Telepresence robot for remote collaboration and consultation support

Abstract

The present application relates to the technical field of telemedicine, in particular to a patrol robot based on remote cooperation and consultation support, which comprises an instruction analysis module, an authority scheduling module, a path switching module, a task execution module and an instruction linkage module.In the present application, the action direction, trigger time period, target coordinates and device part structure in the patrol instruction are disassembled, the continuous task frequency and target range coverage of the doctor's patrol behavior execution record are classified, the doctor's behavior with strong control tendency is sorted, the control focus is gradually focused on the actual master doctor, in the scene of multiple doctors' intervention, the master and cooperative control instruction relationship is established through the synchronous pairing and priority sorting mechanism of voice and focus instruction field, the clarity and independence of the device response link under multiple instruction input are maintained, and the interactive stability and instruction execution structure smoothness of multiple doctor cooperation in remote patrol are comprehensively strengthened.

Description

Technical Field

[0001] This invention relates to the field of telemedicine technology, and in particular to a mobile clinic robot based on remote collaboration and consultation support. Background Technology

[0002] The field of telemedicine technology encompasses systems that utilize information technology and communication networks to achieve long-distance transmission and sharing of medical resources. Its core components include remote diagnosis, remote consultation, remote monitoring, remote surgical guidance, and chronic disease management. By constructing service platforms based on the Internet, the Internet of Things (IoT), and mobile communications, patients can access high-quality medical resources at non-medical institutions or primary healthcare facilities. Telemedicine relies on audio and video transmission technology, clinical data acquisition and transmission equipment, and intelligent scheduling and management methods, covering multiple aspects such as patient health data collection, transmission, storage, analysis, and interaction. It is widely applied in cross-regional collaborative medical services between urban and remote areas, and between primary and higher-level medical institutions, promoting the balanced allocation of medical resources and the construction of a hierarchical medical system.

[0003] Among them, the mobile medical assistance robot supporting remote collaboration and consultation refers to a mobile medical auxiliary device that enables real-time interaction between doctors and patients in different locations through wireless communication and remote control technology, and has the ability to conduct examinations and assist in preliminary diagnosis on-site. The technical matters covered by this patent include multi-source audio and video information acquisition, mobile terminal remote navigation, doctor-end control command parsing, and real-time feedback of consultation data. By integrating a variable-focus high-definition camera component, a high-sensitivity sound pickup device, an image compression encoding module, and a remote feedback interface into the robot platform, the doctor can control the camera angle and microphone sound pickup direction in real time through a remote interface to acquire on-site image and voice information. With the help of mobile chassis navigation, it can move autonomously or under control to the designated location according to the path planning to complete the rounds of visits, while realizing the access of information from remote consultation experts and multi-party voice and video interaction.

[0004] While existing technologies have enabled basic remote operation of telemedicine devices, in scenarios with simultaneous input from multiple experts, there are issues with indiscriminate reception of commands. There is a lack of mechanisms to predict overlap between physical paths and device focus points, leading to frequent interruptions or erroneous responses. In a consultation setting, when one doctor operates the robot to maintain a fixed view, another doctor instructs it to navigate and move. These two commands overlap but have opposing directions, easily causing task execution conflicts. Current systems lack a unified behavioral pattern recognition path in their scheduling mechanisms after doctor command input, ignoring the impact of command frequency and coverage area on actual control intent. They often treat multiple doctors' commands equally, resulting in a lack of hierarchical resource scheduling and reduced consultation command processing efficiency. Regarding path control, existing systems primarily rely on fixed path scheduling, failing to incorporate multi-dimensional service factors such as patient condition and waiting time into the node selection process. This leads to a lack of real-time performance and medical priority response in the selection of rounds nodes. In remote collaboration scenarios, multiple doctors' voices and control focus points can easily compete for control, making it impossible to establish a clear interaction entry point based on control priority, thus affecting the overall collaborative control effect. Summary of the Invention

[0005] To address the problem of indiscriminate instruction reception in existing technologies, which lack a predictive mechanism for overlapping physical paths and device focus between instruction targets, leading to frequent interruptions or erroneous responses in device actions, this invention provides a mobile robot based on remote collaboration and consultation support. In a consultation setting, when one doctor operates the robot to maintain a fixed view, another doctor instructs it to navigate and move. The two action commands overlap but have opposing directions, easily causing task execution conflicts. Current systems lack a unified behavioral pattern recognition path after doctor instruction input, ignoring the reflection of instruction frequency and coverage area on actual control intentions. This often results in the equal processing of instructions from multiple doctors, leading to a lack of hierarchical resource scheduling and reduced consultation instruction processing efficiency. Regarding path control, existing systems primarily rely on fixed path scheduling, failing to incorporate multi-dimensional service factors such as patient condition and waiting time into the node selection process, resulting in a lack of real-time performance and medical priority response capabilities in the selection of patrol nodes. In remote collaboration scenarios, multiple doctors' voices and control focus can easily compete for control, making it impossible to establish a clear interaction entry point based on control priority relationships, thus affecting the overall collaborative control effect. This invention provides a mobile robot based on remote collaboration and consultation support. The technical solution is as follows:

[0006] On the one hand, a mobile clinic robot based on remote collaboration and consultation support was provided, which includes:

[0007] The instruction parsing module is used to extract the action direction field, trigger time period, target position coordinate field and target device part field based on the rounds instructions submitted by the remote consultation terminal, perform direction vector comparison between the action direction field and the navigation direction value in the robot state, and generate a list of parallel control instructions.

[0008] The permission scheduling module is used to extract doctor numbers and consultation stage task execution records through the parallel control instruction list, mark doctors with more than two consecutive tasks as active execution status, mark doctors whose patient number range covers less than two wards as centralized task source doctors, classify task control behaviors, and generate a doctor rounds task control distribution table.

[0009] The path switching module is used to call the doctor's rounds task control distribution table, compare the service status field of the navigation node with the status field of the target node in the set, and generate the change identifier of the rounds node.

[0010] The task execution module is used to use the change identifier content of the patrol node, extract the node path segment instructions, check the availability status of the camera component, the audio pickup component and the signal transmission component, perform device activation operation on the callable components, and generate a patrol task start status registration item.

[0011] As a further aspect of the present invention, the parallel control instruction list includes a control target number sequence, a direction consistency label, a coordinate region pairing result, and a conflict resolution identifier; the doctor's rounds task control distribution table includes an execution frequency classification label, a target region distribution category, a doctor's identity number mapping sequence, and a behavior tendency sorting structure; the rounds node change identifier includes a node update number, a service status label grouping, a navigation redirection identifier, and a path priority execution marker; and the rounds task start status registration item includes a device activation number list, a component status mapping value, a path segment execution identifier, and a task queue update result.

[0012] As a further aspect of the present invention, the instruction parsing module includes:

[0013] The field extraction submodule is used to extract the action direction field, trigger time period, target location coordinate field and target device part field based on the rounds instructions submitted by the remote consultation terminal. The data types are divided and recorded as action vector items and positioning coordinate items respectively, and an instruction data structure set is generated.

[0014] The direction comparison submodule is used to call the action vector items in the instruction data structure set and the navigation direction data to perform vector splitting processing. Based on the difference between the angle between the split direction items and the set direction reference angle, the direction items whose angle falls into the allowable deviation range are marked as navigation inputs, and a direction matching tag set is generated.

[0015] The coordinate matching submodule is used to perform spatial distance calculation based on the positioning coordinate items in the instruction data structure set and the real-time path node coordinate set, calculate the Euclidean distance between coordinate point pairs and perform node attribution determination in combination with trajectory continuity features, mark the attribution node as the target area to be controlled, pair and bind the label matching items with the anchor point set, and generate a parallel path control list.

[0016] As a further aspect of the present invention, the permission scheduling module includes:

[0017] The task extraction submodule is used to extract the corresponding doctor number and consultation stage task execution record based on the parallel control instruction list, number and merge the task records of different time periods under the same doctor number according to the time order, and uniformly mark and filter out interrupted tasks to generate a doctor task list.

[0018] The execution status identification submodule is used to call the length information of each task sequence in the doctor task list, calculate the frequency of consecutive tasks, determine whether the number of consecutive tasks exceeds two, and mark it as an active execution status if the condition is met. It also counts the number of ward numbers covered by the associated patient number in each doctor task, marks doctors with fewer than two ward numbers as centralized task source doctors, and generates a doctor status identification sequence.

[0019] The behavior classification submodule is used to classify doctors in active execution status and doctors from centralized task sources into differentiated task control behavior classification groups according to the doctor status identifier sequence, and to call the task participation time period, task coverage ward number and task type to calculate the task execution time period deviation value, and to compare the horizontal task distribution ratio based on the behavior classification to generate a doctor rounds task control distribution table.

[0020] As a further aspect of the present invention, the deviation value during the task execution period is calculated using the following formula:

[0021] ;

[0022] in, This represents the deviation value during the task execution period. This represents the number of doctors in an active execution state. Representing the The number of real-time task days for each doctor within a real-time task cycle. Representing the The earliest execution time number of the same type of task for each doctor. Representing the The latest execution time number of the same type of task for each doctor. Representing the The doctor's target task time center point.

[0023] As a further aspect of the present invention, the path switching module includes:

[0024] The data processing submodule is used to call the doctor's rounds task control distribution table, extract the patient waiting time record, symptom level code and dense visit record fields corresponding to the navigation node, and merge the three data according to the navigation node number, arrange them into a continuous arrangement structure in chronological order, and generate a node task load set.

[0025] The status comparison submodule is used to call the service status field of the navigation node and the real-time status field of the target node in the set based on the target node number in the node task load set. The service status field of the navigation node and the real-time status field of the target node in the set are encoded and mapped into a unified sequence according to the service occupancy rate, active reception frequency and queue length respectively. The corresponding positions of the status sequences of the navigation node and the target node are compared to generate a node priority switching pair sequence.

[0026] The node adjustment submodule is used to extract the location information, task arrangement time period and patient density frequency of the target node to be replaced in the sequence according to the node priority switching pair sequence. Based on the real-time navigation node path structure, the execution order of replaceable nodes is updated according to the priority switching position and the path segment replacement mark content is marked to generate the patrol node change mark content.

[0027] As a further aspect of the present invention, the task execution module includes:

[0028] The path numbering and sorting submodule is used to extract the marked path segment numbers and corresponding node operation instructions by using the change identifier content of the patrol node, and sort the continuous path segments into a path segment execution sequence according to the node index order, evaluate the trigger index association relationship between the node and the path segment, and generate a path segment instruction sequence set.

[0029] The equipment detection submodule is used to call the path segment instruction sequence set, read the status fields of the camera components, audio pickup components and signal transmission components configured in the path segment, determine whether the component status is marked as callable, remove component entries with unavailable status fields and record node indexes, and generate a callable index group for inspection components.

[0030] The task activation submodule is used to perform device initialization, function interface opening and signal channel connection operations on the camera, audio pickup and signal transmission components respectively, according to the available component number under the node in the callable index group of the roving component, bind the general activation process of the device under the node in the path segment instruction, and generate a roving task start status registration item.

[0031] As a further aspect of the present invention, the robot also includes a command linkage module:

[0032] The instruction linkage module is used to extract the voice instruction fields and operation focus fields that are active in the doctor's interactive interface by using the registration item of the start status of the round of medical tasks, pair the focus fields with the target site fields to control targets, arrange the voice instruction fields, mark the first field of the sort as the master field, mark the remaining fields as the collaboration field, and generate a remote collaborative control instruction set.

[0033] The remote collaborative control instruction set includes the master doctor number, voice control link, synchronous operation mapping structure, and interactive focus association items.

[0034] As a further aspect of the present invention, the instruction linkage module includes:

[0035] The instruction extraction submodule is used to extract the voice instruction fields and operation focus fields that are active in the doctor's interactive interface by using the registration item of the initiation status of the round of visits task, and to filter the fields based on the activation status identifier, calculate the field interaction difference feature value, and generate the interface interaction control sequence.

[0036] The target pairing submodule is used to call the operation focus field in the interface interaction control sequence and the target part field in the patrol node change identifier content, obtain the overlapping area of ​​the focus and part based on the field position code comparison, and generate a control target mapping structure.

[0037] The instruction control submodule is used to rearrange the instruction fields according to the timestamp based on the control target mapping structure, mark the first field as the main control field, and mark the remaining fields as collaborative fields. The main control field is bound to the corresponding target part in the control target mapping structure group and the collaborative fields are linked to generate a remote collaborative control instruction set.

[0038] As a further aspect of the present invention, the formula for calculating the interaction difference feature value of the fields is:

[0039] ;

[0040] in, Represents the interaction difference feature values ​​of the fields. Representing the Normalized weights of the item activation status identifier values Representing the The data item with a positive trigger status value in the voice command field. Representing the The real-time value of the corresponding data item in the focus field of the item operation. Representing the The cumulative number of times an item field is inactive. This represents the total number of fields extracted.

[0041] The beneficial effects of the technical solutions provided in the embodiments of the present invention include at least the following:

[0042] By structurally deconstructing the action direction, trigger time period, target coordinates, and equipment location in rounds instructions, and combining navigation direction values ​​with real-time node coordinates for matching operations, a mechanism for recognizing the parallelism of instructions in rounds tasks is established. A dual comparison of direction vectors and path coordinates is introduced at the initial stage of instruction generation, enabling real-time filtering of rounds instructions from multiple doctors collaborating remotely based on the principle of target conflict avoidance. The execution records of doctors' rounds behavior are categorized by combination of continuous task frequency and target range coverage. After clearly prioritizing doctor behaviors with strong control tendencies, the control focus gradually shifts to the actual controlling doctor. The path target is driven by the behavioral characteristics of the task executor, extracting three status data points—patient waiting time, disease level, and ward visitation density—to achieve dynamic optimization of path nodes. This is then combined with targeted activation of equipment in its callable state, ensuring that the equipment involved in the task path instructions is always responsive. In scenarios where multiple doctors intervene simultaneously, a synchronous pairing and priority ranking mechanism for voice and focus instruction fields establishes the relationship between master control and collaborative control instructions, maintaining a clear and independent device response chain under multiple instruction inputs. This processing logic comprehensively enhances the interactive stability of multi-doctor collaboration and the structural smoothness of instruction execution in remote consultations through measures such as dynamic reconstruction of paths driven by behavioral features, early resolution of conflicting instruction fields, classification and focusing of control intentions, and hierarchical pairing of responses. Attached Figure Description

[0043] To more clearly illustrate the technical solutions in the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0044] Figure 1 This is a schematic diagram of a mobile clinic robot based on remote collaboration and consultation support provided in an embodiment of the present invention;

[0045] Figure 2 This is a schematic diagram of the robot frame of the present invention; Detailed Implementation

[0046] The technical solution of the present invention will now be described with reference to the accompanying drawings.

[0047] In embodiments of the present invention, words such as "exemplarily," "for example," etc., are used to indicate that something is an example, illustration, or description. Any embodiment or design described as "exemplary" in the present invention should not be construed as being more preferred or advantageous than other embodiments or designs. Specifically, the use of the word "exemplary" is intended to present the concept in a concrete manner. Furthermore, in embodiments of the present invention, the meaning expressed by "and / or" can be both, or either one.

[0048] In the embodiments of this invention, the terms "image" and "picture" may sometimes be used interchangeably. It should be noted that, without emphasizing the distinction between them, they convey the same meaning. Similarly, the terms "of," "corresponding (relevant)," and "corresponding" may sometimes be used interchangeably. It should be noted that, without emphasizing the distinction between them, they convey the same meaning.

[0049] In this embodiment of the invention, sometimes a subscript such as W1 may be written in a non-subscript form such as W1. When the difference is not emphasized, the meaning they express is the same.

[0050] To make the technical problems, technical solutions and advantages of the present invention clearer, a detailed description will be given below in conjunction with the accompanying drawings and specific embodiments.

[0051] This invention provides a mobile clinic robot based on remote collaboration and consultation support, such as... Figure 1-2 The diagram shown illustrates a mobile clinic robot based on remote collaboration and consultation support. The robot includes:

[0052] The instruction parsing module is used to extract the action direction field, trigger time period, target position coordinate field and target device part field based on the rounds instructions submitted by the remote consultation terminal. It performs direction vector comparison between the action direction field and the navigation direction value in the robot state, performs coordinate matching between the target position coordinate field and the real-time path node position field, and generates a list of parallel control instructions.

[0053] The permission scheduling module is used to extract doctor IDs and consultation stage task execution records through a list of parallel control instructions. It marks doctors with more than two consecutive tasks as active execution status, marks doctors whose patient ID range covers less than two wards as centralized task source doctors, classifies task control behaviors, performs horizontal comparison and sorting of the classified execution tendencies, and generates a doctor rounds task control distribution table.

[0054] The path switching module is used to call the doctor's rounds task control distribution table, obtain the patient waiting time record, symptom level code and dense visit record fields corresponding to the node, compare the service status field of the navigation node with the status field of the target node in the set, and generate the rounds node change identifier content.

[0055] The task execution module is used to change the identification content of the patrol node, extract the node path segment instructions, check the availability status of the camera component, the audio pickup component and the signal transmission component, activate the device for the component marked as callable, and generate the patrol task start status registration item.

[0056] The instruction linkage module is used to extract the active voice instruction fields and operation focus fields from the doctor's interactive interface by using the registration items of the roving task start status. The focus fields are paired with the target site fields to control targets. The voice instruction fields are arranged, and the first field is marked as the master field and the remaining fields are marked as collaborative fields to generate a remote collaborative control instruction set.

[0057] The parallel control instruction list includes the control target number sequence, direction consistency label, coordinate area pairing result, and conflict elimination identifier. The doctor's rounds task control distribution table includes execution frequency classification label, target area distribution category, doctor identity number mapping sequence, and behavior tendency sorting structure. The rounds node change identifier includes node update number, service status label grouping, navigation redirection identifier, and path priority execution mark. The rounds task start status registration items include the device activation number list, component status mapping value, path segment execution identifier, and task queue update result. The remote collaborative control instruction set includes the master doctor number, voice control link, synchronous operation mapping structure, and interaction focus association item.

[0058] Specifically, such as Figure 2 As shown, the instruction parsing module includes:

[0059] The field extraction submodule extracts the action direction field, trigger time period, target location coordinate field and target device part field based on the rounds instructions submitted by the remote consultation terminal. It divides the data types into action vector items and positioning coordinate items respectively, and generates an instruction data structure set.

[0060] The process requires extracting several key elements from the text, including action direction, trigger time period, target location coordinates, and target device part. In actual execution, the text undergoes preprocessing, including symbol cleaning, syntactic segmentation, and language standardization. A trained named entity recognition model is used to identify semantic units. Verb phrases such as "approach," "move to," and "adjust direction" are identified using a word vector model as action direction-related expressions. The direction vector is determined based on contextual references. For the description "move to the right side of the operating table," "right side" is identified and, combined with a predefined direction dictionary, it is mapped to a standard direction code. The extraction of time periods relies on the time phrase recognition module. Expressions such as "every Wednesday morning" can be converted into specific time period identifiers for unified processing of path planning tasks. The target location coordinates need to be based on the indoor layout map and the calibration of reference object positions. Setting "30 cm to the left of the operating table" can be converted into relative coordinates with the operating table as the origin. Equipment part recognition calls the equipment structure vocabulary, such as "monitor probe" and "infusion pump button". The equipment unit to which it belongs is determined by the joint recognition of keywords and contextual relationships. After the extraction results are encapsulated in standardized fields, they are uniformly summarized into structured data entries to generate an instruction data structure set.

[0061] The direction comparison submodule calls the action vector items in the instruction data structure set and performs vector splitting with the navigation direction data. It judges the difference between the angle between the split direction items and the set direction reference angle, marks the direction items whose angle falls into the allowable range as navigation input, and generates a direction matching label set.

[0062] The direction action field in the round-trip instructions needs to be broken down into direction vector components and compared with the current navigation path direction. In practice, the extracted direction information is expressed in a two-dimensional space. Then, based on the direction changes between path nodes in the navigation record, the direction of each node is extracted to establish the basis for direction comparison. The angle between the round-trip direction and the navigation direction needs to be evaluated. If the angle is within the set allowable deviation range (set to within 30 degrees), the direction is considered to be consistent with the navigation direction. In the application, the system extracts the movement trajectory in the northwest direction in the navigation record for "moving to the left". If the difference between the two directions is small, the trajectory point is marked as a direction matching point. The direction matching label is then added to the matching result set and further participates in the path optimization process. In this judgment process, both direction accuracy and path stability need to be considered. To avoid interference, a minimum action amplitude threshold is set to filter out small offset actions. At the same time, a sliding window mechanism is introduced to jointly judge multiple consecutive direction items to improve the accuracy of direction matching, identify path segments with consistent directions, provide a basis for navigation path selection, and generate a direction matching label set.

[0063] The coordinate matching submodule performs spatial distance calculation based on the positioning coordinate items in the instruction data structure set and the real-time path node coordinate set. It calculates the Euclidean distance between coordinate point pairs and performs node attribution determination in combination with trajectory continuity features. The attribution node is marked as the target area to be controlled. The label matching items are paired and bound with the anchor point set to generate a parallel path control list.

[0064] The target coordinates extracted from the patrol instructions are compared with the distance of the real-time node set in the navigation path to determine whether the target location is within the current control range. During execution, the target coordinates are used as input to traverse the path node set recording points, calculate the spatial interval between them and the target point, and then determine whether they belong to the same area based on the set distance tolerance threshold. To further improve accuracy, the continuity of the trajectory nodes is analyzed to identify the movement change patterns between coordinate points. By calculating the change amplitude of positional differences between continuous nodes, it is determined whether the target point appears on the continuous path segment. For nodes that belong to the same area but are at the turning point of the path, the smoothness of the direction switch is also judged to avoid misidentification. In the example, if the target device part is "monitor probe" and the target coordinates are close to the anchor point marked by the monitor, when there are multiple candidate points in the path, the path point closest to the target anchor point will be selected first to bind and form an effective path control item. The matching relationship also needs to be recorded. In multi-target device instructions, it is necessary to ensure that there is no intersection between the positioning points of each device. The successfully matched path nodes will be added to the parallel path control list.

[0065] Specifically, such as Figure 2 As shown, the permission scheduling module includes:

[0066] The task extraction submodule extracts the corresponding doctor number and consultation stage task execution record based on the parallel control instruction list. It then numbers and merges the task records of different time periods under the same doctor number according to the time order, and uniformly marks and filters out interrupted tasks to generate a doctor task list.

[0067] The robot extracts the corresponding doctor ID and associated consultation stage task execution records from each task. During execution, the robot parses the task entries in the list, identifies the doctor's unique ID, and extracts the doctor ID. It then uses this ID as the primary key to construct doctor task groups. The robot needs to identify the timestamp field in the task records to determine the specific time of each task's execution. Under the same doctor ID, the robot sorts the task records belonging to different time periods according to their chronological order and assigns a sequence number to each task to form a complete task sequence. For entries in the task records that have execution interruption indicators, the robot identifies them through the tag field. If keywords such as "aborted" or "unplanned termination" appear in the task status, the task will be removed from the sequence. The completed doctor task list consists of the doctor ID, task sequence, and corresponding timestamp field. In practical applications, if doctor A executes 5 tasks on different days, the robot organizes the tasks in order according to the recorded time and removes the 3rd task with the interrupted status to obtain the doctor task list.

[0068] The execution status identification submodule calls the length information of each task sequence in the doctor's task list, calculates the frequency of consecutive tasks, and determines whether the number of consecutive tasks exceeds two. If the condition is met, it is marked as an active execution status. The module counts the number of ward numbers covered by the associated patient numbers in each doctor's task, marks doctors with fewer than two ward numbers as centralized task source doctors, and generates a doctor status identification sequence.

[0069] The inspection robot iterates through the task sequences of each doctor in the list, extracts the sequence length information, records the number of consecutive task segments executed by each doctor, and then calculates the frequency of consecutive tasks. If the interval between three consecutive task executions in the task record of doctor D001 is less than a set time threshold (an interval of less than one hour is defined as a consecutive task), the inspection robot determines that the doctor is in an active execution state and marks the doctor's ID as "active". At the same time, the inspection robot reads the patient IDs involved in the tasks associated with the doctor, maps the patient IDs to the ward IDs of the doctor, and counts the number of wards. If the count shows that the total number of wards is less than two, the inspection robot marks the doctor as a "centralized task source" doctor. If all five tasks of doctor B involve patients in the same ward, then the doctor is identified as a centralized task source type. The entire process forms a doctor status identifier sequence after marking.

[0070] The behavior classification submodule categorizes doctors in active execution status and doctors from centralized task sources into differentiated task control behavior classification groups based on the doctor status identifier sequence. It also calls the task participation time period, the ward number covered by the task, and the task type to calculate the task execution time period deviation value. Based on the behavior classification, it compares the horizontal task distribution ratio and generates a doctor's rounds task control distribution table.

[0071] The deviation value during task execution is calculated using the following formula:

[0072] ;

[0073] in, This represents the deviation value during the task execution period. This represents the number of doctors in an active execution state. Representing the The number of real-time task days for each doctor within a real-time task cycle. Representing the The earliest execution time number of the same type of task for each doctor. Representing the The latest execution time number of the same type of task for each doctor. Representing the The doctor's target task time center point;

[0074] Meaning of parameters and derivation of formulas:

[0075] This represents the total number of doctors in an active execution state, which is obtained by extracting the number of doctors with an execution state of "active" from the hospital's dispatching rounds task execution status log;

[0076] Indicates the first The actual number of work days for each doctor in the current task cycle is calculated from the task start and end time fields in the task scheduling record.

[0077] The first doctor's mission will last from May 1st to May 3rd, and will last for 3 days.

[0078] The second doctor: The mission period is from May 2nd to May 4th, and the mission duration is 3 days;

[0079] The third doctor: The mission period is from May 1st to May 2nd, and the mission duration is 2 days;

[0080] Indicates the first The earliest execution time number of the same type of task for each doctor, the numbering rule is to increment the daily numbers within the period starting from 1, with May 1st being number 1, and so on. The data comes from the task database:

[0081] The first doctor: began his mission on May 1st, and was initially numbered 1;

[0082] The second doctor: May 2nd, earliest numbered 2;

[0083] The third doctor: May 1st, earliest numbered 1;

[0084] Indicates the first The latest execution time number of the same type of task for each doctor, the data is also extracted from the task database:

[0085] Doctor #1: Mission ended on May 5th, latest number is 5;

[0086] The second doctor: May 6th, number 6;

[0087] The third doctor: May 3rd, number 3;

[0088] Indicates the first The target task time center point for each doctor within the current cycle, i.e., the preset task standard time point, is data set by the hospital's rounds strategy and uniformly set as the cycle midpoint time (numbered 3):

[0089] Substitute the parameters into the formula and calculate as follows:

[0090] First doctor: ;

[0091] Median value:

[0092] ;

[0093] Second doctor: ;

[0094] Median value:

[0095] ;

[0096] The third doctor: ;

[0097] Median value:

[0098] ;

[0099] Substitute the three intermediate values ​​into the overall formula:

[0100] ;

[0101] The results show that the task execution time deviation value is 6.27, which represents the degree of deviation of the overall time of doctors' task participation from the target center time point. The larger the value, the more dispersed the task execution time is. This value will be used as a weight reference for measuring execution stability in subsequent horizontal task distribution ratio comparisons.

[0102] Specifically, such as Figure 2 As shown, the path switching module includes:

[0103] The data processing submodule calls the doctor's rounds task control distribution table, extracts the patient waiting time record, symptom level code and dense patient record fields corresponding to the navigation node, and merges the three data according to the navigation node number, and arranges them into a continuous arrangement structure in chronological order to generate a node task load set.

[0104] The patrol robot reads the navigation node number associated with each patrol task in the distribution table, and for each navigation node, it calls the relevant database interface to extract the corresponding patient's waiting time record, symptom level code, and dense patient visit record fields. The waiting time record is based on the queuing patrol robot log when the task is triggered, obtaining the total time elapsed for each patient from registration to the current moment, recorded in minutes. The symptom level code is extracted from the automatic classification results of the pre-diagnosis patrol robot of the reception platform, divided into five levels, with Level I for critical, Level II for emergency, Level III for general, etc. The dense patient visit record is obtained by statistically analyzing the number of patients at that node in the past hour. The number of patients received within the time frame is used to determine the three types of data, which are then attached as fields to the data structure of each navigation node. The inspection robot merges these three data items according to the navigation node number and arranges them in chronological order as a continuous time structure based on the timestamp. Each item represents a navigation node and includes the corresponding patient waiting time, current symptom level, and the frequency of patients received per unit time. For example, navigation node A contains three data items: waiting time of 35 minutes, symptom level II, and patient density of 8 times / hour. After sorting, these data items are constructed into a node task load entry, which serves as an important basis for subsequent status judgment and node optimization, forming a node task load set.

[0105] The status comparison submodule, based on the target node number in the node task load set, calls the service status field of the navigation node and the real-time status field of the target node in the set, respectively, and encodes and maps them into a unified sequence according to the service occupancy rate, active reception frequency and queue length. It then compares the corresponding positions of the status sequences of the navigation node and the target node to generate a node priority switching pair sequence.

[0106] The target node number needs to be identified from the node task load set. The inspection robot then sequentially extracts the service status field and real-time status field for this target node. The service status field includes the current doctor's on-duty status, equipment resource utilization, and channel occupancy, while the real-time status field is provided by the current navigation task scheduling results and mainly includes dynamic data such as queue length and patient frequency. The inspection robot maps these two data sources into a unified status evaluation sequence. The mapping method involves encoding fixed indicators in a specific order, with the first indicator representing service utilization, the second representing active patient frequency, and the third representing queue length. Then, each item is assigned a sequence code using a standardized sorting method. The higher the value, the higher the ranking. After the mapping is completed, the inspection robot compares the differences in the ranking of the target node and the current navigation node in each indicator. A large difference in ranking indicates that the target node performs better or worse in some aspects. The comparison results of the indicator differences between the navigation node and the target node are marked. If the target node is better than the current node in both queue length and service utilization rate and the ranking difference reaches two levels or more, then the sequence is marked as "suggested to switch" and this node pair is recorded as a candidate replacement object, generating a node priority switching pair sequence.

[0107] The node adjustment submodule extracts the location information, task scheduling time period and patient density frequency of the target node to be replaced in the sequence according to the node priority switching pair sequence. Based on the real-time navigation node path structure, it updates the execution order of replaceable nodes according to the priority switching position and marks the path segment replacement mark content, generating the patrol node change mark content.

[0108] The system extracts three key elements from each suggested node swap pair: spatial location of the target node, task scheduling time period, and patient frequency. Spatial location information is used to determine if the target node is within the current navigation path. Task scheduling time period is used to determine if there are task conflicts or time overlaps. Patient frequency is used to determine if the node's workload is saturated during the corresponding time period. After integrating these three types of information, the inspection robot performs matching analysis with the real-time navigation path structure to determine a list of replaceable nodes. For replaceable nodes, the inspection robot updates their original execution order and readjusts the path structure according to the recommended order in the priority switching sequence. In the updated navigation path, the original node execution order is replaced by the new target node. At the same time, a label field is inserted into the path segment structure to mark that the path segment has been replaced and adjusted, which facilitates the synchronous update of subsequent task scheduling records. For example, if the original order of a certain ward path is nodes A, B, C, and D, the inspection robot determines that the execution efficiency of node B is lower than that of node D, and node D is located in an adjacent ward with no task overlap. Therefore, the path is updated to ADC, and the "B→D" label is inserted at the B position of the path segment to generate a ward node change label.

[0109] Specifically, such as Figure 2 As shown, the task execution module includes:

[0110] The path numbering and organization submodule uses the change identifier of the patrol node to extract the marked path segment number and the corresponding node operation instruction, and organizes the continuous path segments into a path segment execution sequence according to the node index order. It evaluates the trigger index association relationship between the node and the path segment and generates a path segment instruction sequence set.

[0111] Based on the change markers for the patrol nodes, the patrol robot organizes the path data. It extracts the marked path segment numbers from the change markers and obtains the corresponding node operation instructions for each path segment. The robot then marks "Path segment 3 is replaced by node D" in the node replacement information and updates the operation instructions for path segment 3 to the actions and tasks corresponding to node D. The robot sorts the path segment execution information according to the node index order to ensure the sequential execution of patrol tasks along the physical path. During the sorting process, it identifies whether there are any skips between path segments. If a node operation instruction has a breakpoint between indices, it needs to backtrack the corresponding path segment and insert a placeholder to maintain the sequence integrity. The patrol robot evaluates the node trigger index association relationship of the organized path segment sequence. This evaluation is based on the triggering conditions of the operation instructions in the path segment and the time record of the task completion signal of the previous node. If the trigger time difference between path segments exceeds a set delay threshold, a delay risk field needs to be marked in the path sequence, generating a path segment instruction sequence set.

[0112] The equipment detection submodule calls the path segment instruction sequence set, reads and processes the status fields of the camera components, audio pickup components and signal transmission components configured in the path segment, determines whether the component status is marked as callable, removes component entries with unavailable status fields and records node indexes, and generates a callable index group for inspection components.

[0113] The inspection robot traverses each path segment in the sequence set, reading the status field information of the camera, microphone, and signal transmission components configured in that path segment. This status information comes from the front-end device control interface or the device management platform API. Status fields include status indicators such as "available," "initialization failed," "communication interrupted," and "restricted access." The inspection robot marks these status fields as callable or uncallable based on matching recognition rules. For components marked "uncallable," the robot removes them from the available list and records the corresponding node index in that path segment for subsequent skipping operations. Components with a "available" status field are included in the inspection component callable index group. Each item in this index group contains a node number and a list of corresponding callable component numbers. In practical applications, if a path segment node has a camera status of "communication interrupted," a microphone module of "available," and a signal transmission module of "available," the inspection robot removes the camera device under that node, generating the inspection component callable index group.

[0114] The task activation submodule performs device initialization, function interface activation, and signal channel connection operations on the camera, audio pickup, and signal transmission components respectively, based on the available component numbers under the nodes in the callable index group of the roving components. It also binds the general activation process of the devices under the nodes in the path segment instructions and generates a roving task start status registration item.

[0115] Based on the callable component number of each node, the camera, audio pickup, and signal transmission devices are processed separately. The initialization process for the camera device includes device number confirmation, lens calibration, and image stream transmission channel establishment. The audio pickup device needs to perform audio channel initialization, input gain standard setting, and echo suppression parameter adjustment. The signal transmission component needs to complete channel binding, transmission format setting, and communication path connection. Each type of device is activated by the inspection robot calling the hardware driver module to open the function interface. After the interface response returns to the normal state, it is marked as activated. The inspection robot binds the activation identifier of the activated devices under each node to the path segment instruction, which is written into the task control table as a prerequisite for the task execution phase. At the same time, it records the activation timestamp and response code for each device initialization process. Each record includes the node number, device type, initialization result, and whether it is connected, in order to track the task preparation status in the early stage. If a path segment node E successfully activates the camera and audio pickup module but fails to initialize the signal transmission module, the node will be marked as having a partially successful device activation status in the startup status registration item, providing a recovery mechanism interface for subsequent modules, and obtaining the patrol task startup status registration item.

[0116] Specifically, such as Figure 2 As shown, the command linkage module includes:

[0117] The instruction extraction submodule uses the mobile clinic task start status registration item to extract the voice instruction fields and operation focus fields that are active in the doctor's interactive interface, and filters the fields based on the activation status identifier, calculates the field interaction difference feature value, and generates the interface interaction control sequence.

[0118] The formula for calculating the interaction difference feature value of the fields is:

[0119] ;

[0120] in, Represents the interaction difference feature values ​​of the fields. Representing the Normalized weights of the item activation status identifier values Representing the The data item with a positive trigger status value in the voice command field. Representing the The real-time value of the corresponding data item in the focus field of the item operation. Representing the The cumulative number of times an item field is inactive. Represents the total number of fields extracted;

[0121] Meaning of parameters and derivation of formulas:

[0122] A total of 5 sets of interaction data were monitored for the voice command field, with a total of [number] corresponding fields. Each set of data was based on the speech recognition results and focus operation records automatically collected during the doctor's interactive monitoring cycle. Field items were extracted. The monitoring cycle was once every 10 seconds, and continuous sampling was conducted for 10 minutes. A total of 60 field data were collected, and the following sampled values ​​were obtained by averaging through the data window.

[0123] Group 1 data:

[0124] Activation status identifier normalized weight The normalization is based on the average triggering frequency of each group of fields in the same type of task. The normalized value is between 0 and 1, and the higher the frequency, the greater the weight.

[0125] Positive trigger status value in the voice command field This is obtained by counting the number of commands that were successfully identified and matched with positive actions;

[0126] The current value of the field to be focused on. The count is obtained by accumulating the identification fields of the currently focused operation area on the interface;

[0127] Cumulative number of times the field is inactive This is obtained by accumulating the number of times the same field appeared in the inactive state in past tasks;

[0128] Calculate intermediate terms:

[0129] ;

[0130] ;

[0131] ;

[0132] ;

[0133] ;

[0134] Group 2 data:

[0135] , , , ;

[0136] ;

[0137] ;

[0138] ;

[0139] ;

[0140] ;

[0141] Group 3 data:

[0142] , , , ;

[0143] ;

[0144] ;

[0145] ;

[0146] ;

[0147] ;

[0148] Group 4 data:

[0149] , , , ;

[0150] ;

[0151] ;

[0152] ;

[0153] ;

[0154] ;

[0155] Group 5 data:

[0156] , , , ;

[0157] ;

[0158] ;

[0159] ;

[0160] ;

[0161] ;

[0162] Substitute the calculated values ​​into the main formula:

[0163] ;

[0164] The results show that the field interaction difference feature value is 0.889. The value represents the strength of the difference between the voice command field and the operation focus field in the doctor's interaction interface under the current rounds task. The larger the value, the weaker the interaction correlation between the fields during task activation, the lower the proportion of the fields retained in the control sequence, the stricter the field selection, and the more concentrated the data used to generate the interface interaction control sequence will be on the content with a high degree of command-focus matching.

[0165] The target pairing submodule calls the operation focus field in the interface interaction control sequence and the target part field in the change identifier content of the round-trip node. Based on the field position code comparison, it obtains the overlapping area of ​​the focus and the part and generates a control target mapping structure.

[0166] The robot compares the operation focus field with the target location field in the change identifier of the round-trip node. Each operation focus field corresponds to the interactive element that the doctor actually focuses on in the interface. Its field position code is identified by the two-dimensional or three-dimensional coordinate block number recorded by the graphical interface component. The inspection robot compares the position code with the location code of the target location field recorded in the change identifier of the round-trip node at the field level to identify whether there is a spatial overlap or covering relationship between the two. For items with overlapping or overlapping relationships, the inspection robot marks them as valid matching items and records the mapping relationship between the interface position pointed to by the operation focus and the actual equipment part corresponding to it in the round-trip task. If the doctor sets the operation focus in the interface to the "upper abdominal area", the inspection robot identifies that the area corresponds to the "liver examination area" specified by the change node in the round-trip task. If the two are considered to be matched successfully, the matching item is registered in the structure group. The successfully matched item will become the object of subsequent instruction rearrangement and control linkage to ensure that the interactive behavior is consistent with the actual equipment task and to build a control target mapping structure.

[0167] The instruction control submodule rearranges the instruction fields according to the timestamp based on the control target mapping structure, marks the first field as the main control field, and marks the remaining fields as collaborative fields. It then binds the main control field to the corresponding target part in the control target mapping structure group and links the collaborative fields to generate a remote collaborative control instruction set.

[0168] The inspection robot rearranges the voice command fields corresponding to each group of control targets according to the task record timestamp, extracts the earliest triggered command as the current master control field, and arranges the rest in order of trigger time as collaborative fields. The rearrangement mechanism ensures that the doctor's core intention in task execution is prioritized to be responded to by the inspection robot. The master control field, as the main execution content in the command sequence, will be preferentially bound to the paired target parts in the control target mapping structure group to form a direct control path. The collaborative fields, as auxiliary commands, need to be executed in parallel after the master control field responds, and together with it affect the execution mode of the target node device. In a certain node, if the doctor's first voice command is "start image acquisition" and the second is "synchronize sound recording", the inspection robot will set "start image acquisition" as the master control field and bind it to the "camera device" target part marked on the node. The subsequent "synchronize sound recording" will be executed as a collaborative field in conjunction with the "sound pickup device". The binding relationship between the master control field and its matching collaborative fields and their corresponding target parts is analyzed to drive multiple hardware modules to respond collaboratively to the doctor's operation commands in the round of inspection tasks, ensuring that the task is executed synchronously according to the set process and generating a remote collaborative control command set.

[0169] The above are merely specific embodiments of the present invention, but the scope of protection of the present invention is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the technical scope disclosed in the present invention should be included within the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be determined by the scope of the claims.

Claims

1. A mobile clinic robot based on remote collaboration and consultation support, characterized in that: The robot includes: The instruction parsing module is used to extract the action direction field, trigger time period, target position coordinate field and target device part field based on the rounds instructions submitted by the remote consultation terminal, perform direction vector comparison between the action direction field and the navigation direction value in the robot state, and generate a list of parallel control instructions. The permission scheduling module is used to extract doctor numbers and consultation stage task execution records through the parallel control instruction list, mark doctors with more than two consecutive tasks as active execution status, mark doctors whose patient number range covers less than two wards as centralized task source doctors, classify task control behaviors, and generate a doctor rounds task control distribution table. The path switching module is used to call the doctor's rounds task control distribution table, compare the service status field of the navigation node with the status field of the target node in the set, and generate the change identifier of the rounds node. The task execution module is used to use the change identifier content of the patrol node, extract the node path segment instructions, check the availability status of the camera component, the audio pickup component and the signal transmission component, perform device activation operation on the callable components, and generate a patrol task start status registration item. The instruction linkage module is used to extract the voice instruction fields and operation focus fields that are active in the doctor's interactive interface by using the registration item of the roving task start status. The focus field and the target part field are matched to control targets. The voice instruction fields are arranged, and the first field is marked as the master field and the remaining fields are marked as collaborative fields, so as to generate a remote collaborative control instruction set.

2. The mobile clinic robot based on remote collaboration and consultation support according to claim 1, characterized in that, The parallel control instruction list includes a control target number sequence, direction consistency label, coordinate region pairing result, and conflict resolution identifier. The doctor's rounds task control distribution table includes execution frequency classification label, target region distribution category, doctor identity number mapping sequence, and behavior tendency sorting structure. The rounds node change identifier includes node update number, service status label grouping, navigation redirection identifier, and path priority execution mark. The rounds task start status registration item includes a device activation number list, component status mapping value, path segment execution identifier, and task queue update result.

3. The mobile clinic robot based on remote collaboration and consultation support according to claim 1, characterized in that, The instruction parsing module includes: The field extraction submodule is used to extract the action direction field, trigger time period, target location coordinate field and target device part field based on the rounds instructions submitted by the remote consultation terminal. The data types are divided and recorded as action vector items and positioning coordinate items respectively, and an instruction data structure set is generated. The direction comparison submodule is used to call the action vector items in the instruction data structure set and the navigation direction data to perform vector splitting processing. Based on the difference between the angle between the split direction items and the set direction reference angle, the direction items whose angle falls into the allowable deviation range are marked as navigation inputs, and a direction matching tag set is generated. The coordinate matching submodule is used to perform spatial distance calculation based on the positioning coordinate items in the instruction data structure set and the real-time path node coordinate set, calculate the Euclidean distance between coordinate point pairs and perform node attribution determination in combination with trajectory continuity features, mark the attribution node as the target area to be controlled, pair and bind the label matching items with the anchor point set, and generate a parallel path control list.

4. The mobile clinic robot based on remote collaboration and consultation support according to claim 3, characterized in that, The permission scheduling module includes: The task extraction submodule is used to extract the corresponding doctor number and consultation stage task execution record based on the parallel control instruction list, number and merge the task records of different time periods under the same doctor number according to the time order, and uniformly mark and filter out interrupted tasks to generate a doctor task list. The execution status identification submodule is used to call the length information of each task sequence in the doctor task list, calculate the frequency of consecutive tasks, determine whether the number of consecutive tasks exceeds two, and mark it as an active execution status if the condition is met. It also counts the number of ward numbers covered by the associated patient number in each doctor task, marks doctors with fewer than two ward numbers as centralized task source doctors, and generates a doctor status identification sequence. The behavior classification submodule is used to classify doctors in active execution status and doctors from centralized task sources into differentiated task control behavior classification groups according to the doctor status identifier sequence, and to call the task participation time period, task coverage ward number and task type to calculate the task execution time period deviation value, and to compare the horizontal task distribution ratio based on the behavior classification to generate a doctor rounds task control distribution table.

5. The mobile clinic robot based on remote collaboration and consultation support according to claim 4, characterized in that, The deviation value for the task execution period is calculated using the following formula: ； in, This represents the deviation value during the task execution period. This represents the number of doctors in an active execution state. Representing the The number of real-time task days for each doctor within a real-time task cycle. Representing the The earliest execution time number of the same type of task for each doctor. Representing the The latest execution time number of the same type of task for each doctor. Representing the The doctor's target task time center point.

6. The mobile clinic robot based on remote collaboration and consultation support according to claim 4, characterized in that, The path switching module includes: The data processing submodule is used to call the doctor's rounds task control distribution table, extract the patient waiting time record, symptom level code and dense visit record fields corresponding to the navigation node, and merge the three data according to the navigation node number, arrange them into a continuous arrangement structure in chronological order, and generate a node task load set. The status comparison submodule is used to call the service status field of the navigation node and the real-time status field of the target node in the set based on the target node number in the node task load set. The service status field of the navigation node and the real-time status field of the target node in the set are encoded and mapped into a unified sequence according to the service occupancy rate, active reception frequency and queue length respectively. The corresponding positions of the status sequences of the navigation node and the target node are compared to generate a node priority switching pair sequence. The node adjustment submodule is used to extract the location information, task arrangement time period and patient density frequency of the target node to be replaced in the sequence according to the node priority switching pair sequence. Based on the real-time navigation node path structure, the execution order of replaceable nodes is updated according to the priority switching position and the path segment replacement mark content is marked to generate the patrol node change mark content.

7. The mobile clinic robot based on remote collaboration and consultation support according to claim 6, characterized in that, The task execution module includes: The path numbering and sorting submodule is used to extract the marked path segment numbers and corresponding node operation instructions by using the change identifier content of the patrol node, and sort the continuous path segments into a path segment execution sequence according to the node index order, evaluate the trigger index association relationship between the node and the path segment, and generate a path segment instruction sequence set. The equipment detection submodule is used to call the path segment instruction sequence set, read the status fields of the camera components, audio pickup components and signal transmission components configured in the path segment, determine whether the component status is marked as callable, remove component entries with unavailable status fields and record node indexes, and generate a callable index group for inspection components. The task activation submodule is used to perform device initialization, function interface opening and signal channel connection operations on the camera, audio pickup and signal transmission components respectively, according to the available component number under the node in the callable index group of the roving component, bind the general activation process of the device under the node in the path segment instruction, and generate a roving task start status registration item.

8. The mobile clinic robot based on remote collaboration and consultation support according to claim 1, characterized in that, The remote collaborative control instruction set includes the master doctor number, voice control link, synchronous operation mapping structure, and interactive focus association items.

9. The mobile clinic robot based on remote collaboration and consultation support according to claim 8, characterized in that, The instruction linkage module includes: The instruction extraction submodule is used to extract the voice instruction fields and operation focus fields that are active in the doctor's interactive interface by using the registration item of the initiation status of the round of visits task, and to filter the fields based on the activation status identifier, calculate the field interaction difference feature value, and generate the interface interaction control sequence. The target pairing submodule is used to call the operation focus field in the interface interaction control sequence and the target part field in the patrol node change identifier content, obtain the overlapping area of ​​the focus and part based on the field position code comparison, and generate a control target mapping structure. The instruction control submodule is used to rearrange the instruction fields according to the timestamp based on the control target mapping structure, mark the first field as the main control field, and mark the remaining fields as collaborative fields. The main control field is bound to the corresponding target part in the control target mapping structure group and the collaborative fields are linked to generate a remote collaborative control instruction set.

10. The mobile clinic robot based on remote collaboration and consultation support according to claim 9, characterized in that, The formula for calculating the interaction difference feature value of the fields is: ； in, Represents the interaction difference feature values ​​of the fields. Representing the Normalized weights of the item activation status identifier values Representing the The data item with a positive trigger status value in the voice command field. Representing the The real-time value of the corresponding data item in the focus field of the item operation. Representing the The cumulative number of times an item field is inactive. This represents the total number of fields extracted.

Images