Robot elevator passage method, device, apparatus and mobile tool

Abstract

This application relates to a method, apparatus, device, and mobile tool for robot elevator passage. The method includes: acquiring point cloud data of multiple consecutive frames of the object to be identified; converting and fusing the point cloud data of the multiple consecutive frames based on the robot's real-time odometer information to construct a multi-frame feature point cloud set; evaluating the reliability of the point cloud data based on the multi-frame feature point cloud set using an evaluation function pre-trained based on a data-driven evaluation mechanism; comparing the reliability with preset quality assessment conditions, and updating the elevator's pose information in the robot coordinate system according to different matching update strategies based on the comparison results; when the pose information meets the elevator entry constraints, controlling the robot to enter the elevator according to a preset edge-constraint entry strategy. This application achieves pose perception through multi-frame point cloud fusion, combined with quality constraints and update strategies under different conditions, ultimately controlling the robot to enter along the edge, maintaining passage stability, continuity, and accuracy.

Description

Technical Field

[0001] This application relates to the field of robotics, and in particular to robotic elevator passage methods, devices, equipment and mobility tools. Background Technology

[0002] With the increasing application of service robots, delivery robots, and cleaning robots in multi-story indoor environments such as hospitals and office buildings, robots frequently need to pass through elevators for cross-floor tasks. Stable alignment and reliable entry at elevator doors are crucial to the success rate and safety of passage. The environment at elevator doors is complex, with highly reflective materials such as metal doors and glass decorations. Combined with dynamic interference from frequent elevator door opening and closing and random pedestrian entry and exit, this can easily lead to problems such as abnormal echoes, missing data, or unstable distribution in the robot's LiDAR point cloud, posing significant challenges to pose perception and alignment control. Relying solely on single-frame point clouds or instantaneous perception can easily result in unstable recognition and positioning errors. Furthermore, the significant differences in the structural dimensions of different elevators place higher demands on the robustness and adaptability of the passage method.

[0003] In existing technologies, there are three main approaches for mobile robot elevator access: First, a geometric recognition and fitting scheme based on single-frame laser point clouds. This involves collecting single-frame point clouds by setting reflective markers or utilizing geometric features such as the elevator door frame, and then fitting these points to calculate the elevator pose, guiding the robot to align and enter. However, this approach is heavily reliant on single-frame point clouds. Short-term loss or distortion of the point cloud due to occlusion, abnormal reflection, or other factors can easily lead to abrupt changes in pose estimation, affecting the continuity and stability of alignment. Second, a point cloud screening scheme based on fixed threshold rules. This involves limiting the number of effective point clouds and setting thresholds for fitting residuals to select usable point clouds to update the pose. However, fixed thresholds or rules lack robustness under different elevator structures and usage scenarios, making it difficult to adapt to environmental changes, prone to misjudgments and missed judgments, and resulting in low accuracy. Furthermore, these methods lack a stable pose maintenance strategy when the point cloud is temporarily unavailable, easily interrupting the alignment process or requiring repositioning, reducing access efficiency and posing safety risks.

[0004] It is evident that existing technologies, when robots enter elevators, suffer from problems such as unstable pose estimation, low robustness, and inability to efficiently handle situations where point clouds are temporarily unavailable, due to their reliance on single-frame point cloud judgment and fixed threshold rules for pose updates. Ultimately, these issues affect the robot's success rate in passing through elevators. Summary of the Invention

[0005] This application provides a method, apparatus, device, and mobile tool for robot elevator passage, in order to solve the problems in the prior art, which are prone to sudden changes in pose estimation and have low robustness due to the judgment based on a single frame point cloud and the updating of pose based on a fixed threshold rule, and which cannot efficiently cope with the situation of short-term unavailability of point cloud, ultimately affecting the success rate of robot passage.

[0006] According to one aspect of the embodiments of this application, this application provides a robot elevator passage method, the method comprising: acquiring point cloud data of multiple consecutive frames of an object to be identified; converting and fusing the point cloud data of the multiple consecutive frames based on the robot's real-time odometer information to construct a multi-frame feature point cloud set; evaluating the reliability of the point cloud data based on the multi-frame feature point cloud set using an evaluation function pre-trained based on a data-driven evaluation mechanism; comparing the reliability of the point cloud data with preset quality evaluation conditions, and updating the elevator's pose information in robot coordinates according to different matching update strategies based on the comparison results; and controlling the robot to enter the elevator according to a preset edge constraint entry strategy when the pose information meets the elevator entry constraint conditions.

[0007] Optionally, before acquiring point cloud data of multiple consecutive frames of the object to be identified, the method further includes: controlling the robot to reach a preset initial alignment position at the elevator entrance according to a preset path; adjusting the robot's orientation at the preset initial alignment position so that the robot faces the object to be identified within the orientation angle constraint.

[0008] Optionally, the acquisition of point cloud data of the identified object in multiple consecutive frames includes: when the identified object is within the scanning range of the robot, acquiring multiple consecutive frames of raw echo point cloud data of the identified object through a scanning device; preprocessing each frame of raw echo point cloud data, filtering by reflection intensity threshold to obtain a subset of the identified object point cloud for each frame; transforming each frame of the identified object point cloud subset from the coordinate system of the scanning device to the local coordinate system of the robot, and fusing the transformed multiple consecutive frames of the identified object point cloud, filtering out duplicate point clouds, to obtain multiple consecutive frames of the identified object point cloud data in the local coordinate system of the robot.

[0009] Optionally, the conversion and fusion processing of point cloud data across multiple consecutive frames based on the robot's real-time odometry information to construct a multi-frame feature point cloud set includes: real-time acquisition of odometry information to obtain robot pose data corresponding to the acquisition time of point cloud data across multiple consecutive frames; using the robot pose data corresponding to the first frame of point cloud data as a reference, and combining the robot pose data corresponding to the acquisition time of each frame, performing coordinate transformation on the point cloud data of each subsequent frame to transform the point cloud data of all frames to the robot coordinate system; fusing the coordinate-transformed point cloud data across multiple consecutive frames, filtering out invalid point cloud data, and constructing the multi-frame feature point cloud set based on the retained valid point cloud features.

[0010] Optionally, the step of evaluating the reliability of the point cloud data based on the multi-frame feature point cloud set using an evaluation function pre-trained based on a data-driven evaluation mechanism includes: calculating multi-dimensional feature indicators based on a continuous multi-frame point cloud dataset of the identified object collected and labeled in history. The multi-dimensional feature indicators include point cloud dispersion, point cloud quantity, and point cloud plane fitting residual, wherein each frame of point cloud is labeled with a point cloud quality label; training a model based on the data-driven evaluation mechanism using the point cloud dataset and the extracted point cloud dispersion, point cloud quantity, and point cloud plane fitting residual to obtain the evaluation function, wherein the output target of the evaluation function is the point cloud quality label used to evaluate the reliability of the point cloud data; obtaining the target point cloud dispersion, target point cloud quantity, and target point cloud plane fitting residual from the multi-frame feature point cloud set; and calculating the reliability of the current point cloud data using the evaluation function based on the target point cloud dispersion, target point cloud quantity, and target point cloud plane fitting residual.

[0011] Optionally, the step of comparing the reliability of the point cloud data with preset quality assessment conditions and updating the elevator's pose information in robot coordinates according to different comparison results and matching corresponding update strategies includes: comparing the reliability of the current point cloud data with preset quality assessment conditions to determine whether the reliability of the current point cloud data meets the preset quality assessment conditions; if the reliability of the current point cloud data meets the preset quality assessment conditions within a preset time window, then the quality of the current point cloud data is determined to be qualified, and pose update is performed based on a first update strategy, the first update strategy including updating the elevator's pose information in robot coordinates according to the multi-frame feature point cloud set; if the reliability of the current point cloud data does not meet the preset quality assessment conditions within the preset time window, then the quality of the current point cloud data is determined to be unqualified, and pose update is performed based on a second update strategy, the second update strategy including obtaining the pose information corresponding to when the reliability of the point cloud data was determined to meet the preset quality assessment conditions in the most recent historical period, and updating the elevator's pose information in robot coordinates in combination with the robot's current odometer information.

[0012] Optionally, when the pose information satisfies the elevator entry constraint conditions, controlling the robot to enter the elevator according to a preset edge-constrained entry strategy includes: determining whether the pose information satisfies the preset elevator entry constraint conditions, wherein the preset elevator entry constraint conditions include the horizontal distance between the robot and the elevator door, the angle between the robot's heading and the elevator door, and / or the robot's lateral offset; if the pose information satisfies the preset elevator entry constraint conditions, then switching to the preset edge-constrained entry strategy, using the edge of the elevator door as a constraint reference, controlling the robot to maintain a preset speed and a preset safety distance to enter the elevator along the edge of the elevator door.

[0013] According to another aspect of the embodiments of this application, this application provides a robot elevator access device, the device comprising: a data acquisition module for acquiring point cloud data of multiple consecutive frames of an object to be identified; a feature extraction module for converting and fusing the point cloud data of multiple consecutive frames based on the robot's real-time odometer information to construct a multi-frame feature point cloud set; a feature evaluation module for evaluating the reliability of the point cloud data based on the multi-frame feature point cloud set using an evaluation function pre-trained based on a data-driven evaluation mechanism; a pose update module for comparing the reliability of the point cloud data with preset quality evaluation conditions, and updating the pose information of the elevator in robot coordinates according to the different comparison results and matching corresponding update strategies; and a motion control module for controlling the robot to enter the elevator according to a preset edge constraint entry strategy when the pose information meets the elevator entry constraint conditions.

[0014] According to another aspect of the embodiments of this application, this application provides an electronic device, including: a processor, a memory, and a network interface. The memory stores machine-readable instructions executable by the processor. When the electronic device is running, the processor communicates with the memory through the network interface, and the processor executes the machine-readable instructions to perform the steps of the robot elevator passage method as described above.

[0015] According to another aspect of the embodiments of this application, this application provides a mobile tool, including the electronic device shown above.

[0016] Compared with related technologies, the technical solutions provided in this application have the following advantages: This application provides a method for robot elevator passage, applicable to elevator entry and passage scenarios for service robots, delivery robots, and cleaning robots in complex indoor environments. This application constructs a stable feature point cloud through multi-frame point cloud fusion, adaptively determines point cloud reliability using a data-driven evaluation mechanism, updates elevator pose based on reliability differentiation, and employs an edge-following strategy for elevator entry when constraints are met. This significantly improves the stability and alignment continuity of elevator pose recognition, enhances the robustness and safety of passage in complex environments, and solves the technical problem of low success rate for robot elevator passage. Attached Figure Description

[0017] The accompanying drawings, which are incorporated in and form part of this specification, illustrate embodiments consistent with this application and, together with the description, serve to explain the principles of this application.

[0018] To more clearly illustrate the technical solutions in the embodiments of this application or related technologies, the accompanying drawings used in the description of the embodiments or related technologies will be briefly introduced below. Obviously, those skilled in the art can obtain other drawings based on these drawings without creative effort.

[0019] Figure 1 This is a schematic diagram of the hardware environment for an optional robotic elevator passage method provided according to an embodiment of this application; Figure 2 This is a flowchart illustrating an optional robotic elevator passage method according to an embodiment of this application; Figure 3 This is a schematic diagram of an optional elevator sign provided according to an embodiment of this application; Figure 4 This is a schematic diagram of an optional radar point cloud acquisition method according to an embodiment of this application; Figure 5 This is a flowchart of another optional robot elevator passage method provided according to an embodiment of this application; Figure 6 This is a block diagram of an optional robotic elevator access device according to an embodiment of this application; Figure 7 This is a schematic diagram of an optional electronic device structure provided in an embodiment of this application. Detailed Implementation

[0020] To make the objectives, technical solutions, and advantages of the embodiments of this application clearer, the technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0021] To address the problems mentioned in the background art, according to one aspect of the embodiments of this application, an embodiment of a robot elevator passage method is provided.

[0022] like Figure 1 As shown, the above-described robot elevator access method can be applied to, for example... Figure 1 The hardware environment shown is a system architecture 100 that includes a terminal device 101 and a server 103. The server 103 is connected to the terminal device 101 via a network and can provide services to the terminal device 101 or clients installed on the terminal device 101. A database 105 can be set up on or independently of the server 103 to provide data storage services for the server 103. The network can include various connection types, such as wired, wireless communication links, or fiber optic cables, etc.

[0023] Users can use terminal device 101 to interact with server 103 via a network to receive or send messages. Various communication client applications can be installed on terminal device 101, such as web browsers, search applications, and instant messaging tools. Terminal device 101 can be various electronic devices with a display screen that support web browsing, including but not limited to smartphones, tablets, e-book readers, MP3 players (Moving Picture Experts Group Audio Layer III), MP4 players (Moving Picture Experts Group Audio Layer IV), laptops, and desktop computers. Server 103 can be a server providing various services, such as a backend server supporting the pages displayed on terminal device 101. The server can be installed inside the robot.

[0024] It should be noted that the robot elevator access method provided in this application embodiment is generally executed by a server and / or terminal device, and correspondingly, the robot elevator access device is generally installed in the server / terminal device. Furthermore, it should be understood that... Figure 1The number of terminal devices, networks, and servers shown is merely illustrative. Depending on implementation needs, any number of terminal devices, networks, and servers can be included.

[0025] like Figure 2 As shown, Figure 2 A flowchart of a robot elevator passage method provided in an embodiment of the present invention. Taking the robot elevator passage method being executed by a server as an example, the robot elevator passage method includes the following steps: Step S202: Collect point cloud data of the identified object in multiple consecutive frames.

[0026] In this embodiment, the provided robot elevator access method is applicable to elevator entry and access scenarios for service robots, delivery robots, and cleaning robots in complex indoor environments.

[0027] The object to be identified can refer to a reflective marker with specific reflective properties placed at the elevator entrance, which can be identified and collected by a LiDAR on a robot. Alternatively, it can be a target surface such as the elevator door frame, door edge, door gap, sill, or doorway pillars. When scanned by LiDAR, these surfaces can all receive echoes, generating independent original 3D point cloud data, along with ranging information and reflection intensity information. Based on the characteristics of reflective markers—strong echoes, dense point clouds, low dispersion, and low fitting residuals—making it easier to generate high signal-to-noise ratio, high-quality point clouds, this application uses reflective markers as the object to be identified.

[0028] When the robot reaches the approximate location at the elevator entrance, it continuously collects high-intensity echo point cloud data [0,255] generated by the reflective marker through its own lidar for multiple frames to obtain point cloud data of the reflective marker in the robot's local coordinate system for multiple consecutive frames.

[0029] Step S204: Based on the robot's real-time odometer information, the point cloud data of multiple consecutive frames are converted and fused to construct a multi-frame feature point cloud set.

[0030] The odometry information includes real-time translational coordinates (x, y), heading and yaw angle θ, and timestamp t. Rotational pulses can be collected by the wheel encoders on the robot and the displacement and rotation angle can be calculated by combining them with chassis calibration parameters. This data is then fused with IMU inertial attitude compensation to obtain real-time pose data with timestamps, thus providing the robot's real-time odometry information.

[0031] Furthermore, based on the robot's real-time odometer information, the relative pose transformation relationship of the local point cloud data acquisition time of each frame is determined. Coordinate rigid body transformation is performed on the point cloud data of multiple consecutive frames to unify them into the same reference coordinate system. Deduplication and aggregation processing are then performed to form a stable multi-frame fused feature point cloud, thereby constructing a multi-frame feature point cloud set, which provides a reliable perception basis for subsequent elevator pose calculation and precise alignment.

[0032] Step S206: The reliability of the point cloud data is evaluated based on the multi-frame feature point cloud set using an evaluation function pre-trained based on a data-driven evaluation mechanism.

[0033] Among them, the data-driven evaluation mechanism can refer to the use of a large amount of historically collected point cloud sample data with quality labels, through machine learning, fitting or regression training, to automatically learn and construct the mapping relationship between point cloud quality and various evaluation features, forming a generalizable evaluation model or evaluation function, thereby realizing automatic and adaptive evaluation of point cloud data quality.

[0034] Furthermore, a pre-trained evaluation function is used, with each frame of the feature point cloud set from the multi-frame feature point cloud set as input data. Each frame of the feature point cloud represents the quality of the point cloud. Through autonomous learning on the input data, a reliability score is output to evaluate the quality of the point cloud. The reliability score is expressed numerically, for example, 80%.

[0035] Step S208: Based on the reliability of the point cloud data, compare it with the preset quality assessment conditions, and update the elevator's pose information in robot coordinates according to the different matching update strategies based on the comparison results.

[0036] The preset quality assessment condition can be a reliability threshold, specifically referring to point cloud accuracy. Regardless of whether the reliability meets the preset quality assessment condition, the pose information can still be updated, but the update strategies will differ.

[0037] Furthermore, if the reliability of the point cloud data reaches or exceeds the preset quality assessment conditions, it indicates that the quality of the currently acquired point cloud data is qualified, and the pose information can be directly updated. If it does not reach the conditions, it indicates that the quality of the currently acquired point cloud data is unqualified, and a valid point cloud cannot be obtained within a certain time. Therefore, pose information cannot be updated based on the current point cloud data. In this case, the update can be performed by combining historical point cloud data that meets the preset quality assessment conditions with the current pose. This maintains the continuity and stability of the alignment process, avoiding the direct interruption of the alignment process or re-triggering of localization when a valid point cloud cannot be obtained within a certain time window. This would prevent frequent switching of robot actions, reduced passage efficiency, and potential alignment failures or safety risks.

[0038] Step S210: When the pose information satisfies the elevator entry constraint conditions, the robot is controlled to enter the elevator according to the preset edge constraint entry strategy.

[0039] The elevator entry constraints refer to the conditions required for the robot to enter the elevator from its current position, including but not limited to: the horizontal distance between the robot and the elevator door being within a preset reasonable range, the angle between the robot's heading and the elevator door conforming to a preset standard, and the robot's lateral offset not exceeding a preset threshold. The aforementioned preset edge-constrained entry strategy refers to a motion control strategy that uses the elevator door frame or door edge as a constraint reference, controlling the robot to maintain a relative distance and heading angle from the reference, and smoothly enter the elevator car along the reference.

[0040] Furthermore, through multiple continuous adjustments to the robot's pose information, the robot achieves precise alignment at the elevator door. Once the pose information meets the elevator entry constraints, the robot can switch to the elevator entry mode. After triggering the elevator door to open, the robot is guided to enter the elevator while maintaining a safe posture according to a preset edge constraint entry strategy.

[0041] This application constructs a stable feature point cloud by fusing multiple point clouds, adaptively judges the reliability of the point cloud using a data-driven evaluation mechanism, updates the elevator pose according to the reliability difference, and adopts an edge-following strategy to enter the elevator when the constraints are met. This significantly improves the stability and alignment continuity of elevator pose recognition, enhances the robustness and safety of passage in complex environments, and solves the technical problem of low success rate of robot elevator passage. Specifically, by collecting and fusing multiple consecutive frames of point clouds for pose perception and recognition, the instability of single-frame point clouds in elevator pose recognition can be reduced. By training an evaluation function based on a data-driven evaluation mechanism to assess the reliability of point cloud data, the point cloud screening exhibits stronger adaptability and robustness compared to manually fixed thresholds. By constraining the reliability of point cloud data with preset quality assessment conditions, pose information can be accurately updated based on different data regardless of whether the preset quality assessment conditions are met. Even if the reliability does not reach the preset quality assessment conditions within a short period, a stable pose maintenance strategy is maintained, avoiding issues such as easy interruption of the alignment process or repositioning, thus improving passage efficiency and safety. By constraining the pose with elevator entry constraints, more accurate robot alignment can be ensured. Then, a preset edge-constrained entry strategy is used to control the robot to enter the elevator, ensuring smooth and safe robot communication and ultimately improving the success rate of the robot entering the elevator.

[0042] In some possible embodiments, prior to step S202 described above, the method further includes: S1, control the robot to reach the preset initial alignment position at the elevator entrance according to the preset path; S2, the robot's orientation is adjusted at the preset initial alignment position so that the robot faces the object to be identified within the orientation angle constraint.

[0043] Among them, the orientation angle constraint can refer to the angular position constraint between the robot and the reflective sign. The angle is constrained so that the robot turns to the reflective sign and achieves more accurate scanning. For example, the angle between the robot and the two sides of the reflective sign is constrained to within 30°.

[0044] Specifically, the robot first moves to the preset initial alignment position at the elevator door according to the preset path. The preset initial alignment position can be a position marked by dividing the area on the ground, which can ensure that the LiDAR is in a suitable spatial range for collecting reflective marks, while avoiding problems such as point cloud occlusion and poor viewing angle caused by the robot getting too close to or deviating too far from the elevator door.

[0045] Furthermore, after reaching the preset initial alignment position, the robot performs a small-range attitude rotation in place to fine-tune its orientation, ensuring that the robot's heading angle is within the preset orientation angle constraint range. This guarantees that the main detection direction of the lidar is directly facing the reflective marker, thus ensuring that the reflective marker appears stably in the effective field of view of the lidar. The preset path can be obtained through preset map information or navigation paths. After completing the orientation adjustment, the robot maintains its current stable pose and enters the point cloud acquisition and alignment preparation state, providing stable and consistent observation conditions for subsequent reflective marker recognition, point cloud extraction, multi-frame fusion, and pose calculation.

[0046] In this embodiment, by first reaching the designated initial alignment position according to the preset path and then adjusting the orientation constraint, the robot can have a uniform and standardized initial observation posture every time it enters the elevator alignment process. This effectively avoids the problem of radar being unable to capture the object to be identified, point cloud missing, or recognition failure due to the initial position offset or excessive orientation deviation, thereby improving recognition efficiency.

[0047] In some possible embodiments, step S202 above includes: S2021, When the object to be identified is within the scanning range of the robot, the scanning device collects multiple consecutive frames of raw echo point cloud data of the object to be identified; S2022, preprocess the original echo point cloud data of each frame, and obtain the corresponding object point cloud subset for each frame by filtering through the reflection intensity threshold; S2023, the subset of the point cloud of the identified object in each frame is transformed from the coordinate system of the scanning device to the local coordinate system of the robot, and the point clouds of the identified object in multiple consecutive frames after transformation are fused, and duplicate point clouds are filtered out to obtain the point cloud data of the identified object in multiple consecutive frames in the local coordinate system of the robot.

[0048] In this embodiment, combined with Figure 3 and Figure 4 As shown, a reflective marker is placed directly above the elevator door. Alternatively, it can be placed in other locations on the elevator door, such as directly in front, below, or on the side wall. When the object being identified enters the effective scanning angle and ranging range of the robot's LiDAR, the system triggers a continuous acquisition mode. The LiDAR emits laser beams at a preset fixed scanning frequency, targeting the reflective marker and its surrounding environment, and receives echo signals, for example, 5 times per second. Frame-by-frame raw point cloud data, including 3D coordinates, ranging information, and reflection intensity information, is recorded to obtain multiple consecutive frames of raw echo point cloud data of the reflective marker, providing sufficient temporal data for subsequent processing. Figure 4 In the process, the raw echo point cloud data returned by the reflective mark on the elevator door is obtained through point cloud data acquisition, and the point cloud data generated by the reflective mark is significantly stronger.

[0049] Furthermore, preprocessing operations are sequentially performed on each frame of raw echo point cloud data, including filtering out isolated noise points caused by environmental interference and equipment noise. Furthermore, based on the high reflectivity of reflective markers, preprocessing can be performed according to... Figure 4 The echo intensity is set to a reflection intensity threshold for point cloud filtering. Points with reflection intensity above the threshold are identified as valid reflective markers, while points below the threshold are discarded as background points. This results in a subset of the point cloud containing reflective marker features. For example, the laser reflection intensity of reflective markers is typically between 80 and 100. Therefore, a reflection intensity threshold of 70 is set. When evaluating each point in the original point cloud of the current frame, the reflection intensity of points on reflective markers is mostly 85, 92, and 96, all above the threshold of 70, and are retained. The reflection intensity of stainless steel elevator doors is 40 to 60, walls are approximately 20 to 35, and pedestrian clothing is approximately 10 to 30, all below the threshold of 70, and are directly discarded as background points. By leveraging the physical characteristic that reflective markers have significantly higher reflection intensity than environmental objects such as elevator doors, walls, and pedestrians, setting a discriminative reflection intensity threshold allows for accurate segmentation of the target feature point cloud from the environmental interference point cloud, improving the purity and reliability of the identified object point cloud subset.

[0050] In some examples, based on the pre-calibrated extrinsic parameter matrix between the LiDAR and the robot body, the subset of the object point cloud in each frame can be transformed from the LiDAR coordinate system to the robot's local coordinate system through rotation and translation transformations, achieving a unified description of the point cloud position and the robot's pose. Then, spatial overlay matching is performed on multiple consecutive frames of point clouds in the same coordinate system, removing duplicate point clouds with similar spatial positions and retaining valid feature points, ultimately forming continuous, stable, and non-redundant point cloud data of the reflective marker in the robot's local coordinate system.

[0051] In this embodiment, continuous multi-frame acquisition can compensate for the problems of single-frame point cloud being susceptible to occlusion, abnormal reflection, and sparse sampling, significantly improving the integrity and stability of point cloud data. Reflection intensity threshold-based filtering can effectively separate and identify objects from environmental interference points, improving feature purity and preventing non-target point clouds from interfering with pose calculation. Coordinate transformation achieves a unified reference description between point cloud data and the robot body, ensuring consistency in subsequent pose calculation and motion control. Multi-frame fusion and duplicate point cloud removal enhance feature reliability while reducing data redundancy, improving subsequent point cloud quality evaluation and elevator alignment accuracy, and enhancing the robot's robustness and safety in complex elevator environments.

[0052] In some possible embodiments, step S204 above includes: S2041, Real-time acquisition of the odometer information to obtain robot pose data corresponding to the acquisition time of multiple consecutive frames of point cloud data; S2042, using the robot pose data corresponding to the point cloud data in the first frame as a reference, and combining the robot pose data corresponding to the acquisition time of each frame, perform coordinate transformation on the point cloud data of each subsequent frame, and transform the point cloud data of all frames to the robot coordinate system. S2043, perform fusion processing on the point cloud data of multiple consecutive frames after coordinate transformation, filter out invalid point cloud data, and construct the multi-frame feature point cloud set based on the retained valid point cloud features.

[0053] In some examples, the above-mentioned implementation of acquiring robot odometer information may include: the robot chassis drive unit acquiring wheel encoder pulse data in real time, and statistically analyzing the rotation pulses, rotation angles, and odometer increments of the left and right drive wheels. Combining wheel circumference, reduction ratio, and wheelbase calibration parameters, the chassis displacement and rotational yaw angle increments are calculated in real time. The inertial measurement unit (IMU) outputs angular velocity and attitude angles at high frequency to perform attitude compensation and drift correction on the wheel speed odometer. The wheel speed odometer and IMU data are fused to output the robot's real-time pose (x, y, θ) in the global / local coordinate system frame by frame, i.e., the aforementioned real-time odometer information. Furthermore, the timestamp of the odometer information is aligned with the timestamp of the LiDAR point cloud acquisition, enabling each frame of point cloud to be bound to the corresponding odometer pose at that moment, thereby obtaining robot pose data for subsequent multi-frame point cloud coordinate transformation and fusion.

[0054] Furthermore, the unified transformation of multi-frame point cloud coordinates based on odometry information specifically includes: using the robot's pose at the time of the first frame point cloud acquisition as the reference coordinate system, calculating the rotation matrix and translation vector of the point cloud relative to the reference coordinate system based on the odometry pose changes of each subsequent frame, and uniformly transforming the point clouds of all frames from their respective robot local coordinate systems at the time of acquisition to the same reference coordinate system, eliminating the point cloud position offset caused by the robot's own movement. The unified multi-frame point clouds are then spatially superimposed to filter out invalid point cloud data such as discrete isolated noise points caused by environmental interference and robot jitter. The retained point clouds are then used as effective point cloud features to characterize point cloud quality, constructing a multi-frame feature point cloud set, forming a multi-frame feature point cloud set with higher density, more stable distribution, and stronger integrity for subsequent quality assessment, solving the problems of sparse, jittery, and missing single-frame point clouds. For example, suppose a robot acquires three consecutive frames of reflective marker point clouds at an elevator entrance while simultaneously recording its own pose using odometry. The robot in the first frame of the point cloud is in the reference pose, with translation coordinates X=0, Y=0, and heading angle θ=0°, and is directly used as the reference point cloud. The odometry of the second frame shows that the robot moves slightly forward: X moves forward 5cm, Y has no offset, and heading θ rotates 2°. Based on the pose change, all point clouds in the second frame are translated and rotated in the opposite direction to correct back to the reference coordinate system of the first frame, so that the two frames of point clouds are spatially aligned. The odometry of the third frame shows that the robot has shifted laterally by 3cm and heading θ rotates -1°. Similarly, the odometry pose of the third frame is used to transform the point cloud to the reference coordinate system. After merging and filtering to align the three frames of point clouds, the spatially overlapping parts are deleted as duplicate points, and a few scattered isolated noise points far from the marked area are also removed.

[0055] In this embodiment, by collecting odometer information in real time and matching the robot pose at the time of point cloud acquisition for each frame, the accuracy of coordinate transformation can be ensured, achieving precise unification of multiple frames of local point clouds in the same robot local coordinate system. This solves the problem of pose deviation caused by interference in single-frame point clouds. By eliminating invalid and duplicate point clouds and retaining valid point cloud features, noise points caused by environmental interference are effectively filtered out, improving the integrity and stability of the point cloud set. The final multi-frame feature point cloud set can provide stable and reliable point cloud support for subsequent robot elevator alignment and other operations, avoiding operation interruptions or errors caused by sudden changes in single-frame point clouds, improving the continuity and reliability of robot elevator passage, and increasing the success rate of passage.

[0056] In some possible embodiments, step S206 above includes: S2061, based on the point cloud dataset of the identified object in a series of consecutive frames that have been collected and labeled in history, calculate multi-dimensional feature indicators respectively. The multi-dimensional feature indicators include the point cloud dispersion, the number of point clouds, and the point cloud plane fitting residual. Each frame of point cloud is labeled with a point cloud quality label. S2062, Based on the data-driven evaluation mechanism, the model is trained according to the point cloud dataset and the extracted point cloud discreteness, point cloud quantity and point cloud plane fitting residual to obtain the evaluation function. The output target of the evaluation function is the point cloud quality label that evaluates the reliability of the point cloud data. S2063, obtain the target point cloud discreteness, the number of target point clouds, and the target point cloud plane fitting residual from the multi-frame feature point cloud set; S2064, based on the target point cloud dispersion, the number of target point clouds, and the target point cloud plane fitting residual, the reliability of the current point cloud data is calculated using the evaluation function.

[0057] For training the evaluation function, the first step is to acquire a historically collected and labeled point cloud dataset of reflective markers. This historical dataset includes multiple consecutive frames of reflective marker point cloud data under different environments, such as strong reflection, occlusion, and dynamic interference. Each frame of the point cloud is labeled with a point cloud quality label, which is based on manual judgment or objective standards and is used to characterize the usability of the point cloud for robot localization. Further, for each frame of the reflective marker point cloud in the dataset, three feature indicators are calculated: the discreteness of the reflective marker point cloud (X1), the effective number of reflective marker point clouds (X2), and the fitting residual (X3) obtained by plane fitting calculation of the reflective marker point cloud.

[0058] In some examples, the dataset with extracted X1, X2, and X3 feature indicators and corresponding quality labels can be divided into training and validation sets according to a preset ratio for model training and performance verification. A data-driven evaluation mechanism is adopted, using X1, X2, and X3 of each frame of point cloud in the training set as input features and the corresponding point cloud quality labels as output targets to train the evaluation function. The evaluation function is then validated using the validation set, and the error between the point cloud quality evaluation result output by the function and the labeled label is calculated. If the error exceeds the preset threshold, the training parameters are adjusted and the training steps are repeated until the prediction accuracy of the evaluation function meets the preset requirements. Then, the final values ​​of parameters a, b, c, and d are determined, and the evaluation function training is completed. The formula of the evaluation function is shown in the following formula (1): f(x) = a X1+b X2+c X3+d(1) Where a, b, and c are the weighting coefficients of the corresponding indicators, and d is a constant term.

[0059] Furthermore, for the point cloud data in the current scene, the three feature indices X1, X2, and X3 of the current reflective marker point cloud are calculated from the constructed multi-frame feature point cloud set. Then, the extracted X1, X2, and X3 parameters are substituted into the above equation (1) to obtain the function output value. This output value is the reliability of the current point cloud data. The larger the reliability output value, the more reliable the point cloud data is.

[0060] In this embodiment, a historically collected point cloud dataset with quality labels is constructed to obtain multi-dimensional feature indicators such as point cloud discreteness, point cloud quantity, and plane fitting residual. Based on a data-driven evaluation mechanism, a model is trained to generate a point cloud quality evaluation function, avoiding the subjective limitations of manually setting fixed thresholds. By autonomously learning the intrinsic mapping relationship between multi-dimensional features and point cloud quality, the evaluation function can adapt to the point cloud variation characteristics under different elevator environments and interference conditions. At the same time, by combining point cloud discreteness to represent spatial distribution stability, point cloud quantity to represent feature integrity, and plane fitting residual to represent geometric regularity, a comprehensive quantitative evaluation of point cloud quality is achieved, ensuring the reliability of the output point cloud is objective and accurate. This provides a precise data foundation for judging the effectiveness of subsequent pose calculations and is more conducive to improving the stability and adaptability of the robot in elevator alignment and passage under complex interference scenarios.

[0061] In some possible embodiments, step S208 above includes: S2081, compare the reliability of the current point cloud data with the preset quality assessment conditions, and determine whether the reliability of the current point cloud data meets the preset quality assessment conditions; S2082, if the reliability of the current point cloud data reaches the preset quality assessment condition within the preset time window, then the quality of the current point cloud data is determined to be qualified, and the pose is updated based on the first update strategy. The first update strategy includes updating the pose information of the elevator in robot coordinates according to the multi-frame feature point cloud set. S2083, if the reliability of the current point cloud data does not meet the preset quality assessment condition within the preset time window, then the quality of the current point cloud data is determined to be unqualified, and the pose is updated based on the second update strategy. The second update strategy includes obtaining the pose information corresponding to the point cloud data reliability meeting the preset quality assessment condition in the most recent historical period, and updating the pose information of the elevator in the robot coordinate system in combination with the robot's current odometer information.

[0062] The quality assessment conditions for point clouds are preset, namely, a preset reliability threshold, such as a preset reliability of ≥80 as the standard. At the same time, a preset time window, such as 500ms, is set to judge the stability of reliability and avoid misjudgment caused by single-frame fluctuations.

[0063] In some examples, the reliability of the current point cloud data calculated by the evaluation function is compared in real time with preset quality assessment conditions to determine whether the reliability of the current point cloud meets the standard. The comparison result and corresponding timestamp are recorded. During the comparison, a preset time window is started. If the reliability of the current point cloud data meets the preset quality assessment conditions within the preset time window, the current point cloud data is directly deemed to be of acceptable quality. Then, the first update strategy is executed. Based on the multi-frame feature point cloud set constructed above, the real-time pose information of the elevator relative to the robot's local coordinate system is calculated. The updated real-time pose information directly overwrites the original pose data, completing the pose update. Based on the updated pose information, the robot's position is adjusted to ensure that the pose information completely matches the current actual scene. For example, the quality assessment condition for the point cloud is a reliability threshold ≥ 80 and a preset time window = 500ms. When the quality is qualified, the first update strategy is executed. Based on the multi-frame feature point cloud set, the real-time pose of the elevator is calculated as X = 98cm, Y = 0.5cm, θ = 0.2°. The original pose is directly updated with this real-time pose to complete the first strategy update and ensure that the pose is consistent with the current actual position of the elevator.

[0064] In other examples, if the reliability of the current point cloud does not meet the preset quality assessment conditions within a preset time window, it indicates that the accuracy of the current valid point cloud data is insufficient, and the current point cloud data is deemed unqualified and cannot be used to directly update the pose. A second update strategy is then executed: the elevator pose information corresponding to the most recent point cloud quality assessment from the robot's historical pose cache is extracted—that is, the most recent valid pose with acceptable reliability. This is then combined with the robot's current real-time odometer information to compensate and correct the historical valid pose, obtaining the current elevator pose information in the robot's coordinate system. This completes the pose update, maintaining the continuity and stability of the alignment process and preventing pose information loss or deviation due to short-term point cloud failure. For example, the quality assessment condition for point clouds is a reliability threshold ≥ 80, a preset time window = 500ms, and the most recent qualified pose in the historical cache is: the elevator relative to the robot's local coordinate system X = 100cm, Y = 0cm, and heading angle θ = 0°. When the quality is unqualified, the second update strategy is executed to extract the most recent qualified pose, combine it with the robot's current real-time odometer information (moved forward 2cm, no heading offset), and compensate and correct the historical pose to obtain the current elevator pose as X = 98cm, Y = 0cm, θ = 0°. The corrected pose is used to complete the update, avoiding pose information gaps.

[0065] In this embodiment, by determining the reliability stability through a preset time window, the alignment process can be directly interrupted or re-triggered when the effective point cloud quality that meets the accuracy standard cannot be obtained in a short period of time, which would lead to frequent switching of robot actions. This maintains the continuity and robustness of the alignment process and ensures the stability of elevator alignment. Different pose update strategies are applied according to different alignment conditions. When the quality is qualified, the pose is updated with real-time multi-frame feature point cloud to ensure the real-time and accurate nature of the pose information and to match the actual position changes of the elevator. When the quality is unqualified, the historical effective pose and the current odometer information are combined for compensation and update. Even when the point cloud is temporarily unavailable, the pose maintenance strategy (second update strategy) ensures the continuity of the alignment process.

[0066] In some possible embodiments, step S210 above includes: S2101, determine whether the pose information meets the preset elevator entry constraint conditions, wherein the preset elevator entry constraint conditions include the horizontal distance between the robot and the elevator door, the angle between the robot's heading and the elevator door and / or the robot's lateral offset. S2102, if the pose information satisfies the preset elevator entry constraint conditions, then switch to the preset edge constraint entry strategy, using the elevator door edge as the constraint reference, and control the robot to maintain a preset speed and preset safe distance to enter the elevator along the elevator door edge.

[0067] In this embodiment, after the pose update is completed, it is immediately determined whether the pose information of the robot corresponding to the current point cloud meets the preset elevator entry constraints. Among these constraints, setting the horizontal distance between the robot and the elevator door ensures that the robot can enter the elevator at a preset speed within the waiting time before the door opens; setting the angle between the robot's heading and the elevator door determines whether the deviation generated during the robot's entry from its current position will prevent it from entering the elevator; and setting the lateral offset of the robot determines whether it can enter along the edge.

[0068] Furthermore, if the pose information satisfies the above elevator entry constraints, after the robot arrives at the elevator door and triggers the elevator door to open, the point cloud information corresponding to the reflective mark gradually disappears. The system switches to the edge constraint entry strategy based on lateral lidar, using the edge of the elevator door as the constraint reference, and controls the robot to move at a preset speed while maintaining a preset safe distance from the edge of the elevator door. The robot smoothly enters the elevator along the edge of the door, avoiding deviation or collision throughout the process. For example, it moves forward at a preset speed of 3m / s and at a distance of 20cm from the edge of the elevator door.

[0069] In some examples, if the pose information does not meet the elevator entry constraints, the entry operation is paused and the pose adjustment continues until the pose information meets the constraints. Then, the robot is controlled to enter the elevator based on the edge constraint entry strategy.

[0070] In some examples, during the robot's entry process, odometry information and point cloud data are collected in real time to continuously verify pose stability. If pose deviation or point cloud quality is unacceptable, the robot's motion parameters are immediately adjusted to ensure a smooth entry. When the robot is fully inside the elevator and its pose remains stable, the entry process is considered complete, edge constraint control is stopped, and the robot enters the safe posture maintenance phase inside the elevator.

[0071] In this embodiment, by first determining the matching relationship between the robot's pose information and the elevator entry constraints, and then selectively switching the edge-constrained entry strategy, precise and safe control of the robot's autonomous entry into the elevator is achieved. Using the horizontal distance between the robot and the elevator door, the heading angle, and the lateral offset as constraint judgment indicators, the compliance verification of the robot's relative position and posture with the elevator can be completed before entry, avoiding collisions and jamming caused by excessive position deviations or incorrect orientations, thus ensuring the reliability of the robot's entry. After the constraints are met, switching to the edge-constrained entry strategy based on the elevator door edge allows the robot to overcome the ambiguity of conventional path planning, forming stable constraints based on the fixed door edge. At the same time, with the preset speed and safe distance control, it ensures a smooth movement rhythm and maintains a safe distance from the elevator door and car wall, effectively improving the positioning accuracy and motion stability of the robot when entering the elevator, reducing the passage risk in complex elevator scenarios, and improving passage safety.

[0072] Combination Figure 5 The diagram illustrates the overall process of an optional robotic elevator passage method. First, the robot acquires the elevator's approximate location information and moves to a pre-set approximate alignment position at the elevator door based on a preset map or navigation path. It then adjusts to the corresponding approximate orientation and reaches the elevator's predetermined position. Next, it collects multiple frames of laser point cloud information, performs point cloud regression and filtering, and continues to identify reflective markers. Through reliability analysis, it obtains high-quality, effective point cloud data to determine the elevator door's coordinate position relative to the robot. After correcting horizontal and angular errors, it adjusts its pose for precise alignment. During alignment, each adjustment checks whether the pose information meets the preset elevator entry constraints. If so, the robot begins elevator alignment and continues to identify markers. During this process, the identified point cloud data is also regressed, scored, and filtered until the reflective markers disappear. Upon reaching the door, the robot switches to a preset edge-constrained entry strategy and enters the elevator along the edge.

[0073] According to another aspect of the embodiments of this application, in conjunction with Figure 6 As shown, this application provides a robotic elevator access device, the device comprising: The data acquisition module 601 is used to acquire point cloud data of the identified object in multiple consecutive frames; The feature extraction module 603 is used to convert and fuse the point cloud data of multiple consecutive frames based on the robot's real-time odometer information to construct a multi-frame feature point cloud set. The feature evaluation module 605 is used to evaluate the reliability of the point cloud data based on the multi-frame feature point cloud set using an evaluation function pre-trained based on a data-driven evaluation mechanism. The pose update module 607 is used to compare the reliability of the point cloud data with preset quality assessment conditions, and update the pose information of the elevator in robot coordinates according to the different comparison results and the corresponding update strategy. The motion control module 609 is used to control the robot to enter the elevator according to a preset edge constraint entry strategy when the pose information meets the elevator entry constraint conditions.

[0074] It should be noted that, in this embodiment, the data acquisition module 601 can be used to execute step S202 in this application embodiment, the feature extraction module 603 in this embodiment can be used to execute step S204 in this application embodiment, the feature evaluation module 605 in this embodiment can be used to execute step S206 in this application embodiment, the pose update module 607 in this embodiment can be used to execute step S208 in this application embodiment, and the motion control module 609 in this embodiment can be used to execute step S210 in this application embodiment.

[0075] It should be noted that the examples and application scenarios implemented by the above modules and corresponding steps are the same, but are not limited to the content disclosed in the above embodiments. It should also be noted that the above modules, as part of a device, can operate in environments such as... Figure 1 The hardware environment shown can be implemented either through software or through hardware.

[0076] It should be noted that the suffixes such as "module" used to indicate elements in the above-described apparatus are only for the purpose of illustrative purposes and have no specific meaning in themselves. Therefore, they can be used in combination.

[0077] According to another aspect of the embodiments of this application, a computer program product or computer program is also provided, which includes computer instructions stored in a computer-readable storage medium. A processor of a computer device reads the computer instructions from the computer-readable storage medium and executes the computer instructions, causing the computer device to perform the steps of the robotic elevator passage method in any of the above embodiments.

[0078] According to another aspect of the embodiments of this application, this application also provides an electronic device, such as... Figure 7 As shown, the device includes a memory 701, a processor 703, and a network interface 705. The memory 701 stores a computer program that can run on the processor 703. The memory 701 and the processor 703 communicate through the network interface 705 and a communication bus 707. When the electronic device is running, the processor 703 communicates with the memory 701 through the network interface 705. When the processor 703 executes the computer program, it implements the steps of the above-described robot elevator passage method.

[0079] The memory and processor in the aforementioned electronic device communicate with each other via a communication bus and a communication interface. The communication bus can be a peripheral component interconnect standard (PCI) bus or an extended industry standard structure (EISA) bus, etc. This communication bus can be divided into an address bus, a data bus, a control bus, etc. The memory can include random access memory (RAM) or non-volatile memory, such as at least one disk storage device. Optionally, the memory can also be at least one storage device located remotely from the aforementioned processor. The aforementioned processor can be a general-purpose processor, including a central processing unit (CPU), a network processor (NP), etc.; it can also be a digital signal processor (DSP), an application-specific integrated circuit (ASIC), a field-programmable gate array (FPGA), or other programmable logic devices, discrete gate or transistor logic devices, or discrete hardware components.

[0080] According to another aspect of the embodiments of this application, this application also provides a mobile tool, which includes the electronic device shown above. Among them, mobile vehicles can be all devices with mobility capabilities, including vehicles with autonomous or intelligent driving capabilities (including passenger vehicles (such as cars, buses, coaches, minibuses, etc.), cargo vehicles (such as ordinary trucks, box trucks, trailers, enclosed trucks, tank trucks, flatbed trucks, container trucks, dump trucks, special structure trucks), special vehicles (such as logistics delivery vehicles, automated guided vehicles (AGVs), patrol vehicles, cranes, excavators, bulldozers, loaders, road rollers, off-road engineering vehicles, armored engineering vehicles, sewage treatment vehicles, sanitation vehicles, vacuum trucks, floor cleaning trucks, beverage trucks, sweeping robots, food delivery robots, shopping guide robots, lawnmowers, golf carts, etc.), recreational vehicles (such as amusement vehicles, amusement park autonomous driving devices, balance bikes, etc.), rescue vehicles (such as fire trucks, ambulances, power repair vehicles, engineering emergency rescue vehicles, etc.)) and robots (such as sweeping robots, food delivery robots, etc.).

[0081] It is understood that the embodiments described herein can be implemented using hardware, software, firmware, middleware, microcode, or a combination thereof. For hardware implementation, the processing unit can be implemented in one or more application-specific integrated circuits, digital signal processors, digital signal processing devices, microprocessors, and other electronic units or combinations thereof for performing the functions described herein. For software implementation, the techniques described herein can be implemented by units that perform the functions described herein. Software code can be stored in memory and executed by a processor.

[0082] Those skilled in the art will recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented using electronic hardware, or a combination of computer software and electronic hardware. Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the specific working processes of the systems, devices, and units described above can be referred to the corresponding processes in the foregoing method embodiments, and will not be repeated here.

[0083] In the embodiments provided in this application, it should be understood that the disclosed apparatus and methods can be implemented in other ways. The mutual coupling or direct coupling or communication connection shown or discussed may be an indirect coupling or communication connection through some interface, device or unit, and may be electrical, mechanical or other forms.

[0084] 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. Furthermore, the functional units in the various embodiments of this application may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit.

[0085] If the aforementioned functions are implemented as software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solutions of the embodiments of this application, essentially, or the parts that contribute to the prior art, or parts of the technical solutions, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device to execute all or part of the steps of the methods described in the various embodiments of this application. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, ROM, RAM, magnetic disks, or optical disks.

[0086] It should be noted that, in this document, relational terms such as first, second, etc., are used only to distinguish one entity or operation from another entity or operation. The terms include, encompass, or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, method, article, or apparatus.

[0087] The above description is merely a specific embodiment of this application, enabling those skilled in the art to understand or implement this application. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of this application. Therefore, this application is not to be limited to the embodiments shown herein, but is to be accorded the widest scope consistent with the principles and novel features claimed herein.

Claims

1. A method for navigating a robot elevator, characterized in that, The method includes: Collect point cloud data of the object to be identified in multiple consecutive frames; Based on the robot's real-time odometry information, the point cloud data of multiple consecutive frames are transformed and fused to construct a multi-frame feature point cloud set. The reliability of the point cloud data is evaluated based on the multi-frame feature point cloud set using a pre-trained evaluation function based on a data-driven evaluation mechanism. Based on the reliability of the point cloud data and the preset quality assessment conditions, the elevator pose information in robot coordinates is updated according to the corresponding update strategy matched according to the different comparison results. When the pose information satisfies the elevator entry constraints, the robot is controlled to enter the elevator according to the preset edge constraint entry strategy.

2. The robot elevator passage method according to claim 1, characterized in that, Before acquiring and identifying point cloud data of multiple consecutive frames of the object, the method further includes: The robot is controlled to reach the preset initial alignment position at the elevator entrance according to the preset path; The robot's orientation is adjusted at the preset initial alignment position so that the robot faces the object to be identified within the orientation angle constraint.

3. The robot elevator passage method according to claim 1, characterized in that, The point cloud data of the acquired and identified object, consisting of multiple consecutive frames, includes: When the object to be identified is within the scanning range of the robot, multiple frames of raw echo point cloud data of the object to be identified are collected by the scanning device. The original echo point cloud data of each frame is preprocessed and filtered by reflection intensity threshold to obtain the corresponding object point cloud subset for each frame; The point cloud subset of the identified object in each frame is transformed from the coordinate system of the scanning device to the local coordinate system of the robot, and the point clouds of the identified object in multiple consecutive frames after transformation are fused to remove duplicate point clouds, so as to obtain the point cloud data of the identified object in multiple consecutive frames in the local coordinate system of the robot.

4. The robot elevator passage method according to claim 1, characterized in that, The robot-based real-time odometry information is used to transform and fuse point cloud data from multiple consecutive frames to construct a multi-frame feature point cloud set, including: The odometry information is collected in real time to obtain robot pose data corresponding to the collection time of multiple consecutive frames of point cloud data. Based on the robot pose data corresponding to the point cloud data in the first frame, and combined with the robot pose data corresponding to the acquisition time of each frame, coordinate transformation is performed on the point cloud data of each subsequent frame to transform the point cloud data of all frames into the robot coordinate system. The point cloud data of multiple consecutive frames after coordinate transformation is fused to remove invalid point cloud data, and the set of feature point clouds of the multiple frames is constructed based on the retained valid point cloud features.

5. The robot elevator passage method according to claim 1, characterized in that, The evaluation of the reliability of the point cloud data based on the multi-frame feature point cloud set using a pre-trained evaluation function based on a data-driven evaluation mechanism includes: Based on the point cloud dataset of the identified object in a series of consecutive frames that have been collected and labeled in history, multi-dimensional feature indicators are calculated respectively. The multi-dimensional feature indicators include the point cloud discreteness, the number of points, and the point cloud plane fitting residual. Each frame of point cloud is labeled with a point cloud quality label. Based on the data-driven evaluation mechanism, the model is trained according to the point cloud dataset and the extracted point cloud discreteness, point cloud quantity, and point cloud plane fitting residual to obtain the evaluation function. The output target of the evaluation function is the point cloud quality label that evaluates the reliability of the point cloud data. The discreteness of the target point cloud, the number of target point clouds, and the plane fitting residual of the target point cloud are obtained from the multi-frame feature point cloud set. The reliability of the current point cloud data is calculated using the evaluation function based on the target point cloud dispersion, the number of target point clouds, and the target point cloud plane fitting residual.

6. The robot elevator passage method according to claim 5, characterized in that, The reliability of the point cloud data is compared with preset quality assessment conditions, and the elevator's pose information in robot coordinates is updated according to the corresponding update strategy matched according to the comparison result, including: The reliability of the current point cloud data is compared with the preset quality assessment conditions to determine whether the reliability of the current point cloud data meets the preset quality assessment conditions. If the reliability of the current point cloud data reaches the preset quality assessment condition within the preset time window, the quality of the current point cloud data is determined to be qualified, and the pose is updated based on the first update strategy. The first update strategy includes updating the pose information of the elevator in robot coordinates according to the multi-frame feature point cloud set. If the reliability of the current point cloud data does not meet the preset quality assessment condition within the preset time window, the current point cloud data quality is determined to be unqualified, and the pose is updated based on the second update strategy. The second update strategy includes obtaining the pose information corresponding to the point cloud data reliability meeting the preset quality assessment condition in the most recent historical period, and updating the elevator's pose information in the robot coordinate system in combination with the robot's current odometer information.

7. The robotic elevator passage method according to any one of claims 1 to 6, characterized in that, When the pose information satisfies the elevator entry constraints, the robot is controlled to enter the elevator according to a preset edge constraint entry strategy, including: Determine whether the pose information satisfies the preset elevator entry constraint conditions, wherein the preset elevator entry constraint conditions include the horizontal distance between the robot and the elevator door, the angle between the robot's heading and the elevator door, and / or the robot's lateral offset. If the pose information satisfies the preset elevator entry constraint conditions, then switch to the preset edge constraint entry strategy, using the elevator door edge as the constraint reference, and control the robot to maintain a preset speed and preset safe distance to enter the elevator along the elevator door edge.

8. A robotic elevator access device, characterized in that, The device includes: The data acquisition module is used to acquire point cloud data of the identified object in multiple consecutive frames; The feature extraction module is used to convert and fuse the point cloud data of multiple consecutive frames based on the robot's real-time odometer information to construct a multi-frame feature point cloud set. The feature evaluation module is used to evaluate the reliability of the point cloud data based on the multi-frame feature point cloud set using an evaluation function pre-trained based on a data-driven evaluation mechanism. The pose update module is used to compare the reliability of the point cloud data with preset quality assessment conditions, and update the pose information of the elevator in robot coordinates according to the different comparison results and the corresponding update strategy. The motion control module is used to control the robot to enter the elevator according to a preset edge constraint entry strategy when the pose information meets the elevator entry constraints.

9. An electronic device, comprising: The electronic device comprises a processor, a memory, and a network interface, wherein the memory stores machine-readable instructions executable by the processor, characterized in that: when the electronic device is running, the processor communicates with the memory via the network interface, and the processor executes the machine-readable instructions to perform the steps of the robotic elevator passage method as described in any one of claims 1 to 7.

10. A mobile tool, characterized in that, Includes the electronic device described in claim 9 above.

Images