring the safety and control accuracy of the elevator's autonomous operation.

[0009] Optionally, the step of extracting image features based on the control panel image data to determine the three-dimensional physical coordinates of the target layer control includes: The control panel image data is subjected to grayscale conversion and Gaussian filtering for noise reduction to generate a smooth feature image matrix; Extract the target floor identifier characters contained in the target floor scheduling instruction, perform morphological contour matching calculation on the smooth feature image matrix based on the target floor identifier characters, and extract the two-dimensional plane pixel coordinates of the geometric centroid of the target closed connected domain; The depth measurement data of the control panel is synchronously acquired from the vision sensor. The pre-stored camera intrinsic parameter matrix and distortion correction parameters are retrieved to form an imaging model calibration parameter matrix. Based on the imaging model calibration parameter matrix, the two-dimensional plane pixel coordinates and the depth measurement data are calculated by inverse perspective projection mapping to obtain the three-dimensional spatial physical coordinates of the target layer control panel.

[0010] By adopting the above technical solutions, uneven lighting and image noise interference are effectively suppressed, and the robustness and recognition accuracy of feature extraction are improved. Based on the morphological contour matching of the target floor identification characters, the two-dimensional pixel coordinates of the target control are accurately located. Combined with depth measurement data and camera calibration parameters, inverse perspective projection mapping is performed, realizing high-precision calculation of three-dimensional spatial physical coordinates, providing a reliable spatial positioning basis for the precise pressing operation of the bionic arm.

[0011] Optionally, the step of acquiring real-time operating parameters of the elevator and environmental status monitoring data, performing a fusion threshold comparison based on the real-time operating parameters of the elevator and the environmental status monitoring data, and generating a safety release signal when the comparison result meets preset safety conditions includes: Extract the cage door closing status signal, landing door interlock status signal, and load weight measurement value from the real-time operating parameters of the elevator. Perform a logical AND operation on the cage door closing status signal and the landing door interlock status signal to obtain the door safety verification result. The load weight measurement value is compared with a preset overload threshold to generate a load compliance determination mark. Extract the cage tilt angle measurement value, wind speed measurement value, and obstacle distance measurement value from the environmental status monitoring data. Compare and calculate the cage tilt angle measurement value, wind speed measurement value, and obstacle distance measurement value with the corresponding preset safety thresholds to generate an environmental risk assessment vector. Based on the door safety verification result, the load compliance judgment identifier, and the environmental risk assessment vector, a multi-dimensional logical judgment is performed. When the result of the multi-dimensional logical judgment meets the safety conditions, the safety release signal is generated.

[0012] By adopting the above technical solutions, the risk of overload operation is effectively prevented; by integrating multi-dimensional environmental parameters such as cage tilt angle, wind speed and obstacle distance for threshold comparison, comprehensive perception of overturning, strong wind and collision hazards is achieved; and by generating a safe release signal based on multi-dimensional logical judgment of the door, load and environment, a reliable safety decision basis is provided for the autonomous operation of the elevator.

[0013] Optionally, the step of comparing the measured load weight with a preset overload threshold to generate a load compliance determination identifier includes: Determine whether the measured load weight value is within a preset valid value range. If the measured load weight value is within the preset valid value range, mark it as valid load data. Obtain the rated load parameters stored in the preset elevator equipment configuration database, multiply the rated load parameters with the preset safety factor, and determine the preset overload threshold. Calculate the numerical difference between the effective load data and the preset overload threshold. When the numerical difference is less than zero, set the load compliance judgment identifier to a compliance status code. When the numerical difference is greater than or equal to zero, set the load compliance judgment identifier to an overload status code.

[0014] By adopting the above technical solutions, the validity of load measurement values ​​is verified, and invalid data generated by sensor malfunctions or communication errors are eliminated, thereby improving the reliability of load monitoring. The overload threshold is dynamically determined based on the product of the rated load parameter and the safety factor, which adapts to the safety requirements of different equipment models and operating conditions. By calculating the difference between the effective load data and the threshold, a compliance or overload status code is generated, further enhancing the accuracy and real-time performance of overload risk prevention and control.

[0015] Optionally, the step of calculating the inverse motion joint angle based on the safety release signal and the three-dimensional spatial physical coordinates, and generating the pose approximation trajectory of the end effector of the robot's bionic arm based on the inverse motion joint angle, includes: After receiving the safety release signal, the current angle encoder values ​​of each joint in the bionic arm are obtained, and the current angle encoder values ​​are arranged according to the preset sequence number of each joint to form the current joint angle vector. Obtain the length parameter values ​​of each link in the bionic arm, the maximum rotation angle value of each joint, and the minimum rotation angle value, and establish a positive kinematic calculation formula between the spatial position of each joint and the end effector based on the length parameter values; The three-dimensional spatial physical coordinates are used as the target end position and input into the forward kinematics calculation formula. The reverse motion joint angle that satisfies the constraints of the maximum rotation angle value and the minimum rotation angle value is obtained through numerical iteration. Calculate the joint angle differences between the current joint angle vector and the reverse motion joint angle, and perform polynomial function fitting on the joint angle differences based on preset motion time parameters to generate the pose approximation trajectory.

[0016] By adopting the above technical solution, a forward kinematics formula is established based on the linkage parameters and joint constraints. Combined with numerical iteration to solve the inverse motion joint angles, the target position is accurately mapped to each joint space under physical constraints. By calculating the angle difference and performing polynomial fitting, a smooth and continuous pose approximation trajectory is generated, which takes into account both motion efficiency and mechanical execution stability, providing a high-precision motion control foundation for the precise pressing operation of the bionic arm.

[0017] Optionally, the step of determining the pressing state based on the physical contact feedback torque data, and when the pressing state is determined to be successful, reverse-modulating the drive current waveform to generate a reverse drive current waveform, and driving the transmission component based on the reverse drive current waveform to move the execution end to a preset initial standby position, generating an end-effector retraction trajectory, includes: The physical contact feedback torque data is subjected to time-domain sliding window filtering to generate a smooth torque feature sequence. Peak detection calculation is performed on the smooth torque feature sequence to extract the peak torque value and the peak occurrence time. The peak torque value is compared with a preset successful press threshold. When the peak torque value is greater than the preset successful press threshold and the time difference between the peak occurrence time and the press start time is less than a preset response time threshold, the press state determination is marked as a successful press state code. After receiving the press success status code, a negative mapping transformation is performed on each amplitude component in the amplitude parameter matrix of the drive current waveform, while keeping the frequency parameter matrix in the drive current waveform unchanged, to generate the reverse drive current waveform. The transmission component is driven based on the reverse drive current waveform, and the movement of the execution end is controlled according to the preset multi-segment speed planning curve. When the distance between the execution end and the target layer call control reaches the preset safe distance threshold, the movement stops and the initial standby position is maintained, generating the end-reset trajectory.

[0018] By adopting the above technical solution, the physical contact feedback torque is filtered and peak detected to accurately identify the peak force and response time at the moment of pressing. Combined with threshold comparison, the accuracy and real-time performance of pressing success determination are ensured. After receiving the successful encoding, the reverse motion is smoothly switched by performing a negative mapping transformation on the driving current amplitude while keeping the frequency unchanged. Based on the reverse current waveform and multi-segment speed planning control, the execution end is reset to a safe position, generating a stable and reliable disengagement and reset trajectory, effectively avoiding the risk of secondary collisions and improving the mechanical safety and control accuracy of the operation process.

[0019] A second aspect of this application provides an electronic device including a processor, a memory, a user interface, and a network interface, wherein the memory is used to store instructions, the user interface and the network interface are both used to communicate with other devices, and the processor is used to execute the instructions stored in the memory to cause the electronic device to perform the method as described in any of the foregoing.

[0020] A third aspect of this application provides a computer-readable storage medium storing instructions that, when executed, perform the method described in any of the preceding descriptions.

[0021] A fourth aspect of this application provides a computer program product that, when run on an electronic device, causes the electronic device to perform the method as described in any of the preceding claims.

[0022] In summary, one or more technical solutions provided in the embodiments of this application have at least the following technical effects or advantages: By constructing permission access vectors and a security conflict detection mechanism, precise matching of permission management and scheduling instructions is achieved, effectively avoiding the risk of misoperation and improving docking efficiency. By using image feature extraction and depth data fusion to calculate the three-dimensional spatial coordinates of the target control, and combining the fusion threshold comparison of multi-dimensional operating parameters and environmental monitoring data, the ability to identify abnormal working conditions and docking accuracy are enhanced. Based on inverse kinematics, the pose approximation trajectory of the bionic arm is generated, and precise pressing and smooth reset are achieved through physical contact feedback. This overcomes the shortcomings of traditional solutions, such as strong communication dependence, poor compatibility, and low control accuracy. It realizes the synergistic improvement of multi-dimensional safety perception and mechanical execution stability during the autonomous operation of the elevator, providing technical support for unmanned vertical transportation on construction sites. Attached Figure Description

[0023] Figure 1 This is a schematic diagram of the system architecture of an embodiment of a control method for an embodied humanoid robot used in an elevator, according to the present application. Figure 2 This is a flowchart illustrating a control method for a humanoid robot used in an elevator, as disclosed in an embodiment of this application. Figure 3 This is a schematic diagram of the structure of an electronic device disclosed in an embodiment of this application.

[0024] Explanation of reference numerals in the attached figures: 100, System architecture; 101, First terminal device; 102, Second terminal device; 103, Third terminal device; 104, Network; 105, Server; 301, Processor; 302, Communication bus; 303, User interface; 304, Network interface; 305, Memory. Detailed Implementation

[0025] To enable those skilled in the art to better understand the technical solutions in this specification, the technical solutions in the embodiments of this specification will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments.

[0026] In the description of the embodiments of this application, the words "for example" or "for instance" are used to indicate examples, illustrations, or explanations. Any embodiment or design that is described as "for example" or "for instance" in the embodiments of this application should not be construed as being more preferred or advantageous than other embodiments or design options. Rather, the use of the words "for example" or "for instance" is intended to present the relevant concepts in a specific manner.

[0027] In the description of the embodiments of this application, the term "multiple" means two or more. For example, multiple systems means two or more systems, and multiple screen terminals means two or more screen terminals. Furthermore, the terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the indicated technical features. Thus, a feature defined with "first" or "second" may explicitly or implicitly include one or more of that feature. The terms "comprising," "including," "having," and variations thereof all mean "including but not limited to," unless otherwise specifically emphasized.

[0028] like Figure 1 As shown, system architecture 100 may include terminal devices 101, 102, and 103, a network 104, and a server 105. Network 104 serves as the medium for providing communication links between terminal devices 101, 102, and 103 and server 105. Network 104 may include various connection types, such as wired or wireless communication links, or fiber optic cables, etc.

[0029] Users can use terminal devices 101, 102, and 103 to interact with server 105 via network 104 to receive or send messages, etc. Various communication client applications can be installed on terminal devices 101, 102, and 103, such as model training applications, video recognition applications, web browser applications, social platform software, etc.

[0030] Terminal devices 101, 102, and 103 can be either hardware or software. When terminal devices 101, 102, and 103 are hardware, they can be various electronic devices with displays, including but not limited to smartphones, tablets, e-book readers, MP3 (Moving Picture Experts Group Audio Layer III) players, MP4 (Moving Picture Experts Group Audio Layer IV) players, laptops, and desktop computers, etc. When terminal devices 101, 102, and 103 are software, they can be installed in the aforementioned electronic devices. They can be implemented as multiple software programs or software modules (e.g., multiple software programs or software modules used to provide distributed services) or as a single software program or software module. No specific limitations are imposed here.

[0031] This embodiment discloses a control method for a humanoid robot used in an elevator. Figure 2 This is a flowchart illustrating a control method for a humanoid robot used in an elevator, as disclosed in an embodiment of this application. Figure 2 As shown, the method includes the following steps: S201. Obtain occupant feature data and layer call configuration data, and construct a permission access vector based on the occupant feature data and a preset authorization map; Specifically, to acquire occupant feature data, the system can be implemented in several ways: The first approach is to use a facial recognition camera installed inside the cage or on the floor call button to capture real-time facial images of occupants entering the detection range. Facial feature vectors are extracted using a deep learning model (e.g., FaceNet or ArcFace) built into the edge computing host, and compared with a pre-entered personnel database to output a unique occupant identifier (e.g., employee ID or visitor ID). The second approach is for occupants to use an authorized IC / ID card or NFC-enabled mobile phone to swipe their card in a designated card-reading area. The system obtains the card's unique serial number as an identifier through an RFID / NFC reader. Simultaneously, the system receives one or more target floor information selected by the occupant via a floor call button (either with physical buttons or a touchscreen), and binds these floor numbers to the occupant identifier to form structured floor call configuration data, such as a data object containing {UserID: "EMP007", RequestedFloors: [8, 15]}. Next, the system's embodied intelligent decision-making unit, based on the acquired occupant identification, queries a preset authorization map in local storage or accesses the cloud management platform via a network interface. This map is essentially a multi-dimensional access control list, which can be a relational database table or a JSON / YAML file, defining in detail the permission attributes corresponding to each identification, such as {UserID: "EMP007", Group: "Technician", AllowedFloors: "1-10, 15", TimePolicy: "Weekdays08:00-18:00"}. The decision-making unit compares and verifies the queried permission rules (such as allowed floor ranges and valid time periods) with the current system time, ultimately generating a specific, temporary authorized access vector. This vector can be a Boolean array, where the array index represents the floor number, with a value of true indicating that the floor is authorized for access, and false otherwise; or it can be a set containing all currently accessible floor numbers (e.g., {1, 2, ..., 10, 15}). This vector accurately depicts all the legitimate access rights that the occupant possesses at the current moment, providing a decisive and tamper-proof input basis for subsequent safety scheduling and conflict detection.

[0032] S202. Based on the permitted access vector and the layer call configuration data, perform security conflict detection calculations to generate target floor scheduling instructions; In a preferred embodiment of this application, step S202 serves as the core decision-making step for intelligent scheduling. Specifically, a security check is first performed on permissions: a logical intersection operation is performed on all floors requested by passengers in the floor call configuration data (e.g., the set {8, 15}) and the permitted access vector (e.g., the set {1, 2, ..., 10, 15} representing accessible floors). Through this operation, the system can quickly filter out floors where all permissions and requests are satisfied, forming a conflict-free candidate floor sequence (e.g., the sequence [8, 15]). For any request with conflicting permissions (e.g., a request for the unauthorized floor 12), the system will automatically remove it from the candidate sequence and prompt the user through a human-computer interaction unit (e.g., voice broadcast) that floor 12 is not accessible. After ensuring that all candidate floors are compliant, the system initiates path optimization calculation. One feasible implementation is that the system calls a preset scheduling distance evaluation function. This function comprehensively considers the current location of the elevator, its current direction of travel, and the locations of each candidate floor. Then, it iterates through the floors to determine the stopping order based on the minimum distance cost characteristic, which minimizes the total travel distance or the total time. For example, if the elevator is on the 5th floor and moving upwards, the system will prioritize scheduling it to stop at the 8th floor before proceeding to the 15th floor. Finally, the system encapsulates this optimized ordered list of floors ([8, 15]) along with the pre-stored physical height and other distribution parameters of each floor to generate a standardized target floor scheduling instruction that can be directly parsed and executed by subsequent modules. This could be a JSON object with the content {"schedule_order": [8, 15], "task_id": "..."}.

[0033] Optionally, the step of performing security conflict detection calculation based on the permitted access vector and the layer call configuration data to generate a target floor scheduling instruction includes: performing a logical matching operation between the floor docking record matrix contained in the layer call configuration data and the permitted access vector to extract a conflict-free candidate floor sequence; traversing and calculating the conflict-free candidate floor sequence based on a preset scheduling distance evaluation function to determine the minimum distance cost feature; extracting the corresponding target floor height distribution parameters according to the minimum distance cost feature, and encapsulating the target floor height distribution parameters to generate the target floor scheduling instruction.

[0034] Specifically, the system performs a logical matching operation to extract conflict-free candidate floors. In one implementation, the floor call configuration data can be formatted as a set containing all currently pending floor numbers, i.e., a floor docking record matrix. Simultaneously, the generated permission access vector is also represented as a set containing all authorized floor numbers. The logical matching operation performed by the system is the intersection operation of these two sets. The result of the operation is a new set containing all floors that have been requested by the user and are within the user's permission scope; this result is the conflict-free candidate floor sequence. As an alternative implementation, if the permission access vector is a boolean array (where the index corresponds to the floor and the value indicates whether it is authorized), the system can iterate through each requested floor number in the floor docking record matrix and query the boolean array using that number as an index. Only when the query result is true is the floor added to an initially empty conflict-free candidate floor sequence list. Through this step, the system can automatically filter out all unauthorized elevator requests with extremely high efficiency and accuracy, ensuring that all subsequent scheduling operations are performed within a safe and compliant scope.

[0035] Furthermore, the system optimizes the path of candidate floors based on a preset scheduling distance evaluation function. Specifically, this preset scheduling distance evaluation function is a core scheduling algorithm module designed to maximize the operating efficiency of the elevator while satisfying all compliant requests. A preferred approach is to use the classic elevator scheduling SCAN algorithm. This function takes the elevator's current position, current direction of travel (up / down / stationary), and the generated sequence of conflict-free candidate floors as input. Then, it prioritizes serving all requested floors in the current direction of travel until the furthest request in that direction is satisfied, before turning around to serve the remaining requests. For example, if the elevator is preparing to go up on the 5th floor and the candidate sequence is {2, 9, 12}, the evaluation function will generate a stopping order of [9, 12, 2]. As another more complex implementation, the evaluation function can also be a cost-based model that considers not only the running distance, but also factors such as waiting time on each floor, number of passengers, and power consumption. By performing full permutations or heuristic search on candidate sequences (such as genetic algorithms or simulated annealing algorithms), the comprehensive cost of each stopping order is dynamically calculated, and finally the path with the lowest cost is selected as the optimal solution with the minimum distance cost feature.

[0036] Furthermore, the system generates executable scheduling instructions based on the determined optimal path. After determining the optimal stopping sequence with the minimum distance cost, the decision unit extracts the precise physical location parameters corresponding to each target floor from the pre-stored elevator configuration database, i.e., the target floor height distribution parameters. These parameters may include, but are not limited to: the absolute height value relative to the reference plane (unit: meters), the coordinates of the UWB positioning tag, the precise reading range of the encoder or magnetic ruler, etc. Subsequently, the system encapsulates this ordered list of floors and its corresponding physical parameters according to a predefined protocol format to form a structured and standardized target floor-to-floor scheduling instruction. This instruction is usually represented as a JSON or XML data packet, for example, {"task_id": "T1688...", "schedule": [{"floor": 9, "position_code": 18500}, {"floor": 12, "position_code": 24500}, ...]}. This instruction contains all the precise information required to execute the task, and can be directly parsed and executed by the robot's motion control module, thus ensuring the accuracy and reliability of the entire autonomous driving process.

[0037] S203. According to the target floor scheduling instruction, obtain the control panel image data, and extract image features based on the control panel image data to determine the three-dimensional spatial physical coordinates of the target floor control panel. Specifically, upon receiving a target floor scheduling instruction (e.g., floor 8), the robot's motion control system first drives its arm or body to a pre-set global observation position where the elevator's control panel can be clearly observed. Subsequently, a 3D vision sensor (e.g., an RGB-D depth camera) mounted on the robot's end effector or head is activated to capture one or more frames of high-resolution control panel image data, which includes both color information (RGB image) and pixel-level depth information (depth map). Next, the system's image processing unit extracts image features from the acquired RGB image. A preferred implementation is that the system loads a pre-trained deep learning object detection model (e.g., YOLOv5 or SSD) that can recognize elevator buttons in the image in real time and accurately output the bounding box and center pixel coordinates (u, v) of the number "8" button in the 2D image based on the target floor "8" in the scheduling instruction. Alternatively, the system can first detect all circular or square button outlines using traditional image processing methods such as Hough transform, and then perform optical character recognition on the regions within each outline to find the button matching the target floor number. After successfully locating the 2D pixel coordinates (u, v) of the target button, the system immediately queries the corresponding depth map to obtain its precise depth value d (i.e., the straight-line distance from the camera to the button) at the same coordinates (u, v). Finally, using the camera's pre-calibrated intrinsic parameter matrix, the system calculates and converts the 2D pixel coordinates (u, v) and depth value d into three-dimensional physical coordinates (X, Y, Z) in the robot's own coordinate system using a standard coordinate system transformation formula (e.g., X=(u-cx)×d / fx). This three-dimensional coordinate system accurately describes the target button's position in physical space, providing the necessary target point for subsequent precise touch operations by the robotic arm.

[0038] Optionally, the step of extracting image features based on the control panel image data to determine the three-dimensional physical coordinates of the target floor call control includes: performing grayscale conversion and Gaussian filtering noise reduction on the control panel image data to generate a smooth feature image matrix; extracting the target floor identifier characters contained in the target floor scheduling instruction, performing morphological contour matching calculation on the smooth feature image matrix based on the target floor identifier characters, and extracting the two-dimensional plane pixel coordinates of the geometric centroid of the target closed connected domain; synchronously acquiring the depth measurement data of the control panel from the visual sensor, retrieving the pre-stored camera intrinsic parameter matrix and distortion correction parameters to form an imaging model calibration parameter matrix, and performing inverse perspective projection mapping calculation on the two-dimensional plane pixel coordinates and the depth measurement data based on the imaging model calibration parameter matrix to obtain the three-dimensional physical coordinates of the target floor call control.

[0039] Specifically, to improve the accuracy and robustness of subsequent recognition, the system preprocesses the original control panel image data. In one implementation, the system first receives the raw RGB color image from a vision sensor (e.g., an RGB-D camera) and calls the `cvtColor` function from an image processing library (such as OpenCV) to convert it from a three-channel color space to a single-channel grayscale image. This conversion aims to eliminate interference from color information and significantly reduce the complexity of subsequent calculations. Next, to suppress potential sensor noise or artifacts caused by uneven lighting, the system applies Gaussian filtering to the grayscale image for noise reduction. Specifically, the system uses a Gaussian kernel of a preset size (e.g., 5x5) to perform convolution operations with the image, thereby weighting the pixel values ​​of each pixel to make the overall image smoother, which is beneficial for subsequent contour extraction. The final output of this step is a two-dimensional smooth feature image matrix, laying a solid foundation for accurate feature matching.

[0040] Furthermore, the system performs precise two-dimensional planar localization of the target control on the preprocessed image. Specifically, the system parses and extracts the target floor identifier character, such as the number "8" or the letter "B1," from the target floor scheduling instruction generated in the previous step. Then, the system uses the standardized glyph image of this character as a template to perform morphological contour matching calculations on the smoothed feature image matrix. A preferred implementation is that the system first uses the Canny edge detection algorithm to extract all potential contours in the image, and then applies a contour matching algorithm (such as the matchShapes function) to each contour, comparing its shape with the contour of the template character. When the similarity score of a contour is higher than a preset threshold, the contour is confirmed as the target control. As another optional solution, the system can also use template matching technology to perform a sliding window search on the entire smoothed feature image to find the region with the highest relevance. Once a match is successful, the system calculates the geometric centroid of the closed connected region of the target (i.e., the matched button contour region), thereby obtaining a two-dimensional planar pixel coordinate accurate to the sub-pixel level (e.g., (u=325.4, v=240.1)).

[0041] Furthermore, the system upgrades the pixel coordinates of the two-dimensional plane to precise coordinates in three-dimensional physical space. During this process, the system synchronously acquires a frame of depth measurement data (i.e., a depth map) strictly aligned with the RGB image from the vision sensor. This depth map records the straight-line distance from each pixel in the scene to the camera. Simultaneously, the system retrieves pre-stored camera intrinsic parameter matrices and distortion correction parameters from local storage. These two parameters are precisely calculated during camera manufacturing or installation using calibration procedures (such as checkerboard calibration), and together they constitute the camera's imaging model calibration parameter matrix. Specifically, the system first uses the distortion correction parameters to correct the two-dimensional pixel coordinates (u, v) obtained in the previous step to eliminate errors caused by lens distortion. Then, the system queries the depth map for the corresponding depth value d based on the corrected coordinates. Finally, the system applies mathematical formulas for inverse perspective projection mapping (e.g., X=(u_corrected-cx)×d / fx and Y=(v_corrected-cy)×d / fy, where fx, fy, cx, and cy are key parameters of the intrinsic parameter matrix) to convert two-dimensional image points (u_corrected, v_corrected) and depth d into three-dimensional spatial points (X, Y, Z) in the camera coordinate system. These three-dimensional coordinate points are the three-dimensional physical coordinates of the target layer control in the physical world, which can be directly positioned and manipulated by the robotic arm.

[0042] S204. Obtain the real-time operating parameters of the elevator and the environmental status monitoring data, perform a fusion threshold comparison based on the real-time operating parameters of the elevator and the environmental status monitoring data, and generate a safe release signal when the comparison result meets the preset safety conditions. Specifically, before the robot extends its robotic arm to perform button operations, its main control system acquires data required for safety decisions from two dimensions in parallel. On one hand, the robot continuously acquires real-time operating parameters of the elevator through its communication interface (e.g., CAN bus, RS485, or wireless IoT module), such as the current floor of the elevator, the status of the car door (fully open, closed, etc.), the operating status (stationary, moving), and whether there are any fault alarms. On the other hand, the robot's own sensor array (e.g., LiDAR, IMU (Inertial Measurement Unit), depth camera) collects environmental status monitoring data in real time, such as whether there are obstacles on the robotic arm's planned movement trajectory and whether the robot itself is in a stable posture. Subsequently, the system performs a fusion threshold comparison, rigorously verifying these real-time acquired multi-source data against a set of preset safety conditions. A preferred implementation is that the safety condition is defined as a logical AND gate circuit, which must simultaneously satisfy: [Elevator door status == fully open] AND [Elevator operating status == stationary] AND [Obstacle monitoring result == none] AND [Robot posture tilt angle < preset threshold]. Only when all conditions are met—that is, when the comparison results meet the preset safety conditions—will the system's safety decision module generate a Boolean value of true for safe passage and transmit it to the motion control module, authorizing it to perform subsequent physical contact operations. If any condition is not met, the system will remain in standby mode and continuously monitor until all safety conditions are met, thereby eliminating any risk of collision due to unexpected movement of the elevator or changes in the environment.

[0043] Optionally, the step of acquiring real-time operating parameters of the elevator and environmental status monitoring data, and performing a fusion threshold comparison based on the real-time operating parameters of the elevator and the environmental status monitoring data, and generating a safety release signal when the comparison result meets the preset safety conditions, includes: extracting the cage door closure status signal, the landing door interlock status signal, and the load weight measurement value from the real-time operating parameters of the elevator; performing a logical AND operation on the cage door closure status signal and the landing door interlock status signal to obtain the door safety verification result; performing a numerical comparison operation on the load weight measurement value with a preset overload threshold to generate a load compliance judgment identifier; extracting the cage tilt angle measurement value, wind speed measurement value, and obstacle distance measurement value from the environmental status monitoring data; comparing the cage tilt angle measurement value, the wind speed measurement value, and the obstacle distance measurement value with the corresponding preset safety thresholds to generate an environmental risk assessment vector; performing a multi-dimensional logical judgment based on the door safety verification result, the load compliance judgment identifier, and the environmental risk assessment vector, and generating the safety release signal when the result of the multi-dimensional logical judgment meets the safety conditions.

[0044] Specifically, the system extracts and performs preliminary verification of key safety parameters of the elevator body. In one specific implementation, the robot's main control system actively requests or passively receives real-time operating parameters of the elevator through its communication interface (such as a CAN bus or RS485 interface) connected to the elevator control system. The system extracts two core Boolean state signals from this data stream: the cage door closure status signal and the landing door interlock status signal, where "1" represents closure / interlock and "0" represents opening / unlocking. Subsequently, the system performs a logical AND operation on these two signals. The result of this logical AND operation is true only when the cage door itself is fully closed (signal 1) and its mechanical and electrical interlock with the landing door of the current floor is also active (signal 1). This result is stored as a temporary door safety verification result (e.g., a Boolean variable isDoorSafe), providing the first layer of safety assurance regarding the physical state of the door for subsequent decisions, aiming to eliminate the risk of accidental movement or clamping when the door is not fully closed. In this step, the system will also simultaneously extract the load weight measurement value (e.g., a floating-point number in kilograms) for use in the next step.

[0045] Furthermore, the system performs a compliance review of the elevator's load status. The system compares the measured load weight (e.g., 750.5 kg) obtained in the previous step with a preset overload threshold (e.g., 1000 kg, typically provided by the elevator manufacturer) stored in the system configuration. Specifically, the system determines whether the current load weight is less than or equal to the overload threshold. If the condition is met, the system generates a compliance flag (e.g., a Boolean value of true or an integer 1); conversely, if the load weight exceeds the threshold, it generates an non-compliance flag (e.g., a Boolean value of false or an integer 0). As an optional implementation, in addition to overload, the system can also set a light load threshold to determine whether personnel are entering or leaving, thus assisting in determining the timing of operations. The purpose of this step is to prevent the robot from attempting to execute invalid floor button operations when the elevator refuses to operate due to overload or triggers the safety protection mechanism.

[0046] Furthermore, the system performs a comprehensive real-time risk assessment of the robot's external environment. The system extracts key environmental status monitoring data from its integrated multiple sensors. A preferred implementation involves obtaining the cage tilt angle measurement from the inertial measurement unit (to determine if the elevator is stable and stationary), wind speed measurement from a digital anemometer installed externally to the robot (especially important in outdoor or ventilated environments), and calculating the nearest obstacle distance measurement from point cloud data obtained by scanning the predetermined motion path using a 3D vision sensor or LiDAR. Subsequently, the system compares these three real-time measurements with corresponding preset safety thresholds. For example, it determines whether the tilt angle is less than 2 degrees, the wind speed is less than 5 m / s, and the obstacle distance is greater than 0.5 meters. Each comparison result (true or false) constitutes an environmental risk assessment vector, such as [true, true, true], which intuitively reflects whether the current external environment is suitable for precise robotic arm operations.

[0047] Furthermore, based on all the aforementioned verification results, the system executes the final, comprehensive safety decision. The system gathers all Boolean values ​​in the "Door Safety Verification Result" (isDoorSafe), "Load Compliance Judgment Identifier" (isLoadOk), and "Environmental Risk Assessment Vector" (isEnvSafe) and performs "multi-dimensional logical judgment" on them. The most common implementation is that the system performs a total logical "AND" operation, i.e., isDoorSafe AND isLoadOk AND isEnvSafe[0] AND isEnvSafe[1] AND .... Only when the results of all verification items are "true", i.e. "the results of the multi-dimensional logical judgment meet the safety conditions", will the system safety module finally generate the safety release signal. This signal can be a specific instruction code or a semaphore set to the allowed state, which is sent to the robot's motion control subsystem as the final authorization to perform subsequent physical interaction actions. As a more complex judgment scheme, the system can adopt a weighted scoring mechanism to assign different weights to risks in different dimensions. As long as the total risk score is lower than the preset threshold, it can be released, thereby providing more flexible adaptability. This final, deliberate decision-making step ensures that every step of the robot's operation is protected by multiple safety redundancies.

[0048] Optionally, the step of comparing the measured load weight with a preset overload threshold to generate a load compliance determination identifier includes: determining whether the measured load weight is within a preset valid value range; when the measured load weight is within the preset valid value range, marking it as valid load data; obtaining the rated load parameters stored in a preset elevator equipment configuration database; multiplying the rated load parameters with a preset safety factor to determine the preset overload threshold; calculating the numerical difference between the valid load data and the preset overload threshold; when the numerical difference is less than zero, setting the load compliance determination identifier as a compliance status code; when the numerical difference is greater than or equal to zero, setting the load compliance determination identifier as an overload status code.

[0049] Specifically, the system verifies the validity of the input load data to ensure the accuracy and reliability of subsequent calculations. In one implementation, after obtaining the raw load weight measurement value from the elevator's sensor interface, the system does not use it immediately but first determines whether it falls within a preset valid value range. The lower limit of this range is typically set to zero or a very small positive number close to zero (e.g., 1 kg) to filter out negative drift or invalid readings when the sensor is unloaded; the upper limit can be set to the maximum value of the sensor's range or an empirical value far exceeding the rated load (e.g., 150% of the rated load). Only when the load weight measurement value falls within this reasonable range will the system mark it as valid load data and pass it to the next step. If the measurement value exceeds this range, the system can determine it as an invalid acquisition and process it according to a preset strategy (e.g., using the previous valid data or triggering a sensor fault alarm), thereby avoiding misjudgments based on erroneous data.

[0050] Furthermore, the system dynamically and accurately calculates the safety threshold for the current elevator operation scenario. Unlike using a fixed threshold, the system in this embodiment accesses a preset elevator equipment configuration database, either locally or in the cloud. This database stores key parameters for different elevators within the robot's service range. Based on the elevator ID currently being interacted with, the system retrieves the corresponding rated load parameter (e.g., 1000kg). To reserve safety redundancy, the system does not directly use this rated value but multiplies it by a configurable preset safety factor (e.g., 0.95). For example, if the rated load is 1000kg and the safety factor is 0.95, the system calculates 1000 × 0.95 to determine the preset overload threshold for this judgment as 950kg. This dynamic threshold calculation method not only allows the robot to adapt to various elevator specifications but also ensures operational conservatism and safety by introducing a safety factor.

[0051] Furthermore, based on valid load data and a precise overload threshold, the system generates a clear compliance status code. The system calculates the difference between the valid load data and the preset overload threshold. Specifically, this is a simple subtraction operation: Value difference = Valid load data - Preset overload threshold. The system then makes a final determination based on the sign of this value difference. When the value difference is less than zero, it indicates that the current load is below the safety threshold, and the elevator is in a safe and operable state. In this case, the system sets the load compliance determination identifier to a specific code representing the compliance status (e.g., the integer 0 or the string "NORMAL"). Conversely, when the value difference is greater than or equal to zero, it indicates that the current load has reached or exceeded the allowable safety limit, and the system sets the load compliance determination identifier to a code representing the overload status (e.g., the integer 1 or the string "OVERLOAD"). This finally generated status code will serve as a crucial input, directly participating in subsequent multi-dimensional safety decision-making logic.

[0052] S205. Calculate the reverse motion joint angle based on the safety release signal and the three-dimensional spatial physical coordinates, and generate the pose approximation trajectory of the end effector of the robot's bionic arm according to the reverse motion joint angle. Specifically, the robot is authorized to execute subsequent motion planning only after receiving a generated safety release signal indicating that all preconditions are met. At this point, the system first reads the 3D physical coordinates of the calibrated target button, which preferably constitutes a target pose including position (X, Y, Z) and orientation (e.g., the end effector needs to be perpendicular to the elevator panel). Next, the system calls its built-in inverse kinematics solver to accurately calculate and convert this target pose in Cartesian space into a set of target inverse kinematic joint angles (θ1, θ2, ..., θn)_target that each joint on the bionic arm needs to achieve. To ensure the universality and accuracy of the calculation, a preferred implementation is to use an iterative numerical solution based on the Jacobian matrix. After obtaining the target joint angles, the system does not directly drive the motors to the position, but instead generates a smooth pose approximation trajectory based on these angles. In practice, the system uses the current real-time joint angle of the robotic arm as the starting point and the calculated target joint angle as the ending point. It employs advanced algorithms, such as fifth-order polynomial interpolation, to interpolate the trajectory within the joint space. This method ensures that the velocity and acceleration of the generated trajectory are zero at both the starting and ending points, allowing the end effector of the bionic arm (such as a fingertip) to precisely approach the target button in a shock-free, extremely smooth, and highly controllable manner, preparing for the subsequent pressing action.

[0053] Optionally, the step of calculating the inverse motion joint angles based on the safety release signal and the three-dimensional spatial physical coordinates, and generating the pose approximation trajectory of the end effector of the robot's bionic arm based on the inverse motion joint angles, includes: after receiving the safety release signal, obtaining the current angle encoder values ​​of each joint in the bionic arm, arranging the current angle encoder values ​​according to the preset sequence number of each joint to form a current joint angle vector; obtaining the length parameter values ​​of each link in the bionic arm, the maximum rotation angle value and the minimum rotation angle value of each joint, and establishing a forward kinematics calculation formula between the spatial position of each joint and the end effector based on the length parameter values; inputting the three-dimensional spatial physical coordinates as the target end effector position into the forward kinematics calculation formula, and obtaining the inverse motion joint angles that satisfy the constraints of the maximum rotation angle value and the minimum rotation angle value through numerical iteration; calculating the joint angle differences between the current joint angle vector and the inverse motion joint angles, and performing polynomial function fitting on the joint angle differences based on preset motion time parameters to generate the pose approximation trajectory.

[0054] Specifically, upon receiving a clear safety clearance signal, the robot's motion control unit immediately initiates a self-sensing program for the current state. This unit polls or reads in parallel the current angle encoder values ​​mounted on the servo motors of each movable joint of the bionic arm. These encoders, such as high-precision absolute encoders, provide the precise rotation angle of each joint relative to its zero point. To integrate these discrete readings into a mathematically usable entity, the system arranges the current angle encoder values ​​according to a predefined joint chain order (e.g., from the shoulder joint closest to the robot base to the wrist joint at the far end), thus forming a current joint angle vector, which can be represented as θ_current = [θ1_current, θ2_current, ..., θ...]. n [_current]. This vector accurately describes the initial pose of the bionic arm before initiating the approximation motion and is a necessary starting point for subsequent trajectory planning.

[0055] Furthermore, to perform kinematic calculations, the system needs to establish a mathematical model describing its own physical structure. In this step, the system retrieves the inherent physical parameters of the bionic arm, which have been precisely calibrated at the factory, from a pre-set configuration file or database. These parameters mainly include the length parameters of each link (i.e., the precise distance between the rotation axes of two adjacent joints) and the range of motion limits of each joint, i.e., the maximum and minimum rotation angle values. Based on these parameters, especially the link lengths, the system establishes a forward kinematic calculation formula for the spatial position of each joint and the end effector. A common and efficient implementation method in the industry is to use the Denavit-Hartenberg (DH) parametric method to establish this model. This model can give a definite function P_end=ƒ(θ1, θ2, ..., θn) through a series of coordinate transformation matrices. This function can uniquely map any given joint angle vector to the position and orientation of the end effector in three-dimensional space.

[0056] Furthermore, the system performs the core inverse kinematics solution to calculate the joint configuration required to reach the target position. In this step, the system uses the acquired three-dimensional physical coordinates of the target button (preferably also including the desired end pose, together forming a target pose) as the target end position P_target, and inputs it into the forward kinematics calculation formula. Since inverse kinematics problems typically do not have analytical solutions, the system searches for the answer through numerical iteration. Specific solution methods may include, but are not limited to, the Jacobian matrix pseudo-inverse method, damped least squares method, or cyclic coordinate descent method. In each step of the iterative solution, the system checks whether the calculated intermediate solutions satisfy the constraints of the maximum rotation angle value and the minimum rotation angle value. Only solutions where all joint angles are within their effective range of motion are considered feasible solutions. Finally, through multiple iterations, the system converges to a target joint angle vector that meets the accuracy requirements and conforms to the physical constraints, i.e., the inverse kinematic joint angle.

[0057] Furthermore, to ensure smoothness and safety of the motion, the system generates a complete motion trajectory based on the calculated start and end points. The system first calculates the joint angle differences between the current joint angle vector and the reverse motion joint angle. This difference vector represents the total angle each joint needs to rotate. Then, the system introduces a configurable preset motion time parameter T, which defines the total time required to complete the entire approximation motion. Next, the system performs polynomial function fitting on the joint angle differences, thereby independently planning a motion curve that varies with time for each joint. A preferred implementation is to use a fifth-order polynomial function, as it ensures that the velocity and acceleration of the trajectory are zero at both the start and end points, thus achieving smooth start and stop and avoiding mechanical shock. This fitting process ultimately generates the pose approximation trajectory, which is essentially a sequence of joint angle commands that vary with time t (from 0 to T). The servo system will strictly follow this sequence to drive the bionic arm, causing its end effector to move smoothly and accurately to the target position.

[0058] S206. Output a driving current waveform according to the pose approximation trajectory, and adjust the transmission component based on the driving current waveform to make the execution end press the target layer call control, thereby generating physical contact feedback torque data. Specifically, the pose approximation trajectory is essentially a series of high-density target angle setpoints that vary over time. In each control cycle (e.g., every millisecond), the robot's servo drive system takes one setpoint on the trajectory as the current target and compares it with the actual angle read in real-time from the joint encoder, thus obtaining a position error. A built-in controller, employing, for example, a PID (proportional-integral-derivative) algorithm, calculates and outputs a precise drive current waveform in real-time based on this error. This current waveform is applied to the servo motor of the corresponding joint, generating a precise drive torque through electromagnetic effects. This torque, after being adjusted and amplified by transmission components such as a reducer, drives the joint to rotate, thereby ensuring that the entire bionic arm strictly follows the pose approximation trajectory. It is worth noting that the endpoint of the pose approximation trajectory is typically designed to slightly cross the surface plane of the target button. This virtual overdrive command ensures that when the end effector (e.g., a bionic finger) reaches and contacts the target layer call control, the control system continues to output torque to further reduce the position error, thus naturally completing the pressing action. When a button is pressed, due to the reaction force from the button, the servo motor's PID controller automatically increases the output current to overcome resistance and maintain or continue approaching the target position. The system monitors this actual output current at high frequency and compares it with the theoretical model current required to perform the same action in collision-free free space. The difference between the two can be accurately converted and generated as physical contact feedback torque data. As an alternative implementation, a multi-axis force / torque sensor can be added to the actuator or wrist. When contact occurs, this sensor will directly measure and output more accurate torque data. Regardless of the method used, this set of real-time generated torque data provides crucial and quantitative evidence for subsequent decisions such as whether the press was successful or whether abnormal resistance was encountered.

[0059] S207. Based on the physical contact feedback torque data, determine the pressing state. When the pressing state is determined to be successful, reverse modulate the driving current waveform to generate a reverse driving current waveform. Based on the reverse driving current waveform, drive the transmission component to move the execution end to the preset initial standby position and generate the end-to-end retraction reset trajectory.

[0060] After the bionic arm's end effector makes physical contact with the target control, step S207 begins real-time pressing status determination. Specifically, the motion control system continuously and frequently samples the physical contact feedback torque data generated in the previous step and compares it with a preset torque threshold. This threshold is pre-calibrated based on the typical pressing force of an elevator button (e.g., equivalent to 3-5 Newtons of motor torque). To ensure robustness of the determination and effectively filter out torque spikes caused by jitter or instantaneous collisions, a preferred determination logic is: the system only determines a successful pressing state if and only if the feedback torque data not only exceeds the preset threshold but also remains above this threshold for a "preset time window" (e.g., 100 milliseconds). Once the determination is successful, the system immediately triggers a retraction procedure, i.e., reverse modulation of the drive current waveform. Here, "reverse modulation" does not simply mean reversing the current, but rather that the system immediately plans and generates a completely new end effector retraction and reset trajectory. The trajectory begins at the position where the button is currently pressed and ends at a preset initial standby position—a specific spatial coordinate or joint angle configuration located in front of the elevator control panel, safe and without interfering with passengers. The system then drives the transmission component based on the reverse drive current waveform (i.e., the drive current calculated in real time to achieve this new trajectory), causing the execution end to smoothly and quickly retract along the newly generated reset trajectory until it stably stops at the standby position, thus efficiently and safely completing a full button interaction loop.

[0061] Optionally, the step of determining the pressing state based on the physical contact feedback torque data, and when the pressing state is determined to be successful, performing reverse modulation on the drive current waveform to generate a reverse drive current waveform, and driving the transmission component based on the reverse drive current waveform to move the execution end to a preset initial standby position, generating an end-effector retraction trajectory, includes: performing time-domain sliding window filtering on the physical contact feedback torque data to generate a smooth torque feature sequence; performing peak detection calculation on the smooth torque feature sequence to extract the peak torque value and the peak occurrence time; comparing the peak torque value with a preset pressing success threshold; and when the peak torque value is greater than the preset pressing success threshold... When the time difference between the peak occurrence time and the press start time is less than a preset response time threshold, the press state determination is marked as a press success state code; after receiving the press success state code, a negative mapping transformation is performed on each amplitude component in the amplitude parameter matrix of the drive current waveform, keeping the frequency parameter matrix in the drive current waveform unchanged, and the reverse drive current waveform is generated; based on the reverse drive current waveform, the transmission component is driven, and the movement of the execution end is controlled according to a preset multi-segment speed planning curve. When the distance between the execution end and the target layer call control reaches a preset safety distance threshold, the movement stops and the initial standby position is maintained, generating the end-reset trajectory.

[0062] Specifically, the acquired physical contact feedback torque data is analyzed in real time to determine whether the pressing state is successful. This analysis process includes comparing the high-frequency sampled torque data with a preset torque threshold, which is pre-calibrated based on the physical characteristics of the target control (such as the trigger pressure of an elevator button). To improve the accuracy of the determination results and filter out instantaneous torque disturbances caused by mechanical vibration or initial contact, at least two determination schemes can be adopted. Scheme 1 is a combined amplitude-duration determination, meaning that a pressing is considered successful only when the amplitude of the feedback torque exceeds the preset threshold and the duration above the threshold meets a preset duration (e.g., 50 to 200 milliseconds). Scheme 2 is torque curve feature matching, meaning that by analyzing the time series of torque data, a pressing is considered successful when the curve shape matches the characteristics of a typical successful pressing process (e.g., the torque rises rapidly and then enters a stable high-value plateau region). When a pressing state is determined to be successful according to either scheme, the subsequent disengagement and reset operation is triggered.

[0063] Furthermore, upon successful press confirmation, a motion trajectory is generated for the bionic arm's end effector to disengage and reset. This process involves active replanning of the drive commands, aiming to generate a completely new motion sequence. The starting point of this reset trajectory is set to the current pose of the end effector (i.e., the spatial position and posture of the button being successfully pressed), and its ending point is set to a preset initial standby position. This position is typically a safe, open spatial coordinate point in front of the control panel or a specific set of joint angles to ensure that the bionic arm can safely complete its current task and enter standby mode. The trajectory can be generated using algorithms such as fifth-order polynomial interpolation to ensure smooth velocity and acceleration curves at both the starting and ending points of the motion. The series of time-varying drive current commands calculated to achieve this reset trajectory constitute the aforementioned reverse drive current waveform.

[0064] Furthermore, the generated reverse drive current waveform is output to the servo drive circuit to precisely drive the transmission components to perform a reset action. Based on this current waveform, the servo drive circuit applies precise current to the servo motors of each joint, thereby generating the required driving torque. Under the action of these torques, the transmission components (including but not limited to reducers, linkages, etc.) drive the end effector of the bionic arm, making it move smoothly and quickly, strictly following the aforementioned generated reset trajectory. Since the trajectory itself ensures the smoothness of the movement, the retraction process of the end effector is stable and controllable, effectively avoiding equipment vibration or instability that may be caused by rapid movement. Finally, the end effector moves to and stably stops at the preset initial standby position, marking the completion of a complete button interaction operation loop.

[0065] It should be noted that the above embodiments of the apparatus are only illustrated by the division of the above functional modules. In practical applications, the above functions can be assigned to different functional modules as needed, that is, the internal structure of the device can be divided into different functional modules to complete all or part of the functions described above. In addition, the apparatus and method embodiments provided in the above embodiments belong to the same concept, and the specific implementation process can be found in the method embodiments, which will not be repeated here.

[0066] This embodiment also discloses an electronic device, as shown in the reference. Figure 3 The electronic device may include: at least one processor 301, at least one communication bus 302, user interface 303, network interface 304, and at least one memory 305.

[0067] The communication bus 302 is used to enable communication between these components.

[0068] The user interface 303 may include a display screen and a camera. Optionally, the user interface 303 may also include a standard wired interface and a wireless interface.

[0069] The network interface 304 may optionally include a standard wired interface or a wireless interface (such as a Wi-Fi interface).

[0070] The processor 301 may include one or more processing cores. The processor 301 connects to various parts of the server using various interfaces and lines, and performs various server functions and processes data by running or executing instructions, programs, code sets, or instruction sets stored in memory 305, and by calling data stored in memory 305. Optionally, the processor 301 may be implemented using at least one hardware form of Digital Signal Processing (DSP), Field-Programmable Gate Array (FPGA), or Programmable Logic Array (PLA). The processor 301 may integrate one or a combination of several of the following: Central Processing Unit (CPU), Graphics Processing Unit (GPU), and modem. The CPU primarily handles the operating system, user interface, and applications; the GPU is responsible for rendering and drawing the content required for display; and the modem handles wireless communication. It is understood that the modem may also not be integrated into the processor 301 and may be implemented as a separate chip.

[0071] The memory 305 may include random access memory (RAM) or read-only memory. Optionally, the memory 305 may include a non-transitory computer-readable storage medium. The memory 305 can be used to store instructions, programs, code, code sets, or instruction sets. The memory 305 may include a program storage area and a data storage area, wherein the program storage area may store instructions for implementing an operating system, instructions for at least one function (such as touch function, sound playback function, image playback function, etc.), instructions for implementing the above-described method embodiments, etc.; the data storage area may store data involved in the above-described method embodiments, etc. Optionally, the memory 305 may also be at least one storage device located remotely from the aforementioned processor 301. Figure 3As shown, the memory 305, which serves as a computer storage medium, may include an operating system, a network communication module, a user interface module, and an application program for a control method of a humanoid robot for an elevator.

[0072] exist Figure 3 In the electronic device shown, the user interface 303 is mainly used to provide an input interface for the user and to obtain the user input data; while the processor 301 can be used to call an application program stored in the memory 305 for a control method of an embodied humanoid robot for an elevator. When executed by one or more processors 301, the electronic device performs one or more methods as described in the above embodiments.

[0073] It should be noted that, for the sake of simplicity, the foregoing method embodiments are all described as a series of actions. However, those skilled in the art should understand that this application is not limited to the described order of actions, as some steps may be performed in other orders or simultaneously according to this application. Furthermore, those skilled in the art should also understand that the embodiments described in the specification are preferred embodiments, and the actions and modules involved are not necessarily essential to this application.

[0074] In the several embodiments provided in this application, it should be understood that the disclosed apparatus can be implemented in other ways. For example, the apparatus embodiments described above are merely illustrative; for instance, the division of units is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the shown or discussed mutual couplings or direct couplings or communication connections may be through some service interfaces; indirect couplings or communication connections between apparatuses or units may be electrical or other forms.

[0075] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.

[0076] Furthermore, the functional units in the various embodiments of this application can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. If the integrated unit is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage device (CMD). Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, or all or part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a memory 305 and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods in the various embodiments of this application. The aforementioned memory 305 includes various media capable of storing program code, such as a USB flash drive, external hard drive, magnetic disk, or optical disk.

[0077] The foregoing description is merely an exemplary embodiment of this disclosure and should not be construed as limiting the scope of this disclosure. Any equivalent changes and modifications made in accordance with the teachings of this disclosure shall still fall within the scope of this disclosure. Other embodiments of this disclosure will be readily apparent to those skilled in the art upon consideration of the disclosure in this specification. This application is intended to cover any variations, uses, or adaptations of this disclosure that follow the general principles of this disclosure and include common knowledge or customary techniques in the art not described in this disclosure. The specification and embodiments are to be considered exemplary only, and the scope and spirit of this disclosure are defined by the claims.

Claims

1. A control method for a humanoid robot used in an elevator, characterized in that, The method includes: Acquire occupant feature data and layer call configuration data, and construct a permission access vector based on the occupant feature data and a preset authorization map; Based on the permitted access vector and the layer call configuration data, a security conflict detection calculation is performed to generate a target floor scheduling instruction; According to the target floor dispatch instruction, obtain the control panel image data, and perform image feature extraction based on the control panel image data to determine the three-dimensional spatial physical coordinates of the target floor call control. The system acquires real-time operating parameters of the elevator and environmental status monitoring data, performs a fusion threshold comparison based on the real-time operating parameters of the elevator and the environmental status monitoring data, and generates a safe release signal when the comparison result meets the preset safety conditions. The reverse motion joint angle is calculated based on the safety release signal and the three-dimensional spatial physical coordinates, and the pose approximation trajectory of the end effector of the robot's bionic arm is generated based on the reverse motion joint angle. Based on the pose approximation trajectory, a drive current waveform is output, and the transmission component is adjusted based on the drive current waveform to make the execution end press the target layer call control, thereby generating physical contact feedback torque data. The pressing state is determined based on the physical contact feedback torque data. When the pressing state is determined to be successful, the driving current waveform is reverse modulated to generate a reverse driving current waveform. Based on the reverse driving current waveform, the transmission component is driven to move the execution end to the preset initial standby position, generating the end to disengage from the reset trajectory.

2. The method according to claim 1, characterized in that, The step of performing security conflict detection calculations based on the permitted access vector and the layer call configuration data to generate target floor scheduling instructions includes: The floor stop record matrix contained in the floor call configuration data is logically matched with the permitted access vector to extract a conflict-free candidate floor sequence. The conflict-free candidate floor sequence is traversed and calculated based on a preset scheduling distance evaluation function to determine the minimum distance cost feature; Based on the minimum distance cost feature, the corresponding target layer height distribution parameters are extracted, and the target layer height distribution parameters are encapsulated to generate the target layer scheduling instruction.

3. The method according to claim 1, characterized in that, The step of extracting image features based on the control panel image data to determine the three-dimensional physical coordinates of the target layer control includes: The control panel image data is subjected to grayscale conversion and Gaussian filtering for noise reduction to generate a smooth feature image matrix; Extract the target floor identifier characters contained in the target floor scheduling instruction, perform morphological contour matching calculation on the smooth feature image matrix based on the target floor identifier characters, and extract the two-dimensional plane pixel coordinates of the geometric centroid of the target closed connected domain; The depth measurement data of the control panel is synchronously acquired from the vision sensor. The pre-stored camera intrinsic parameter matrix and distortion correction parameters are retrieved to form an imaging model calibration parameter matrix. Based on the imaging model calibration parameter matrix, the two-dimensional plane pixel coordinates and the depth measurement data are calculated by inverse perspective projection mapping to obtain the three-dimensional spatial physical coordinates of the target layer control panel.

4. The method according to claim 1, characterized in that, The process involves acquiring real-time operating parameters of the elevator and environmental status monitoring data, performing a fusion threshold comparison based on the real-time operating parameters and environmental status monitoring data, and generating a safety release signal when the comparison result meets preset safety conditions. This includes: Extract the cage door closing status signal, landing door interlock status signal, and load weight measurement value from the real-time operating parameters of the elevator. Perform a logical AND operation on the cage door closing status signal and the landing door interlock status signal to obtain the door safety verification result. The load weight measurement value is compared with a preset overload threshold to generate a load compliance determination mark. Extract the cage tilt angle measurement value, wind speed measurement value, and obstacle distance measurement value from the environmental status monitoring data. Compare and calculate the cage tilt angle measurement value, wind speed measurement value, and obstacle distance measurement value with the corresponding preset safety thresholds to generate an environmental risk assessment vector. Based on the door safety verification result, the load compliance judgment identifier, and the environmental risk assessment vector, a multi-dimensional logical judgment is performed. When the result of the multi-dimensional logical judgment meets the safety conditions, the safety release signal is generated.

5. The method according to claim 4, characterized in that, The step of comparing the measured load weight with a preset overload threshold to generate a load compliance determination identifier includes: Determine whether the measured load weight value is within a preset valid value range. If the measured load weight value is within the preset valid value range, mark it as valid load data. Obtain the rated load parameters stored in the preset elevator equipment configuration database, multiply the rated load parameters with the preset safety factor, and determine the preset overload threshold. Calculate the numerical difference between the effective load data and the preset overload threshold. When the numerical difference is less than zero, set the load compliance judgment identifier to a compliance status code. When the numerical difference is greater than or equal to zero, set the load compliance judgment identifier to an overload status code.

6. The method according to claim 1, characterized in that, The step of calculating the inverse motion joint angle based on the safety release signal and the three-dimensional spatial physical coordinates, and generating the pose approximation trajectory of the robot's bionic arm end effector based on the inverse motion joint angle, includes: After receiving the safety release signal, the current angle encoder values ​​of each joint in the bionic arm are obtained, and the current angle encoder values ​​are arranged according to the preset sequence number of each joint to form the current joint angle vector. Obtain the length parameter values ​​of each link in the bionic arm, the maximum rotation angle value of each joint, and the minimum rotation angle value, and establish a positive kinematic calculation formula between the spatial position of each joint and the end effector based on the length parameter values; The three-dimensional spatial physical coordinates are used as the target end position and input into the forward kinematics calculation formula. The reverse motion joint angle that satisfies the constraints of the maximum rotation angle value and the minimum rotation angle value is obtained through numerical iteration. Calculate the joint angle differences between the current joint angle vector and the reverse motion joint angle, and perform polynomial function fitting on the joint angle differences based on preset motion time parameters to generate the pose approximation trajectory.

7. The method according to claim 1, characterized in that, The step of determining the pressing state based on the physical contact feedback torque data, and when the pressing state is determined to be successful, reverse modulation of the drive current waveform to generate a reverse drive current waveform, and driving the transmission component based on the reverse drive current waveform to move the execution end to a preset initial standby position, generating an end-effector disengagement trajectory, includes: The physical contact feedback torque data is subjected to time-domain sliding window filtering to generate a smooth torque feature sequence. Peak detection calculation is performed on the smooth torque feature sequence to extract the peak torque value and the peak occurrence time. The peak torque value is compared with a preset successful press threshold. When the peak torque value is greater than the preset successful press threshold and the time difference between the peak occurrence time and the press start time is less than a preset response time threshold, the press state determination is marked as a successful press state code. After receiving the press success status code, a negative mapping transformation is performed on each amplitude component in the amplitude parameter matrix of the drive current waveform, while keeping the frequency parameter matrix in the drive current waveform unchanged, to generate the reverse drive current waveform. The transmission component is driven based on the reverse drive current waveform, and the movement of the execution end is controlled according to the preset multi-segment speed planning curve. When the distance between the execution end and the target layer call control reaches the preset safe distance threshold, the movement stops and the initial standby position is maintained, generating the end-reset trajectory.

8. An electronic device, characterized in that, The device includes a processor, a memory, a user interface, and a network interface. The memory is used to store instructions. The user interface and the network interface are both used to communicate with other devices. The processor is used to execute the instructions stored in the memory to cause the electronic device to perform the method as described in any one of claims 1-7.

9. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores instructions that, when executed, perform the method as described in any one of claims 1-7.

10. A computer program product, characterized in that, When the computer program product is run on an electronic device, it causes the electronic device to perform the method as described in any one of claims 1-7.

Images