A self-service robot-based guidance method and system

Abstract

This application relates to the field of patient guidance technology, specifically a patient guidance method and system based on a self-service robot. The system employs a path planning + real-time status monitoring approach for patient guidance. First, a multi-dimensional planning strategy is used for path planning to avoid potentially congested areas as much as possible. Then, to address potential road congestion, a 3D grid map of the hospital is constructed to achieve precise navigation across floors. Radar-visual perception and a grid occupancy model are deeply integrated to analyze dynamic environmental information in real time. Based on the comprehensive recognition results of the grid ahead, the system intelligently determines the congestion status and dynamically selects a strategy of continuing, dynamically traversing, detouring, or waiting. Dynamic traversal uses a risk-deviation scoring mechanism to balance safety and path efficiency. This effectively solves the obstacle avoidance problem in highly dynamic pedestrian environments, significantly improving patient guidance efficiency, traffic reliability, and patient experience, providing a verifiable and implementable intelligent patient guidance solution for smart hospitals.

Description

Technical Field

[0001] This application relates to the field of patient guidance technology, specifically a patient guidance method and system based on a self-service robot. Background Technology

[0002] Patient guidance refers to the process in medical institutions (such as hospitals and clinics) where professionals (usually called "patient guides" or "medical guides") provide guidance and consultation services to patients. Its main purpose is to help patients complete the medical process more efficiently and systematically, improve their medical experience, and reduce the workload of medical staff. With the development of robotics technology, more and more hospitals are beginning to use self-service robots for patient guidance.

[0003] The core challenges of using self-service robots for patient guidance lie in intent understanding, triage, path planning, and guidance. While existing technologies offer mature solutions for intent understanding and triage, they still fall short in path planning and guidance. In some scenarios, after the robot plans its path, guidance may be limited by congestion, preventing the patient from moving forward despite the robot's ability to do so. In such cases, if the robot lacks an effective mechanism for determining and executing detours or handling congestion, both the robot and the patient may experience prolonged, ineffective waiting, thus reducing guidance efficiency and the patient's experience. Summary of the Invention

[0004] In view of this, the purpose of this application is to provide a triage method and system based on a self-service robot to solve the problems in the background art.

[0005] To achieve the above objectives, this application adopts the following technical solution:

[0006] This application discloses a triage method based on a self-service robot, comprising the following steps:

[0007] A three-dimensional grid map of the hospital is obtained, and the target location corresponding to the triage result is obtained. The three-dimensional grid map includes two-dimensional grid maps of multiple floors, and the two-dimensional grid maps of different floors are connected by an elevator grid.

[0008] Based on the current location of the self-service robot and the target location, a route is planned in the three-dimensional grid map to obtain the planned path;

[0009] Control the self-service robot to move along the planned path, and during the movement, acquire the occupied grid output by the occupied grid model inside the self-service robot, and acquire the visual data sequence and radar point cloud data sequence in front of the self-service robot. The occupied grid includes multiple original grids and the occupied probability of multiple original grids.

[0010] The visual data sequence and the radar point cloud data sequence are identified and fused to obtain a fused identification result. The fused identification result is then projected onto an occupied grid to obtain a comprehensive identification result for each occupied grid. The comprehensive identification result includes the identification label, motion speed, and direction.

[0011] A passage strategy is determined based on the comprehensive recognition results of multiple grids in front of the self-service robot, and the self-service robot is controlled to perform guidance based on the passage strategy. The passage strategy includes continuing to drive, dynamic passage, detouring, and waiting.

[0012] In one embodiment of this application, route planning is performed on the three-dimensional grid map based on the current location of the self-service robot and the target location to obtain a planned path, including:

[0013] Construct the edge cost function for global path planning, wherein the mathematical expression of the edge cost function is:

[0014]

[0015] In the formula, Indicates the current grid The corresponding edge cost, For route weight cost, For the current grid With the previous grid European distance, This represents the normalized Euclidean distance. For congestion weighting, For the current grid Historical congestion coefficient, For elevator waiting weight, This represents the elevator grid indicator function. To estimate elevator waiting time, This represents the normalized duration;

[0016] Path planning is performed based on the planner, metaheuristic function, edge cost function, and 3D mesh map to obtain the planned path. The edge cost function is embedded in the cumulative calculation process of the planner to obtain the total path cost.

[0017] In one embodiment of this application, the visual data sequence and the radar point cloud data sequence are identified and fused to obtain a fused identification result, including:

[0018] The visual data at each time point in the visual data sequence is identified based on a pre-built recognition model to obtain a visual recognition sequence, wherein the visual recognition sequence includes recognition labels at multiple time points and the confidence levels of the recognition labels;

[0019] Clustering is performed on the radar point cloud data at each time point in the radar point cloud data sequence to obtain a clustering result sequence, wherein the clustering result sequence includes clustering results at multiple time points;

[0020] The visual recognition sequence and the clustering result sequence are aligned, and the aligned visual recognition sequence and the clustering result sequence are fused based on the point cloud-visual fusion module to obtain a fused sequence, wherein the fused sequence includes the fusion results at multiple time points and the fusion confidence.

[0021] Motion attributes are extracted for each target corresponding to each identification label in the fusion sequence to obtain the motion speed and direction of multiple targets.

[0022] In one embodiment of this application, a passage strategy is determined based on the comprehensive recognition results of multiple grids in front of the self-service robot, including:

[0023] Calculate the average probability of occupancy for multiple grids in front of the self-service robot. ;

[0024] In the average value If the probability is less than the preset probability threshold, it is determined that the space ahead is open and the passage strategy is to continue passing.

[0025] In the average value When the probability is greater than or equal to a preset probability threshold, calculate the average movement speed of the multiple grids in front. and the variance of the direction of motion ;

[0026] In the average value The average motion speed is greater than or equal to a preset probability threshold. The speed is greater than a preset speed threshold, and the variance of the motion direction is greater than a preset speed threshold. When the variance is less than the preset variance threshold, the passage strategy will be determined as dynamic passage.

[0027] In the average value The average motion speed is greater than or equal to a preset probability threshold. If the speed is less than or equal to the preset speed threshold and the duration exceeds the preset duration threshold, the traffic strategy will be determined as detour.

[0028] In the average value The average motion speed is greater than or equal to a preset probability threshold. The speed is greater than a preset speed threshold, and the variance of the motion direction is greater than a preset speed threshold. When the variance is greater than or equal to the preset variance threshold, the passage strategy is set to wait. If other strategy conditions are met during the waiting period, the passage strategy is set to the corresponding option. When the waiting time exceeds the preset duration threshold, the passage strategy is set to detour.

[0029] In one embodiment of this application, when the access strategy is detour, controlling the self-service robot to provide guidance based on the access strategy includes:

[0030] In the three-dimensional grid map, the cost of the grid of the congested section in the current planned path is set to the maximum value, and the planner is returned to update the starting point and re-plan the path to obtain the updated planned path. The self-service robot is then controlled to travel along the updated planned path to perform guidance.

[0031] In one embodiment of this application, when the access strategy is dynamic passage, controlling the self-service robot to perform triage based on the access strategy includes:

[0032] S1, determine the planning domain for the current control cycle, wherein the planning domain includes grids within the target radius surrounding the current grid as candidate grids. ;

[0033] S2, for each candidate grid within the planning domain A comprehensive score is performed, and the candidate grid with the highest comprehensive score that is greater than a preset score threshold is selected as the target grid. The comprehensive score includes a security score and a deviation score.

[0034] S3, based on the preset speed limit calculation model, control the self-service robot to move to the target grid and enter the next control cycle;

[0035] S4. Repeat steps S1-S3 until you get out of the congested area.

[0036] In one embodiment of this application, for each candidate grid within the planning domain A comprehensive evaluation will be conducted, including:

[0037] Calculate each candidate grid in the planning domain Safety rating The security score The mathematical expression is:

[0038]

[0039] In the formula, Indicates the predicted duration. This represents the risk sensitivity coefficient. Represents the target set. Represents the Gaussian probability density. For the target index, fusion confidence for the target For the goal Current location For the goal The motion vector, Representing candidate grids The center coordinates, For the covariance of motion uncertainty;

[0040] Calculate each candidate grid in the planning domain Deviation score The deviation score The mathematical expression is:

[0041]

[0042] In the formula, and All are weighting factors. The direction angle of the slide rail path, To shift from the current position to the candidate mesh Candidate orientation angles, Candidate grid With global path The minimum vertical distance, This is the normalized distance parameter;

[0043] Based on the security score and the deviation score Calculate each candidate grid Overall rating The comprehensive score The mathematical expression is:

[0044]

[0045] In the formula, This is the weighting factor.

[0046] In one embodiment of this application, the mathematical expression of the speed limit calculation model is:

[0047]

[0048] In the formula, Indicates the upper limit of speed. This indicates the absolute safe speed limit. For speed following coefficient, The average velocity of the grid in front. This is the velocity offset. Based on speed, This is the density attenuation coefficient. This indicates the density of people in a given area.

[0049] In one embodiment of this application, the average value of the occupancy probability of multiple preceding grids over multiple consecutive control cycles. If all probabilities are less than the preset probability threshold, it is determined that the person has passed through the congested area.

[0050] This application also provides a triage system based on a self-service robot, including:

[0051] The acquisition module is used to acquire a three-dimensional grid map of the hospital and acquire the target location corresponding to the triage results. The three-dimensional grid map includes two-dimensional grid maps of multiple floors, and the two-dimensional grid maps of different floors can be switched through elevator grids.

[0052] The path planning module is used to plan a route on the three-dimensional grid map based on the current location of the self-service robot and the target location, and obtain the planned path.

[0053] The movement control module is used to control the self-service robot to move along the planned path, and during the movement, it acquires the occupied grid output by the occupied grid model inside the self-service robot, and acquires the visual data sequence and radar point cloud data sequence in front of the self-service robot. The occupied grid includes multiple original grids and the occupied probability of multiple original grids.

[0054] The feature extraction module is used to identify and fuse the visual data sequence and the radar point cloud data sequence to obtain a fused identification result, and to project the fused identification result onto the occupied grid to obtain a comprehensive identification result for each occupied grid, wherein the comprehensive identification result includes identification label, motion speed and direction;

[0055] The triage control module is used to determine the passage strategy based on the comprehensive recognition results of multiple grids in front of the self-service robot, and to control the self-service robot to perform triage based on the passage strategy. The passage strategy includes continuing to drive, dynamic passage, detouring, and waiting.

[0056] The beneficial effects of this application are as follows: This application discloses a patient guidance method and system based on a self-service robot, which generally adopts a path planning + real-time status monitoring approach to achieve patient guidance. First, path planning is performed based on a multi-dimensional planning strategy to avoid potentially congested areas as much as possible. Then, for potential road congestion, a three-dimensional grid map of the hospital is constructed to achieve precise navigation across floors, and radar-visual perception and a grid occupancy model are deeply integrated to analyze dynamic environmental information in real time. Based on the comprehensive recognition results of the grid ahead, the system intelligently determines the congestion status and dynamically selects a strategy of continuing, dynamically traversing, detouring, or waiting. Dynamic traversal employs a risk-deviation scoring mechanism to balance safety and path efficiency. This effectively solves the obstacle avoidance problem in highly dynamic pedestrian environments, significantly improves patient guidance efficiency, traffic reliability, and patient experience, and provides a verifiable and implementable intelligent patient guidance solution for smart hospitals. Attached Figure Description

[0057] The present application will be further described below with reference to the accompanying drawings and embodiments:

[0058] Figure 1 This is a logical framework diagram of an intelligent triage system shown in one embodiment of this application;

[0059] Figure 2 This is a logical framework diagram of a triage system based on a self-service robot according to one embodiment of this application;

[0060] Figure 3 This is a flowchart illustrating a triage method based on a self-service robot in one embodiment of this application;

[0061] Figure 4 This is a schematic diagram of the perception data fusion process in one embodiment of this application;

[0062] Figure 5 This is a flowchart of the traffic strategy analysis in one embodiment of this application.

[0063] Figure 6 This is a structural diagram of a triage system based on a self-service robot, as shown in one embodiment of this application. Detailed Implementation

[0064] The following specific examples illustrate the implementation of this application. Those skilled in the art can easily understand other advantages and effects of this application from the content disclosed in this specification. This application can also be implemented or applied through other different specific embodiments, and various details in this specification can also be modified or changed based on different viewpoints and applications without departing from the spirit of this application. It should be noted that, unless otherwise specified, the following embodiments and features in the embodiments can be combined with each other.

[0065] It should be noted that the illustrations provided in the following embodiments are only schematic representations of the basic concept of this application. Therefore, the drawings only show the layers related to this application and are not drawn according to the actual number, shape and size of the layers in the actual implementation. In the actual implementation, the form, number and proportion of each layer can be arbitrarily changed, and the layer layout may also be more complex.

[0066] Numerous details are explored in the following description to provide a more thorough explanation of embodiments of this application; however, it will be apparent to those skilled in the art that embodiments of this application may be practiced without these specific details.

[0067] Figure 1 This is a logical framework diagram of an intelligent triage system shown in one embodiment of this application, such as... Figure 1 As shown, the complete patient guidance system of this application includes a voice receiving module 110, an intent understanding module 120, a triage module 130, and a path planning and guidance module 140. The voice receiving module 110 receives the user's voice information and converts it into text. The intent understanding module 120 extracts intent information from the text. The triage module 130 performs triage based on the intent information, determining the target department and address. The path planning and guidance module 140 performs path planning according to the target department and address, and controls the patient guidance robot to perform guidance actions.

[0068] This application primarily focuses on innovations in the path planning and guidance modules. Figure 2 This is a logical framework diagram of a triage system based on a self-service robot according to one embodiment of this application, such as... Figure 2 As shown, the triage system in this application includes a cloud-based path planning component 210 and a self-service robot 220. The path planning component 210 is deployed on a server, performs path planning by receiving address information uploaded by the self-service robot 220, and distributes the planned path to the self-service robot 220. The self-service robot 220 includes a sensing unit 221, a decision-making unit 222, and an execution unit 223. The decision-making unit 222 controls the execution unit 223 to move according to the planned path, and senses environmental information through the sensing unit 221 during the movement. The decision-making unit 222 formulates new decisions based on the environmental information and executes them through the execution unit 223.

[0069] Figure 3 This is a flowchart illustrating a triage method based on a self-service robot in one embodiment of this application, as shown below. Figure 3 As shown, a triage method based on a self-service robot in this application mainly includes the following steps:

[0070] S310, Obtain a three-dimensional grid map of the hospital and obtain the target location corresponding to the triage result, wherein the three-dimensional grid map includes two-dimensional grid maps of multiple floors, and the two-dimensional grid maps of different floors are switched through elevator grids;

[0071] In this application, the hospital environment is modeled as a hierarchical graph structure, and its mathematical expression is:

[0072]

[0073] In the formula, Represents a 3D grid map. For floor indexing, For the number of floors, Indicates the first Two-dimensional grid map, Connect the edge set for the elevator.

[0074] In this application, the subsequent actual path planning adopts the strategy of "2D planning within the floor + elevator node jump" to avoid the explosion of real 3D grid calculation.

[0075] S320, Based on the current location of the self-service robot and the target location, route planning is performed on the three-dimensional grid map to obtain the planned path;

[0076] This application uses the A* algorithm for path planning. The execution of the A* algorithm depends on the cost function and the metaheuristic function. The path planning process based on the A* algorithm includes:

[0077] S321, Construct the edge cost function for global path planning, wherein the mathematical expression of the edge cost function is:

[0078]

[0079] In the formula, Indicates the current grid The corresponding edge cost, For route weight cost, For the current grid With the previous grid European distance, This represents the normalized Euclidean distance. For congestion weighting, For the current grid Historical congestion coefficient, For elevator waiting weight, This represents the elevator grid indicator function (1 if the grid is an elevator, 0 otherwise). To estimate elevator waiting time, This represents the normalized duration;

[0080] Among them, distance cost term The system penalizes the longer planned path and encourages the selection of a shorter path.

[0081] Congestion cost item In This is based on historical hospital traffic data (such as the average number of people passing through per hour) and normalized to [0, 1], used to penalize grids with congested pathways.

[0082] Elevator waiting area Activation is only performed at the elevator node to reflect the time cost of traveling across floors. If the hospital's elevator management system can be accessed to read the current waiting time of all elevators, the algorithm can be estimated to select the elevator with the shorter waiting time.

[0083] S322, Path planning is performed based on the planner, metaheuristic function, edge cost function and the three-dimensional grid map to obtain the planned path, wherein the edge cost function is embedded in the cumulative calculation process of the planner to obtain the total path cost.

[0084] In this application, the metaheuristic function adopts an acceptable normalized Euclidean distance heuristic, for example:

[0085]

[0086] In the formula, Indicates starting from the current position To the target location The straight-line distance This represents the maximum displacement distance in a single step. This represents the minimum waiting time for the elevator (a conservative estimate of the lower limit). This refers to the elevator's one-way travel time. Indicates the current floor and the target floor. The difference in the number of floors, This represents the maximum total time taken for the elevator.

[0087] The path planning method based on the A* planner, metaheuristic function, edge cost function, and 3D mesh map can be briefly described as follows:

[0088] initialization:

[0089] Add the starting point s to the open list ( ), assuming cumulative cost , Close list ( () is empty.

[0090] Main loop:

[0091] from Take out smallest node ;

[0092] like For the target point, backtrack the path generated by the parent pointer, and the process ends;

[0093] Otherwise, move in and visit its neighbors ;

[0094] Neighbor expansion: For each legal neighbor (Non-obstacle, within the map):

[0095] Calculate edge cost (Using the normalization formula above)

[0096] Calculate the new total cost ;

[0097] like Not accessed or :

[0098] renew ;

[0099] set up ;

[0100] Will join in (or update its priority);

[0101] Termination conditions:

[0102] Target found → Return to optimal path;

[0103] Empty → No feasible path (triggers exception handling).

[0104] In the above process, To the point The cost, Starting point Metaheuristic cost estimation or This is the cumulative cost.

[0105] S330, control the self-service robot to move along the planned path, and during the movement, acquire the occupied grid output by the occupied grid model inside the self-service robot, and acquire the visual data sequence and radar point cloud data sequence in front of the self-service robot, wherein the occupied grid includes multiple original grids and the occupied probability of multiple original grids.

[0106] After obtaining the planned path, the path is sent to the robot. The planned path includes multiple waypoints (path grid). The robot moves forward point by point according to its own positioning information and the coordinates of the waypoints, and prompts the patient to follow as it moves forward.

[0107] During movement, the sensing unit (including a LiDAR module and a camera) collects information about the environment ahead and inputs it into the internally integrated occupancy grid model, outputting occupancy grid data. The occupancy grid data includes the occupancy probability of multiple grids. .

[0108] In this application, in order to achieve local / dynamic occupancy grids output by the occupancy grid model, which are typically aligned with the grids in the global 3D grid map, prior calibration is required, including:

[0109] (1) Unified coordinate system and resolution

[0110] The global 3D mesh map (for path planning) and the local occupied mesh (for real-time obstacle avoidance) must use the same origin, coordinate axis direction, and mesh resolution (e.g., 0.1 m / mesh).

[0111] For example, if the global map takes a pillar in the lobby of the hospital as its origin (0,0,0) and has a resolution of 0.1m, then the local occupancy grid built inside the robot must also be based on this.

[0112] (2) Pose alignment (Localization)

[0113] The robot needs to estimate its precise pose in the global map in real time through a localization module (such as AMCL, laser + IMU fusion). .

[0114] The local occupancy grid is constructed with the robot's current position as the center, and then projected onto the global coordinate system through pose transformation, so that it corresponds one-to-one with the global grid.

[0115] (3) Multi-floor consistency processing

[0116] Each floor's two-dimensional grid map is independently numbered (e.g., floor=2), and the elevator location serves as the cross-floor connection point.

[0117] Locally occupied grids are generated only on the current floor and are aligned with the global grid for that floor.

[0118] In addition, the visual data and point cloud data collected in this application are stored in the form of sequences.

[0119] S340, the visual data sequence and the radar point cloud data sequence are identified and fused to obtain a fused identification result, and the fused identification result is projected onto the occupied grid to obtain a comprehensive identification result for each occupied grid, wherein the comprehensive identification result includes identification label, motion speed and direction;

[0120] Figure 4 This is a schematic diagram of the perception data fusion process in one embodiment of this application, as shown below. Figure 4 As shown, in this application, visual data and radar point cloud data need to be fused and mapped onto a grid to achieve spatiotemporal alignment of the perception data, providing a data foundation for subsequent access decisions. The specific fusion and projection process includes:

[0121] S341, Based on the pre-built recognition model, the visual data at each time point in the visual data sequence is recognized to obtain a visual recognition sequence, wherein the visual recognition sequence includes recognition labels at multiple time points and the confidence level of the recognition labels;

[0122] Specifically, pre-trained deep learning models (such as YOLO, DETR, or Transformer-based detectors) are used to detect objects in each frame of the image, outputting the category label (such as "pedestrian" or "wheelchair"), bounding box, and confidence score for each object. This forms a visual recognition sequence organized by timestamps.

[0123] S342, cluster the radar point cloud data at each time point in the radar point cloud data sequence to obtain a clustering result sequence, wherein the clustering result sequence includes clustering results at multiple time points;

[0124] Specifically, clustering algorithms (such as DBSCAN and EuclideanClustering) are applied to the radar point cloud (such as LiDAR or millimeter-wave radar) at each time moment to group points belonging to the same physical target into one class and output a set of clusters (each cluster represents the location and shape of a potential obstacle).

[0125] S343, Align the visual recognition sequence and the clustering result sequence, and fuse the aligned visual recognition sequence and the clustering result sequence based on the point cloud-visual fusion module to obtain a fused sequence, wherein the fused sequence includes the fusion results at multiple time points and the fusion confidence.

[0126] Specifically, the principles of the alignment and blending described above are as follows:

[0127] (1) Time alignment: By synchronizing sensor timestamps, visual recognition results and radar clustering results at the same physical moment are matched;

[0128] (2) Spatial alignment: Project the visual detection box onto the radar coordinate system (or vice versa) using calibration parameters to establish cross-modal correspondence;

[0129] (2) Fusion strategy: Use existing point cloud-visual fusion modules (such as IoU matching, Kalman filtering, or neural network fusion head) to perform semantic label inheritance (from vision), position / size optimization (from radar) and fusion confidence calculation (such as weighted average or Bayesian update) on the successfully matched objects.

[0130] S344, extract motion attributes for the target corresponding to each identification label in the fusion sequence to obtain the motion speed and direction of multiple targets.

[0131] Specifically, based on the pose (position + orientation) of the same target at multiple time steps in the fused sequence, its instantaneous velocity and motion direction (i.e., two-dimensional / three-dimensional velocity vector) are estimated by difference, linear regression or filtering (such as Kalman Filter).

[0132] S345. By projecting different objects onto the grid according to their positional relationships, a comprehensive recognition result can be obtained.

[0133] Each fused target (including category, location, velocity, and confidence level) is projected onto a predefined 3D mesh map (such as a voxel grid or occupancy grid) based on its world coordinates. Each mesh cell can store: occupancy status, dominant target category, average velocity vector, and fusion confidence level.

[0134] S350, a passage strategy is determined based on the comprehensive recognition results of multiple grids in front of the self-service robot, and the self-service robot is controlled to perform guidance based on the passage strategy. The passage strategy includes continuing to drive, dynamic passage, detour, and waiting.

[0135] The self-service robot in this application needs to continuously assess the surrounding environment and determine its passage strategy during its movement. Although congestion costs have already been factored into the path planning process described above... This is to punish behaviors that pass through historically congested areas. However, hospitals have many pedestrians, and the robots may encounter crowds that prevent them from moving forward. In such cases, a traffic strategy needs to be developed through a process like this.

[0136] Figure 5 This is a flowchart of the access strategy analysis in one embodiment of this application, such as... Figure 5 As shown, the traffic strategy analysis process of this application includes:

[0137] (1) Calculate the average probability of multiple grids in front of the self-service robot being occupied. ;

[0138] (2) In the average value If the probability is less than the preset probability threshold, it is determined that the space ahead is open and the passage strategy is to continue passing.

[0139] In this application, the average value of the occupied probabilities of multiple grids in front of the self-service robot is first used. To determine whether the path ahead is passable, since the self-service robot is mostly in open and passable areas during its movement, directly calculating the movement speed and direction consistency of the grid ahead in each control cycle would cause unnecessary calculations and reduce efficiency.

[0140] In the average value When the probability is less than the preset threshold, it indicates that there are no obstacles ahead, requiring no complex analysis and ensuring smooth passage. Obstacle avoidance for continued movement can employ existing, commonly used obstacle avoidance algorithms, which will not be elaborated upon here.

[0141] (3) In the average value When the probability is greater than or equal to a preset probability threshold, calculate the average movement speed of the multiple grids in front. and the variance of the direction of motion ;

[0142] In the average value A probability value greater than or equal to a preset probability threshold indicates congestion ahead. In potentially congested areas, the dynamic characteristics of the environment are quantified using a two-dimensional approach:

[0143] Average speed of motion Distinguish between static barriers (queuing crowds) and dynamic barriers (flowing pedestrian traffic);

[0144] Variance of direction of motion : Measures the consistency of motion (low variance = ordered flow in the same direction, high variance = chaotic flow with cross paths).

[0145] (3-1) In the average value The average motion speed is greater than or equal to a preset probability threshold. The speed is greater than a preset speed threshold, and the variance of the motion direction is greater than a preset speed threshold. When the variance is less than the preset variance threshold, the passage strategy will be determined as dynamic passage.

[0146] Among them, high speed + low directional variance = stable flow of people in the same direction (such as a one-way corridor), and the robot can predict the gaps in the trajectory to pass through.

[0147] (3-2) In the average value The average motion speed is greater than or equal to a preset probability threshold. If the speed is less than or equal to the preset speed threshold and the duration exceeds the preset duration threshold, the traffic strategy will be determined as detour.

[0148] Low speed + continuous presence = queues / lingering crowds; detouring is more time-efficient than waiting. Alternatively, existing methods of outputting voice prompts plus waiting for environmental updates can be used as a supplement.

[0149] (3-3) In the average value The average motion speed is greater than or equal to a preset probability threshold. The speed is greater than a preset speed threshold, and the variance of the motion direction is greater than a preset speed threshold. When the variance is greater than or equal to the preset variance threshold, the passage strategy is set to wait. If other strategy conditions are met during the waiting period, the passage strategy is set to the corresponding option. When the waiting time exceeds the preset duration threshold, the passage strategy is set to detour.

[0150] High speed + high directional variance = intersecting pedestrian flow / disordered movement (such as intersections), high risk of crossing; wait and observe dynamic evolution, and detour to prevent deadlock if time exceeds the limit.

[0151] In the above strategy, open scenes (accounting for >70%) only require lightweight occupancy probability calculation, avoiding over 90% of motion attribute analysis, extending robot endurance, and meeting real-time control cycle requirements. This is achieved through motion direction variance... Identify high-risk, chaotic scenarios to prevent blindly navigating through intersecting crowds; waiting strategies provide a reaction window for unexpected obstacles, effectively reducing the risk of collisions. Dynamic navigation strategies proactively utilize gaps in crowd flow while ensuring safety, effectively reducing waiting time compared to conservative detour strategies; detour strategies avoid ineffective waiting in front of static obstacles.

[0152] In one embodiment of this application, when the traffic strategy is detour, the self-service robot is controlled to perform guidance based on the traffic strategy, including:

[0153] In the three-dimensional grid map, the cost of the grid of the congested section in the current planned path is set to the maximum value, and the planner is returned to update the starting point and re-plan the path to obtain the updated planned path. The self-service robot is then controlled to travel along the updated planned path to perform guidance.

[0154] If the strategy is to detour, the planner described above is also used for replanning. However, the cost of the grid in the congested area in the existing path cost needs to be maximized so that the planner can automatically select the detour path.

[0155] In one embodiment of this application, when the access strategy is dynamic passage, controlling the self-service robot to perform triage based on the access strategy includes:

[0156] S1, determine the planning domain for the current control cycle, wherein the planning domain includes grids within the target radius surrounding the current grid as candidate grids. ;

[0157] Specifically, the 3×3 area surrounding the current grid is defined as the planning domain, and the grids within the planning domain are candidate grids. This range covers the area reachable by the robot in a single step (typically one grid step), which aligns with the short-term local replanning concept.

[0158] S2, for each candidate grid within the planning domain A comprehensive score is performed, and the candidate grid with the highest comprehensive score that is greater than a preset score threshold is selected as the target grid. The comprehensive score includes a security score and a deviation score.

[0159] The comprehensive scoring process includes:

[0160] S21, Calculate each candidate grid in the planning domain. Safety rating The security score The mathematical expression is:

[0161]

[0162] In the formula, Indicates the predicted duration. This represents the risk sensitivity coefficient. Represents the target set. Represents the Gaussian probability density. For the target index, fusion confidence for the target For the goal Current location For the goal The motion vector, Representing candidate grids The center coordinates, For the covariance of motion uncertainty;

[0163] The above calculation formula is based on kinematic prediction ( The model uses Gaussian uncertainty modeling (after position) to convert multi-objective risk density into safe return value (value ∈ (0,1], the larger the value, the safer the return) through exponential decay.

[0164] Low confidence target ( Small (increased) levels have a weak impact on scoring and suppress perceived false alarm interference. The index form conforms to the cognitive model that "risk increases exponentially with density," and the parameters... Adjustable sensitivity. Additionally, a security score. Naturally normalize to (0,1] to avoid decision-making anomalies caused by negative values / out-of-bounds errors.

[0165] S22, Calculate each candidate grid in the planning domain. Deviation score The deviation score The mathematical expression is:

[0166]

[0167] In the formula, and All are weighting factors. The direction angle of the slide rail path, To shift from the current position to the candidate mesh Candidate orientation angles, Candidate grid With global path The minimum vertical distance, This is the normalized distance parameter;

[0168] The two-dimensional deviations between the candidate grid and the global path are quantified: azimuth deviation (normalized to [0,1]) and vertical distance deviation (normalized to [0,1]). This prevents local decisions from causing trajectory divergence and ensures that the trajectory remains anchored to the global target throughout the traversal process.

[0169] S23, based on the security score and the deviation score Calculate each candidate grid Overall rating The comprehensive score The mathematical expression is:

[0170]

[0171] In the formula, This is the weighting factor.

[0172] Finally, the security score was evaluated. and the deviation score The scores are weighted to obtain a comprehensive score. Finally, the grid with the highest score (indicating passage is possible) is selected as the target grid for the next step.

[0173] S3, based on the preset speed limit calculation model, control the self-service robot to move to the target grid and enter the next control cycle;

[0174] Specifically, the mathematical expression for the speed limit calculation model is as follows:

[0175]

[0176] In the formula, Indicates the upper limit of speed. This indicates the absolute safe speed limit. For speed following coefficient, The average velocity of the grid in front. This is the velocity offset. Based on speed, This is the density attenuation coefficient. The local pedestrian density can be calculated using the comprehensive identification results mentioned above.

[0177] The above speed limit calculation model uses a "minimum value" mechanism to ensure that the robot speed simultaneously satisfies three independent constraints:

[0178] 1. Absolute safety constraints: ;

[0179] 2. Constraints on pedestrian flow coordination: When pedestrian flow accelerates, the robot speeds up appropriately to avoid blocking the passageway; when pedestrian flow slows down, the robot slows down synchronously to reduce the risk of "rear-end collisions" and the feeling of pressure on users.

[0180] 3. Density suppression constraint: based on local pedestrian density The exponential decay rate actively slows down before entering high-density areas to avoid the sudden risk of "rushing into the crowd at high speed".

[0181] S4. Repeat steps S1-S3 until you get out of the congested area.

[0182] Specifically, the average value of the occupied probability of multiple preceding grids over multiple consecutive control cycles. If all probabilities are less than the preset probability threshold, it is determined that the person has passed through the congested area.

[0183] Specifically, the average occupancy probability of multiple grids preceding each of the three consecutive control cycles. When all values ​​are less than the preset probability threshold, it means that the vehicle has already left the congested area.

[0184] The above solution can effectively address the patient guidance problem in congested areas, avoiding issues such as patients not understanding the system and low efficiency caused by robots staying on the system for extended periods.

[0185] This application presents a patient guidance method based on a self-service robot, which generally employs a path planning + real-time status monitoring approach. First, a multi-dimensional planning strategy is used for path planning to avoid potentially congested areas as much as possible. Then, to address potential road congestion, a 3D grid map of the hospital is constructed to achieve precise navigation across floors. Furthermore, radar-visual perception and a grid occupancy model are deeply integrated to analyze dynamic environmental information in real time. Based on the comprehensive recognition results of the grid ahead, the system intelligently determines the congestion status and dynamically selects a strategy of continuing, dynamically traversing, detouring, or waiting. Dynamic traversal employs a risk-deviation scoring mechanism to balance safety and path efficiency. This effectively solves the obstacle avoidance problem in highly dynamic pedestrian environments, significantly improving patient guidance efficiency, traffic reliability, and patient experience, providing a verifiable and implementable intelligent patient guidance solution for smart hospitals.

[0186] like Figure 6 As shown, this application also provides a triage system based on a self-service robot, including:

[0187] The acquisition module is used to acquire a three-dimensional grid map of the hospital and acquire the target location corresponding to the triage results. The three-dimensional grid map includes two-dimensional grid maps of multiple floors, and the two-dimensional grid maps of different floors can be switched through elevator grids.

[0188] The path planning module is used to plan a route on the three-dimensional grid map based on the current location of the self-service robot and the target location, and obtain the planned path.

[0189] The movement control module is used to control the self-service robot to move along the planned path, and during the movement, it acquires the occupied grid output by the occupied grid model inside the self-service robot, and acquires the visual data sequence and radar point cloud data sequence in front of the self-service robot. The occupied grid includes multiple original grids and the occupied probability of multiple original grids.

[0190] The feature extraction module is used to identify and fuse the visual data sequence and the radar point cloud data sequence to obtain a fused identification result, and to project the fused identification result onto the occupied grid to obtain a comprehensive identification result for each occupied grid, wherein the comprehensive identification result includes identification label, motion speed and direction;

[0191] The triage control module is used to determine the passage strategy based on the comprehensive recognition results of multiple grids in front of the self-service robot, and to control the self-service robot to perform triage based on the passage strategy. The passage strategy includes continuing to drive, dynamic passage, detouring, and waiting.

[0192] This application presents a patient guidance system based on a self-service robot, which employs a path planning + real-time status monitoring approach. First, a multi-dimensional planning strategy is used for path planning to avoid potentially congested areas. Then, to address potential road congestion, a 3D grid map of the hospital is constructed for precise cross-floor navigation. This system deeply integrates radar-visual perception with a grid occupancy model to analyze dynamic environmental information in real time. Based on the comprehensive recognition results of the grid ahead, the system intelligently determines the congestion status and dynamically selects a strategy: continue driving, dynamically navigate, detour, or wait. Dynamic navigation employs a risk-deviation scoring mechanism to balance safety and path efficiency. This effectively solves the obstacle avoidance problem in highly dynamic pedestrian environments, significantly improving patient guidance efficiency, traffic reliability, and patient experience, providing a verifiable and implementable intelligent patient guidance solution for smart hospitals.

[0193] This embodiment also provides an electronic terminal, including: a processor and a memory;

[0194] The memory is used to store computer programs, and the processor is used to execute the computer programs stored in the memory so that the terminal performs any of the methods in this embodiment.

[0195] As will be understood by those skilled in the art, the computer-readable storage medium described in this embodiment allows for the implementation of all or part of the steps in the above method embodiments by computer program-related hardware. The aforementioned computer program can be stored in a computer-readable storage medium. When executed, the program performs the steps of the above method embodiments; and the aforementioned storage medium includes various media capable of storing program code, such as ROM, RAM, magnetic disks, or optical disks.

[0196] The electronic terminal provided in this embodiment includes a processor, a memory, a transceiver, and a communication interface. The memory and the communication interface are connected to the processor and the transceiver and complete communication between them. The memory is used to store computer programs, the communication interface is used to perform communication, and the processor and the transceiver are used to run the computer programs, so that the electronic terminal performs the steps of the above method.

[0197] In this embodiment, the memory may include random access memory (RAM) and may also include non-volatile memory, such as at least one disk storage device.

[0198] The processors mentioned above can be general-purpose processors, including central processing units (CPUs), network processors (NPs), etc.; they can also be digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, or discrete hardware components.

[0199] In the above embodiments, although the present application has been described in conjunction with specific embodiments thereof, many substitutions, modifications, and variations of these embodiments will be apparent to those skilled in the art based on the foregoing description. The embodiments of the present application are intended to cover all such substitutions, modifications, and variations falling within the broad scope of the appended claims.

[0200] The above embodiments are merely illustrative of the principles and effects of this application and are not intended to limit this application. Any person skilled in the art can modify or alter the above embodiments without departing from the spirit and scope of this application. Therefore, all equivalent modifications or alterations made by those skilled in the art without departing from the spirit and technical concept disclosed in this application should still be covered by the claims of this application.

Claims

1. A triage method based on a self-service robot, characterized in that, Including the following steps: A three-dimensional grid map of the hospital is obtained, and the target location corresponding to the triage result is obtained. The three-dimensional grid map includes two-dimensional grid maps of multiple floors, and the two-dimensional grid maps of different floors are connected by an elevator grid. Based on the current location of the self-service robot and the target location, a route is planned in the three-dimensional grid map to obtain the planned path; Control the self-service robot to move along the planned path, and during the movement, acquire the occupied grid output by the occupied grid model inside the self-service robot, and acquire the visual data sequence and radar point cloud data sequence in front of the self-service robot. The occupied grid includes multiple original grids and the occupied probability of multiple original grids. The visual data sequence and the radar point cloud data sequence are identified and fused to obtain a fused identification result. The fused identification result is then projected onto an occupied grid to obtain a comprehensive identification result for each occupied grid. The comprehensive identification result includes the identification label, motion speed, and direction. A passage strategy is determined based on the comprehensive recognition results of multiple grids in front of the self-service robot, and the self-service robot is controlled to perform guidance based on the passage strategy. The passage strategy includes continuing to drive, dynamic passage, detouring, and waiting.

2. The triage method based on a self-service robot according to claim 1, characterized in that, Based on the current location of the self-service robot and the target location, route planning is performed on the 3D grid map to obtain the planned path, including: Construct the edge cost function for global path planning, wherein the mathematical expression of the edge cost function is: In the formula, Indicates the current grid The corresponding edge cost, For route weight cost, For the current grid With the previous grid European distance, This represents the normalized Euclidean distance. For congestion weighting, For the current grid Historical congestion coefficient, For elevator waiting weight, This represents the elevator grid indicator function. To estimate elevator waiting time, This represents the normalized duration; Path planning is performed based on the planner, metaheuristic function, edge cost function, and 3D mesh map to obtain the planned path. The edge cost function is embedded in the cumulative calculation process of the planner to obtain the total path cost.

3. The triage method based on a self-service robot according to claim 1, characterized in that, The visual data sequence and radar point cloud data sequence are identified and fused to obtain a fused identification result, including: The visual data at each time point in the visual data sequence is identified based on a pre-built recognition model to obtain a visual recognition sequence, wherein the visual recognition sequence includes recognition labels at multiple time points and the confidence levels of the recognition labels; Clustering is performed on the radar point cloud data at each time point in the radar point cloud data sequence to obtain a clustering result sequence, wherein the clustering result sequence includes clustering results at multiple time points; The visual recognition sequence and the clustering result sequence are aligned, and the aligned visual recognition sequence and the clustering result sequence are fused based on the point cloud-visual fusion module to obtain a fused sequence, wherein the fused sequence includes the fusion results at multiple time points and the fusion confidence. Motion attributes are extracted for each target corresponding to each identification label in the fusion sequence to obtain the motion speed and direction of multiple targets.

4. The triage method based on a self-service robot according to claim 3, characterized in that, The passage strategy is determined based on the comprehensive recognition results of multiple grids in front of the self-service robot, including: Calculate the average probability of occupancy for multiple grids in front of the self-service robot. ; In the average value If the probability is less than the preset probability threshold, it is determined that the space ahead is open and the passage strategy is to continue passing. In the average value When the probability is greater than or equal to a preset probability threshold, calculate the average movement speed of the multiple grids in front. and the variance of the direction of motion ; In the average value The average motion speed is greater than or equal to a preset probability threshold. The speed is greater than a preset speed threshold, and the variance of the motion direction is greater than a preset speed threshold. When the variance is less than the preset variance threshold, the passage strategy will be determined as dynamic passage. In the average value The average motion speed is greater than or equal to a preset probability threshold. If the speed is less than or equal to the preset speed threshold and the duration exceeds the preset duration threshold, the traffic strategy will be determined as detour. In the average value The average motion speed is greater than or equal to a preset probability threshold. The speed is greater than a preset speed threshold, and the variance of the motion direction is greater than a preset speed threshold. When the variance is greater than or equal to the preset variance threshold, the passage strategy is set to wait. If other strategy conditions are met during the waiting period, the passage strategy is set to the corresponding option. When the waiting time exceeds the preset duration threshold, the passage strategy is set to detour.

5. A triage method based on a self-service robot according to claim 2, characterized in that, When the traffic strategy is detour, the self-service robot is controlled to provide guidance based on the traffic strategy, including: In the three-dimensional grid map, the cost of the grid of the congested section in the current planned path is set to the maximum value, and the planner is returned to update the starting point and re-plan the path to obtain the updated planned path. The self-service robot is then controlled to travel along the updated planned path to perform guidance.

6. The triage method based on a self-service robot according to claim 1, characterized in that, When the access strategy is dynamic passage, the self-service robot is controlled to provide guidance based on the access strategy, including: S1, determine the planning domain for the current control cycle, wherein the planning domain includes grids within the target radius surrounding the current grid as candidate grids. ; S2, for each candidate grid within the planning domain A comprehensive score is performed, and the candidate grid with the highest comprehensive score that is greater than a preset score threshold is selected as the target grid. The comprehensive score includes a security score and a deviation score. S3, based on the preset speed limit calculation model, control the self-service robot to move to the target grid and enter the next control cycle; S4. Repeat steps S1-S3 until you get out of the congested area.

7. A triage method based on a self-service robot according to claim 6, characterized in that, For each candidate grid within the planning domain A comprehensive evaluation will be conducted, including: Calculate each candidate grid in the planning domain Safety rating The security score The mathematical expression is: In the formula, Indicates the predicted duration. This represents the risk sensitivity coefficient. Represents the target set. Represents the Gaussian probability density. For the target index, fusion confidence for the target For the goal Current location For the goal The motion vector, Representing candidate grids The center coordinates, For the covariance of motion uncertainty; Calculate each candidate grid in the planning domain Deviation score The deviation score The mathematical expression is: In the formula, and All are weighting factors. The direction angle of the slide rail path, To shift from the current position to the candidate mesh Candidate orientation angles, Candidate grid With global path The minimum vertical distance, This is the normalized distance parameter; Based on the security score and the deviation score Calculate each candidate grid Overall rating The comprehensive score The mathematical expression is: In the formula, This is the weighting factor.

8. A triage method based on a self-service robot according to claim 6, characterized in that, The mathematical expression for the speed limit calculation model is: In the formula, Indicates the upper limit of speed. This indicates the absolute safe speed limit. For speed following coefficient, The average velocity of the grid in front. This is the velocity offset. Based on speed, This is the density attenuation coefficient. This indicates the density of people in a given area.

9. A triage method based on a self-service robot according to claim 6, characterized in that, The average value of the occupied probability of multiple grids in front over multiple consecutive control cycles. If all probabilities are less than the preset probability threshold, it is determined that the person has passed through the congested area.

10. A triage system based on a self-service robot, characterized in that, include: The acquisition module is used to acquire a three-dimensional grid map of the hospital and acquire the target location corresponding to the triage results. The three-dimensional grid map includes two-dimensional grid maps of multiple floors, and the two-dimensional grid maps of different floors can be switched through elevator grids. The path planning module is used to plan a route on the three-dimensional grid map based on the current location of the self-service robot and the target location, and obtain the planned path. The movement control module is used to control the self-service robot to move along the planned path, and during the movement, it acquires the occupied grid output by the occupied grid model inside the self-service robot, and acquires the visual data sequence and radar point cloud data sequence in front of the self-service robot. The occupied grid includes multiple original grids and the occupied probability of multiple original grids. The feature extraction module is used to identify and fuse the visual data sequence and the radar point cloud data sequence to obtain a fused identification result, and to project the fused identification result onto the occupied grid to obtain a comprehensive identification result for each occupied grid, wherein the comprehensive identification result includes identification label, motion speed and direction; The triage control module is used to determine the passage strategy based on the comprehensive recognition results of multiple grids in front of the self-service robot, and to control the self-service robot to perform triage based on the passage strategy. The passage strategy includes continuing to drive, dynamic passage, detouring, and waiting.

Images