Nursing task dynamic scheduling and path optimization collaborative management method and system

Abstract

The invention discloses a nursing task dynamic scheduling and path optimization collaborative management method and system, particularly relates to the technical field of nursing task scheduling, and is used for solving the problems of unnecessary movement distance accumulation and emergency response delay caused by task transmission and path planning disjunction in the existing cross-shift handover process. Generating a to-be-handed-over set by obtaining uncompleted task points; evaluating the complexity of the path movement decision tree of each task point; mining a cross-shift task chain and calculating a time efficiency entropy; generating a space-time value based on the complexity and the aging entropy to select a dynamic handover anchor point; tracking an anchor point value attenuation rate in real time and triggering anchor point updating; generating an initial path and an emergency re-planning path based on the dynamic anchor point position; a redundant path moving from a fixed starting point to a first task point is eliminated through a space-time value driven anchor point dynamic decision, and rapid path response of an emergency task is realized by using the real-time position of an anchor point.

Description

Technical Field

[0001] This invention relates to the field of nursing task scheduling technology, and more specifically, to a collaborative management method and system for dynamic scheduling and path optimization of nursing tasks. Background Technology

[0002] In nursing task scheduling and path optimization, existing systems typically use a discrete management model to handle shift handover scenarios. When shifts change, the system reassigns unfinished tasks to the executors of the subsequent shift and independently generates task sequences and movement paths for each executor. In this mode, scheduling decisions and path planning are limited to closed-loop optimization within a single shift time window, and task transfer between different shifts is only achieved through task list handover.

[0003] However, the above methods can lead to significant systemic losses for executors during the handover process. Because the handover process does not establish a spatiotemporal correlation between the geographical attributes of the task and the executor's movement path, the incoming executor must move from a fixed starting point (such as the duty room) to the first task point, resulting in a large accumulation of unnecessary travel distances. At the same time, the emergency task insertion mechanism is limited by the discrete scheduling framework and cannot utilize the geographical continuity of the handover phase to achieve rapid response, making it difficult to balance task transfer efficiency and path optimization goals. Summary of the Invention

[0004] In order to overcome the above-mentioned defects of the prior art, the present invention provides a method and system for collaborative management of dynamic scheduling and path optimization of nursing tasks to solve the problems mentioned in the background art.

[0005] To achieve the above objectives, the present invention provides the following technical solution:

[0006] A collaborative management method for dynamic scheduling and path optimization of nursing tasks includes:

[0007] S1. Obtain the geographic coordinates and task attributes of all unfinished task points in the current shift, and generate a set of tasks to be handed over.

[0008] S2. Generate the optimal path movement decision tree from the successor to each task point in the set of tasks to be handed over and evaluate the complexity of the movement decision tree;

[0009] S3. Identify cross-shift task chains in the set of tasks to be handed over by mining association rules, calculate the time interval distribution entropy value of adjacent tasks in the cross-shift task chain, and obtain the time efficiency entropy of the task chain.

[0010] S4. Based on the complexity of the moving decision tree of each task point and the time entropy of the task chain, generate the handover spatiotemporal value, and select the task point with the highest value as the dynamic handover anchor point.

[0011] S5. Real-time tracking of the spatiotemporal value decay rate of dynamic handover anchor points. When the spatiotemporal value decay rate is greater than the decay rate threshold, the second highest value task point is updated as the dynamic handover anchor point.

[0012] S6. Generate the initial optimized path for the successor based on the current dynamic handover anchor point, and replan the path based on the real-time position of the dynamic handover anchor point when a new emergency task is added.

[0013] Furthermore, obtain the geographic coordinates and task attributes of all unfinished task points for the current shift, and generate a set of tasks to be handed over, including:

[0014] Obtain the geographic coordinates and task attributes of all unfinished tasks in the current shift, including task type, task urgency, and task dependency; identify task points with cross-shift continuous execution requirements based on task dependency and mark them as mandatory handover task points;

[0015] Based on the urgency of the tasks, select non-mandatory handover task points whose urgency level is higher than a preset urgency threshold; merge mandatory handover task points with the selected non-mandatory handover task points to generate a set of tasks to be handed over.

[0016] Furthermore, the optimal path movement decision tree from the successor to each task point in the set of tasks to be handed over is generated, and the complexity of the movement decision tree is evaluated, including:

[0017] A weighted path network was constructed based on the hospital's indoor navigation map, including weights for passage length, access permit status, and elevator waiting time.

[0018] The weighted path network is input into the Monte Carlo tree search algorithm to simulate the process of generating the optimal path movement decision tree from the current position of the successor to each task point in the set of tasks to be handed over.

[0019] During the generation of the optimal path movement decision tree, environmental constraints such as temporary closed channels and sudden changes in pedestrian density are dynamically injected.

[0020] Extract the maximum depth and average number of branches of all nodes in the optimal path movement decision tree;

[0021] The complexity of the moving decision tree is calculated by multiplying the maximum depth by the average number of branches.

[0022] Furthermore, by using association rule mining to identify cross-shift task chains in the set of tasks to be handed over, the time interval distribution entropy values ​​of adjacent tasks in the cross-shift task chain are calculated to obtain the task chain time efficiency entropy, including:

[0023] Analyze the task dependency attributes of all task points in the set of tasks to be handed over, identify the combination of task points that have the requirement of continuous execution across shifts, and define it as a cross-shift task chain.

[0024] Obtain historical execution timestamp data for each task point in the cross-shift task chain from the hospital information system;

[0025] Calculate the historical time intervals between adjacent task points in a cross-shift task chain and generate a set of time interval distribution probability.

[0026] Based on the probability set of time interval distributions, the entropy value of the time interval distribution is calculated using the Shannon entropy formula.

[0027] Multiplying the time interval distribution entropy value by the environmental decay factor yields the task chain time-dependent entropy.

[0028] Furthermore, the environmental degradation factor is derived from a weighted fusion of weather indices and seasonal disease outbreak coefficients.

[0029] Furthermore, based on the complexity of the decision tree for each task point and the time entropy of the task chain, a handover spatiotemporal value is generated, and the task point with the highest value is selected as the dynamic handover anchor point, including:

[0030] Normalize the movement decision tree complexity and task chain time entropy of each task point to the same numerical dimension.

[0031] Construct a graph structure with task points as nodes, where the node feature vectors contain the normalized moving decision tree complexity and task chain time entropy;

[0032] The feature vectors of adjacent task points are aggregated through the message passing mechanism of a graph neural network.

[0033] For each task point, a graph readout function is applied to generate the handover spatiotemporal value, where the graph readout function includes a multilayer perceptron and a sigmoid activation function;

[0034] Compare the handover spatiotemporal value values ​​of all task points, and select the task point with the highest handover spatiotemporal value value as the dynamic handover anchor point.

[0035] Furthermore, the spatiotemporal value decay rate of the dynamic handover anchor point is tracked in real time. When the spatiotemporal value decay rate exceeds the decay rate threshold, the second-highest value task point is updated as the dynamic handover anchor point, including:

[0036] Establish a time series of spatiotemporal values ​​of dynamic handover anchor points and record the spatiotemporal value values ​​of handover at a fixed sampling frequency;

[0037] A linear regression model was used to fit the decay slope of the spatiotemporal value time series of the handover, which was taken as the spatiotemporal value decay rate.

[0038] The spatiotemporal value decay rate is compared with a preset decay rate threshold in real time.

[0039] When the spatiotemporal value decay rate is detected to be greater than the decay rate threshold, the current second highest value task point is obtained from the spatiotemporal value sorting table.

[0040] Update the dynamic handover anchor point to the second highest value task point, and mark the original anchor point's status as decayed in the handover spatiotemporal value ranking table.

[0041] Furthermore, an initial optimized path for the successor is generated based on the current dynamic handover anchor point, and the path is replanned based on the real-time position of the dynamic handover anchor point when a new emergency task is added, including:

[0042] The real-time geographic coordinates of the current dynamic handover anchor point are obtained as the starting point of the path. Based on the moving decision tree generation mechanism, an initial optimized path is constructed from the starting point of the path to the remaining task points in the set of tasks to be handed over.

[0043] After the initial optimized path is generated, path conflict detection is performed to identify path segments that overlap with fixed obstacles and temporary closed passages in the hospital.

[0044] When a new emergency task is detected, the real-time location of the dynamic handover anchor point is obtained as the starting point for replanning.

[0045] A multi-objective optimization algorithm is used to generate a replanning path from the replanning starting point to the newly added urgent task point and the incomplete task points in the set of tasks to be handed over.

[0046] On the other hand, the present invention provides a collaborative management system for dynamic scheduling and path optimization of nursing tasks, comprising:

[0047] The task acquisition module is used to obtain the geographical coordinates and task attributes of all unfinished task points in the current shift and generate a set of tasks to be handed over.

[0048] The path decision module is used to generate the optimal path movement decision tree from the successor to each task point in the set of tasks to be handed over and to evaluate the complexity of the movement decision tree.

[0049] The entropy calculation module is used to identify cross-shift task chains in the set of tasks to be handed over through association rule mining, calculate the time interval distribution entropy of adjacent tasks in the cross-shift task chain, and obtain the time-sensitivity entropy of the task chain.

[0050] The value generation module is used to generate the handover spatiotemporal value based on the moving decision tree complexity and task chain time entropy of each task point, and selects the task point with the highest value as the dynamic handover anchor point.

[0051] The anchor point update module is used to track the spatiotemporal value decay rate of dynamic handover anchor points in real time. When the spatiotemporal value decay rate is greater than the decay rate threshold, the second highest value task point is updated as the dynamic handover anchor point.

[0052] The path planning module is used to generate an initial optimized path for the successor based on the current dynamic handover anchor point, and to replan the path based on the real-time location of the dynamic handover anchor point when a new emergency task is added.

[0053] Compared with the prior art, the present invention has the following beneficial effects:

[0054] 1. By using dynamic handover anchor points, deep collaboration between cross-shift task transfer and executor path planning is achieved in the spatiotemporal dimension. Specifically, anchor points are dynamically generated and updated based on the path complexity of task points and the time entropy of cross-shift task chains. This allows the incoming executor to start executing tasks and planning paths directly from the most reasonable dynamic spatial node (i.e., anchor point), rather than the traditional fixed starting point (such as the duty room). This effectively eliminates unnecessary path accumulation caused by shift changes and movement from a fixed starting point to the first task point, significantly improving the geographical efficiency of task transfer and the executor's movement efficiency.

[0055] 2. By combining dynamic handover anchor points with real-time tracking and updating strategies, the system provides advantages for emergency task response. When a new emergency task is added, the system can replan the path based on the real-time location of the executor's current dynamic anchor point. This real-time response mechanism based on the executor's actual movement status and optimal spatial nodes makes full use of the geographical continuity of the handover phase, enabling emergency tasks to be inserted into the ongoing task sequence and path more quickly and reasonably. This significantly improves the system's dynamic adaptability and response agility to sudden demands while ensuring overall handover efficiency. Attached Figure Description

[0056] Figure 1 This is a flowchart of a collaborative management method for dynamic scheduling and path optimization of nursing tasks according to the present invention;

[0057] Figure 2 This is a schematic diagram of the structure of a collaborative management system for dynamic scheduling and path optimization of nursing tasks according to the present invention. Detailed Implementation

[0058] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0059] Example 1: Figure 1 This invention provides a collaborative management method for dynamic scheduling and path optimization of nursing tasks, comprising:

[0060] S1. Obtain the geographic coordinates and task attributes of all unfinished task points in the current shift, and generate a set of tasks to be handed over.

[0061] S2. Generate the optimal path movement decision tree from the successor to each task point in the set of tasks to be handed over and evaluate the complexity of the movement decision tree;

[0062] S3. Identify cross-shift task chains in the set of tasks to be handed over by mining association rules, calculate the time interval distribution entropy value of adjacent tasks in the cross-shift task chain, and obtain the time efficiency entropy of the task chain.

[0063] S4. Based on the complexity of the moving decision tree of each task point and the time entropy of the task chain, generate the handover spatiotemporal value, and select the task point with the highest value as the dynamic handover anchor point.

[0064] S5. Real-time tracking of the spatiotemporal value decay rate of dynamic handover anchor points. When the spatiotemporal value decay rate is greater than the decay rate threshold, the second highest value task point is updated as the dynamic handover anchor point.

[0065] S6. Generate the initial optimized path for the successor based on the current dynamic handover anchor point, and replan the path based on the real-time position of the dynamic handover anchor point when a new emergency task is added.

[0066] S1. Obtain the geographic coordinates and task attributes of all unfinished task points in the current shift, and generate a set of tasks to be handed over. The specific implementation is as follows:

[0067] When obtaining the geographic coordinates and task attributes of all unfinished tasks in the current shift, the geographic coordinates are collected in real time using Bluetooth beacon triangulation technology of the hospital's indoor positioning system, with positioning accuracy controlled within 1 meter. The task type in the task attributes is classified into standard categories such as intravenous infusion, vital sign monitoring, and medication dispensing according to the hospital's nursing operation specifications. The urgency of the task is quantified according to the grading standards formulated by the hospital's nursing department. The quantification method is to map the text priority defined by the nursing department to a numerical range of 0 to 1. For example, "critical" is mapped to 1.0, "high" to 0.8, "medium" to 0.5, and "low" to 0.3. The task dependency relationship is generated by parsing the medical order execution sequence of the electronic medical record system. Specifically, each task point records its set of preceding task identifiers.

[0068] When identifying task points with cross-shift continuous execution requirements based on task dependencies, cross-shift continuous execution requirements are defined as situations where the completion status of the current task directly affects the start of the next shift's task. For example, in a patient continuous blood glucose monitoring task, if the monitoring data collection is not completed in the current shift, the next shift will not be able to generate an analysis report. The identification logic is to traverse the set of successor task identifiers in the task dependency relationship. If the shift to which the successor task belongs is different from the current shift, it is marked as a mandatory handover task point. The marking result is stored in the handover flag field of the task point attribute table.

[0069] When filtering non-mandatory handover task points based on task urgency, the preset urgency threshold is set to 0.7. This threshold is derived from the analysis of historical nursing records in the hospital. Specifically, through statistical analysis of nursing task execution records over the past 12 months, it was found that when the task urgency value is below 0.7, the incidence of adverse events due to delayed execution is less than 5%. The filtering process excludes records marked as mandatory handover task points, sorts the remaining task points in descending order of urgency value, and selects task points with urgency values ​​greater than 0.7 to generate a subset of high-urgency task points.

[0070] When generating a set of tasks to be handed over by combining the mandatory handover task points and the high-urgency task points, a database join query is used to merge the unique identifiers of the two types of task points. The merged set contains task point identifiers, geographic coordinates, and complete task attribute fields. After the set is generated, data deduplication is performed by comparing the hash values ​​of the task point identifiers to ensure that the same task point does not appear repeatedly.

[0071] Verify the topological consistency between geographic coordinates and the hospital's indoor positioning map: Load the navigation node data of the hospital's indoor positioning map, calculate the Euclidean distance between the geographic coordinates of each task point in the task handover set and the nearest navigation node. The distance calculation formula is the square root of the sum of the squares of the coordinate differences between two points in a Cartesian coordinate system. When the Euclidean distance exceeds 2 meters, it is determined to be a topological inconsistency, triggering the coordinate correction process: Call the indoor positioning system application interface to re-collect coordinate data. If the distance still exceeds 2 meters after re-collection, remove the task point from the task handover set and generate an exception log.

[0072] In the above process, the quantitative value of task urgency comes from the nursing task priority label in the hospital information system, which is automatically labeled by the medical order system according to the severity of the patient's condition; the preset urgency threshold is set based on regression analysis of the hospital's historical nursing data; the 2-meter threshold setting in the topology consistency verification is based on the technical characteristics of the indoor positioning system, for example, the standard deviation of the positioning error of a certain model of Bluetooth positioning system is 0.5 meters; the judgment rule for cross-shift continuous execution requirements is formulated according to the continuity requirements of medical workflow, specifically, when the task status field in the electronic medical record system is "in progress" and the associated medical order contains the "continuous execution" attribute, a cross-shift continuous execution identifier is automatically added.

[0073] The task dependency attribute is generated by parsing the medical order execution sequence in the electronic medical record system. Specifically, each task point records its set of direct preceding task identifiers. The set of successor task identifiers is generated through reverse lookup: traversing the preceding task identifier sets of all task points, if the preceding task set of task point A contains the identifier of task point B, then the identifier of task point A is added to the successor task set of task point B. The task point identifier adopts the standard task coding format in the hospital information system, consisting of 6 digits. The first 2 digits represent the department code, the middle 2 digits represent the task type code, and the last 2 digits represent the sequence number. The hash value is calculated by encrypting the complete identifier string using the MD5 algorithm to generate a 32-bit hexadecimal hash code. During deduplication verification, the values ​​of these hash codes are compared to see if they are the same.

[0074] S2. Generate the optimal path movement decision tree from the successor to each task point in the set of tasks to be handed over, and evaluate the complexity of the movement decision tree. The specific implementation is as follows:

[0075] When constructing a weighted path network based on the hospital's indoor navigation map, which includes weights for passage length, access permission status, and elevator waiting time, the hospital's indoor navigation map is derived from navigation grid data exported from the Building Information Model (BIM). The navigation grid data includes node coordinates and the connections between nodes. Passage length is obtained by calculating the Euclidean distance between adjacent node coordinates, specifically by taking the square root of the sum of the squares of the differences in the X / Y coordinates of two nodes, with the result in meters. The weighting method for access permission status is as follows: The access permission status is accessed in real-time via the hospital's access control system's channel status interface. When a channel is disabled, it is assigned a very high weight value (specifically, the maximum floating-point number allowed by the system); when enabled, it is assigned a basic weight value of 1.0. The process for setting the elevator waiting time weight includes: extracting historical waiting time records from the hospital's elevator operation database, grouping them by hourly time periods, and calculating the 90th percentile value of the waiting time for each time period. For example, if the 90th percentile value for the 8:00-9:00 time period is 85 seconds, then the elevator edge weight for that time period is set to 85.

[0076] The weighted path network construction method is as follows: Navigation nodes of the hospital's indoor navigation map are used as vertices, and the connecting channels between nodes are used as edges. Each edge is assigned a comprehensive weight value, calculated as: channel length multiplied by access permission weight, and then multiplied by elevator waiting weight. Before calculating the comprehensive weight value, the channel length, access permission weight, and elevator waiting weight need to be normalized to ensure dimensional consistency. The maximum value of the access permission weight is specifically represented using the IEEE 754 standard single-precision floating-point maximum value of 3.4028235e38. The elevator waiting time weight is set to an actual value in seconds; for example, 85 seconds has a weight of 85. Because the weight calculation uses actual time values ​​rather than proportional coefficients, this setting ensures that the path cost is proportional to the actual time consumption.

[0077] When inputting the weighted path network into the Monte Carlo Tree Search algorithm, the algorithm parameter setting process is as follows: The base number of iterations is determined based on the hospital's building area and the number of task points. For every additional 100 square meters of area or 5 task points, 50 iterations are added. The final iteration count is rounded up to the nearest hundred. The simulation process uses the real-time location coordinates of the successor as the root node, specifically by matching the location coordinates to the nearest navigation node. Decision tree generation includes four stages: 1. Node selection stage: The Unified Confidence Limit (UCT) formula is used to calculate the node score, with the exploration weight coefficient set to the square root of 2; 2. Expansion stage: Unvisited child nodes are added to the current node; 3. Simulation stage: Random walks are performed from the new node until the target node is reached; 4. Backtracking stage: The number of visits to path nodes and the cumulative value are updated.

[0078] The Monte Carlo Tree Search algorithm is specifically configured as follows: the algorithm outputs the optimal path sequence from the root node to the target node; the confidence upper limit formula used in the node selection phase is: UCT value = average node value + √2 × √(parent node visit count) / current node visit count; the cumulative value update method is: add the reciprocal of the total cost of the simulated path to the node value; the reason for rounding the number of iterations to the hundreds is to ensure that the algorithm can obtain stable solutions under different map sizes, and to avoid insufficient search due to too few iterations.

[0079] Average node value: Defined as the cumulative value of a node divided by the number of times the node is visited. The cumulative value of a node is obtained by summing the reward values ​​of all simulated paths that pass through the node. Specifically, during the algorithm backtracking phase, the reward value of the simulated path is added to the cumulative value of each node in the path, and then the average value of each node is recalculated.

[0080] The total cost of this simulated path refers to the sum of the weights of all edges traversed during the simulation phase when randomly walking from the current node to the target node. The weights include the comprehensive costs such as the path length, access permission status, and elevator waiting time. Specifically, it is obtained by traversing each edge in the simulated path and accumulating its current weight value.

[0081] Node value: Represents the expected efficiency of the path starting from this node, with an initial value of zero. During the algorithm's operation, the node value is updated through multiple simulated reward values. The specific update method is: Node value = (historical cumulative value × historical visit count + current reward value) / (historical visit count + 1).

[0082] When dynamically injecting environmental constraints during the optimal path movement decision tree generation process, the detection mechanism for temporarily closed channels is as follows: every 10 seconds, the channel status interface of the hospital facility management system is polled. When the returned status code is "disabled", the path network update function is called to set the weight of the channel to the maximum value. The sudden flow density handling process includes: collecting real-time video streams through visual sensors deployed on the corridor ceiling, and using a background subtraction algorithm to calculate the number of people per unit area; the density grading threshold is set based on the hospital's historical passage efficiency data. When the number of people per unit area exceeds 0.3 people / square meter, the passage efficiency declines to an inflection point, so 0.3 is used as the lower limit of the medium density threshold; the delay coefficient is set by multiplying the basic passage time by (1 + density level coefficient), where the medium density level coefficient is 0.2.

[0083] Before setting the delay coefficient, the density level coefficient and the basic passage time need to be normalized to ensure that the dimensions are consistent.

[0084] The path network update function is implemented as follows: traverse all edges in the weighted path network, and when the channel identifier corresponding to an edge matches a disabled channel, update the weight value of that edge to 3.4028235e38. The specific steps of the background subtraction algorithm to calculate the number of people per unit area are as follows: first, establish a background model of the corridor area; then, extract the foreground from real-time video frames, count the number of pixels in connected regions, and convert the number of pixels into the actual number of people based on camera calibration parameters. The medium density level coefficient is set to 0.2 based on hospital corridor passage experiment data: when the density exceeds 0.3 people / square meter, the passage speed decreases by 20%, so the delay coefficient increases accordingly by 20%. The implementation method of dynamically injecting environmental constraints is as follows: during each simulation of the Monte Carlo tree search, the current channel status and pedestrian density are queried in real time, and the passage cost of the edges is dynamically adjusted.

[0085] When extracting the maximum depth of all nodes in the optimal path movement decision tree, a depth-first traversal algorithm is used to visit all leaf nodes, and the length of the longest edge from the root node to a leaf node is recorded. The average number of branches is calculated as follows: traverse all non-leaf nodes and record the number of their child nodes, sum the number of child nodes of all non-leaf nodes, and then divide by the total number of non-leaf nodes.

[0086] The complexity calculation process for a moving decision tree follows the formula: Moving decision tree complexity = maximum depth × average number of branches. The result is a dimensionless value. For example, when the maximum depth obtained from traversal is 15 and the average number of branches is 3.2, the multiplication operation is 15 × 3.2 = 48.

[0087] In a moving decision tree, depth is defined as the number of edges traversed from the root node to a leaf node. In a large hospital path network, the path from the nurses' station to a remote ward may pass through multiple intermediate nodes, such as corridor nodes, elevator nodes, and partition corridor nodes. Therefore, the maximum depth may reach 15. When calculating the average number of branches, the arithmetic mean is calculated by counting the number of direct child nodes of all non-leaf nodes.

[0088] To verify the logical consistency between the complexity of the moving decision tree and the weights of the weighted path network, a preset threshold of 1.0 is set based on the following: analysis of historical path planning data revealed that when complexity / maximum single-sided weight > 1.0, 95% of cases exhibit path anomalies. The verification process specifically includes: 1. Traversing all weight values ​​of the weighted path network and taking the maximum value MAX_W; 2. Calculating the ratio R = complexity / MAX_W; 3. If R ≤ 1.0, it is considered consistent; 4. If R > 1.0, the current decision tree is deleted and regenerated.

[0089] In the above process, the 90th percentile value of elevator waiting time is calculated as follows: all waiting time samples for that period are sorted in ascending order, and the 90th percentile value is taken. The basis for setting the pedestrian density level coefficient of 0.2 is: through hospital corridor passage experiments, it was found that the average passage speed decreases by 20% under medium density. When processing visual sensor data, if three consecutive acquisitions fail, the system switches to backup infrared sensor data. The dynamic setting method for elevator waiting time weight is as follows: the hospital elevator operation log analysis system collects the stopping floor data of each elevator car every minute and calculates the average waiting time for each floor; the iteration number of the Monte Carlo tree search algorithm is determined through pre-experimentation: the iteration number is gradually increased in the test map until the optimal path no longer changes after 100 consecutive iterations; the pedestrian density classification threshold is set based on the analysis of historical monitoring data of the hospital, and the visual sensor outputs the density analysis result every 5 seconds; the threshold of 1.0 for logical consistency verification is determined through path planning anomaly detection experiments. When the product of the decision tree depth and the branch exceeds the maximum single-sided weight, it indicates that there is an abnormal path expansion. During the path network construction process, if the channel length data is missing, the nearest neighbor navigation node distance interpolation is used; during the decision tree generation process, if all weights are extremely large, the path is determined to be unreachable and an alarm is triggered.

[0090] S3. Identify cross-shift task chains in the set of tasks to be handed over through association rule mining, calculate the time interval distribution entropy of adjacent tasks in the cross-shift task chain, and obtain the task chain time efficiency entropy. The specific implementation is as follows:

[0091] When parsing the task dependency attributes of all task points in the task handover set, these attributes originate from the medical order execution sequence data in the hospital's electronic medical record system. Specifically, this involves parsing the identifier field in the medical order records that indicates the order of tasks. The process of identifying combinations of task points with cross-shift continuous execution requirements is as follows: traverse the set of successor task identifiers for each task point in the task handover set. If the shift field of a successor task differs from the current shift, then that task and its successor task form a task chain, defined as a cross-shift task chain. For example, in postoperative rehabilitation training tasks, an unfinished pain assessment task in the current shift and the rehabilitation training task in the next shift form a cross-shift task chain. This identification process is achieved through database association queries.

[0092] When retrieving historical execution timestamp data for each task point in a cross-shift task chain from the hospital information system, the application programming interface of the hospital nursing execution record database is called. The query condition is the execution records of the same task chain combination within the past 90 days. The timestamp data is accurate to the minute, and the data filtering rule is to exclude records with abnormal execution status. For example, records with execution interruption time exceeding 120 minutes are marked as invalid data. The 90-day query time range is set based on the hospital nursing quality assessment cycle, which covers nursing task execution data for three full months.

[0093] When calculating the historical time interval between adjacent task points in a cross-shift task chain, the time interval is defined as the difference between the completion timestamp of the preceding task and the start timestamp of the subsequent task, in minutes. The specific method for generating the probability set of time interval distribution is as follows: group the historical time interval data according to a fixed time interval, and set the group width to 15 minutes according to the maximum time tolerance stipulated by the medical shift handover system; count the frequency of each interval, and the frequency calculation formula is the number of samples in the group divided by the total number of samples; for example, if the time interval occurs 30 times in the interval between 0 and 15 minutes in 100 valid historical records, then the probability value of this interval is 0.3; if the number of samples in a certain interval is zero, assign a minimum probability value of 0.001 to avoid calculation anomalies.

[0094] When calculating the entropy value of the time interval distribution using the Shannon entropy formula based on the probability set of the time interval distribution, the following operations are performed:

[0095] For each interval in the probability set, calculate the base-2 logarithm of the probability value, multiply each probability value by its corresponding logarithm, sum all the products, and multiply the sum by -1. The final time interval distribution entropy value is in bits. When all time intervals are concentrated in a single interval, the entropy value is zero bits, and the entropy value reaches its maximum value when uniformly distributed.

[0096] When multiplying the time interval distribution entropy value by the environmental decay factor to obtain the task chain time-effect entropy, the environmental decay factor is generated as follows: the weather index is derived from real-time data from the meteorological service application interface, with a value between 0 and 1, and the specific mapping rule is 1.0 for sunny days, 0.7 for cloudy days, and 0.3 for heavy rain; the seasonal disease outbreak coefficient is derived from the daily reports of the hospital infection monitoring system, with a value between 0 and 1, 1.0 for no outbreak, 0.8 for local outbreaks, and 0.5 for hospital-wide outbreaks; the weighted fusion formula is: the environmental decay factor equals 0.7 times the weather index plus 0.3 times the seasonal disease outbreak coefficient, where the weight coefficients 0.7 and 0.3 are derived from the regression analysis of the hospital's task delay data over the past three years; the final task chain time-effect entropy equals the time interval distribution entropy value multiplied by the environmental decay factor, with the result in bits.

[0097] Verify the logical consistency between the task chain time-dependent entropy calculation result and the task dependency attribute. The logical consistency rule is set as follows: when the task dependency attribute is marked as a strong dependency, the task chain time-dependent entropy should be less than the preset baseline entropy value. The verification process includes: first, extracting the dependency strength flag from the task dependency attribute, with a strong dependency flag value of 1 and a weak dependency flag value of 0; second, setting the baseline entropy value to 0.8 bits, which is determined by taking the median of the entropy value distribution of 1000 non-dependent task chains; finally, when a strong dependency flag is detected and the task chain time-dependent entropy is greater than 0.8 bits, a logical inconsistency is determined and a data review process is triggered. The review content includes the integrity of the timestamp data and the accuracy of the environmental decay factor value.

[0098] In the above process, the 15-minute time interval grouping width is set based on the maximum task interval tolerance stipulated by the hospital's shift handover system; the minimum probability value of 0.001 is set by taking the inverse upper limit of the total sample size; the weighting coefficients of the environmental degradation factor, 0.7 and 0.3, are determined by analyzing historical task delay data; and the baseline entropy value of 0.8 bits for logical consistency verification is obtained by statistically analyzing the entropy distribution of the independent task chain dataset. The environmental degradation factor is updated hourly, triggering a real-time update when the monitored weather index or disease outbreak coefficient changes by more than 20%. When querying historical timestamps, if the data volume is less than 50 records, the query range is expanded to the past 180 days to ensure statistical significance.

[0099] S4. Based on the complexity of the decision tree for each task point and the time entropy of the task chain, generate the spatiotemporal value of handover, and select the task point with the highest value as the dynamic handover anchor point. The specific implementation is as follows:

[0100] When normalizing the movement decision tree complexity and task chain time entropy of each task point in the task handover set to the same numerical dimension, a max-min normalization method is used. The specific implementation process is as follows: First, obtain the movement decision tree complexity values ​​of all task points in the task handover set to form a first numerical set, and identify the maximum and minimum values ​​in this set; at the same time, obtain the task chain time entropy values ​​of all task points to form a second numerical set, and similarly identify its maximum and minimum values; then, calculate the movement decision tree complexity of each task point: subtract the minimum value of the first numerical set from the complexity value, and then divide by the first numerical set. The difference between the maximum and minimum values ​​of the set; the same logic is applied to the task chain time entropy of each task point: subtract the minimum value of the second set of values ​​from the entropy value and divide by the difference between the maximum and minimum values ​​of the second set of values; finally, all normalized results are mapped to a closed interval between 0 and 1 to form a dimensionless standardized value; for example, if the original movement decision tree complexity of a task point is 48, the minimum complexity of the entire set is 20, and the maximum complexity is 80, then the normalized result is (48-20) / (80-20)=0.467; if a special case occurs where the maximum value is equal to the minimum value, then all normalized values ​​are set to 0.5.

[0101] When constructing a graph structure with task points as nodes, a node is defined as each task point entity in the set of tasks to be handed over. The rule for generating connecting edges between nodes is: calculate the planar Euclidean distance between the geographic coordinates of any two task points, and establish a connecting edge when the distance is less than 50 meters. The node feature vector is constructed as a two-dimensional vector structure. The first dimension stores the normalized moving decision tree complexity value, and the second dimension stores the normalized task chain time entropy value. The graph structure data is stored through an adjacency matrix. The matrix row and column indices correspond to the task point numbers. A matrix element value of 1 indicates the existence of a connecting edge, and a value of 0 indicates no connection relationship.

[0102] When aggregating feature vectors of adjacent task points using the message passing mechanism of a graph neural network, the following operations are performed: In the message generation phase, for each node, its set of neighboring nodes is traversed, and the feature vector of the neighboring node is multiplied by the weight value of the connecting edge to generate a message vector, where the edge weight value is the reciprocal of the distance between the two nodes; In the message aggregation phase, the message vectors of all neighbors of the current node are processed using an arithmetic mean function; In the node update phase, the aggregated result vector is concatenated with the original feature vector of the current node, and the result is input into a fully connected layer containing 32 neurons to generate the updated feature vector; The above message passing process is executed iteratively twice, and after each iteration, a modified linear unit activation function is applied to the feature vector for nonlinear transformation.

[0103] When applying the graph readout function to generate the handover spatiotemporal value for each task point, the graph readout function is implemented as a three-layer fully connected neural network architecture: the input layer receives the node feature vector updated through message passing and sets 16 neurons; the hidden layer sets 8 neurons and applies a modified linear unit activation function for feature transformation; the output layer sets 1 neuron and applies a sigmoid activation function to compress the output value to the range of 0 to 1; the final output value is the handover spatiotemporal value, which is positively correlated with the comprehensive value of the task point as a handover anchor point; for example, when the input feature vector is [0.6, 0.8], the neural network may calculate an output handover spatiotemporal value of 0.92.

[0104] When comparing the spatiotemporal value values ​​of all task points, the spatiotemporal value values ​​corresponding to each node in the graph structure are traversed; the task point with the highest spatiotemporal value value is selected as the dynamic handover anchor point. When multiple task points have the same highest value, the task point with the larger time entropy normalization value of the task chain is selected first; the attribute record of the dynamic handover anchor point includes the unique identifier of the task point, the precise geographic coordinates, and the spatiotemporal value value of the handover.

[0105] To verify the existence of the geographic coordinates of the dynamic handover anchor point in the set of tasks to be handed over, the following verification process is executed: extract the geographic coordinate data of the dynamic handover anchor point; traverse the geographic coordinate records of all task points in the set of tasks to be handed over; calculate the planar Euclidean distance between the coordinates of the dynamic handover anchor point and the coordinates of each task point; if a distance value less than 2 meters is found, it is considered to exist; if no distance value less than 2 meters is found, the exception handling process is triggered: re-execute the message passing mechanism and check the graph structure construction parameter settings.

[0106] In the above process, the node connection distance threshold of 50 meters was set based on the measured data of the maximum physical span of the hospital nursing unit; the number of message passing iterations was determined to be two based on the results of a ten-fold cross-validation experiment, which showed that more than two iterations would lead to model overfitting; the optimization method for setting the number of neurons in the hidden layer of the neural network to 8 was as follows: in the validation set test, when the number of neurons increased from 4 to 8, the prediction accuracy improved by 15%, but when it was further increased to 12, the improvement was less than 2%; the reason for choosing the sigmoid activation function is that its output range perfectly matches the normalized dimension; the distance validation threshold of 2 meters corresponds to the upper limit of the accuracy of the hospital's indoor positioning system. In the graph structure construction, the distance calculation uses the actual metric distance after the plane coordinate system transformation; the edge weight calculation in the message passing mechanism uses the formula of 1 divided by the actual distance value plus 1 to avoid division by zero error. The neural network weight initialization adopts the normal distribution initialization method with the standard deviation being the reciprocal of the square root of the number of input and output nodes, and the training data comes from the hospital's historical handover task record database. The pre-trained model parameters are loaded each time the system starts. When an isolated node appears in the graph structure, the node directly participates in the graph readout function calculation without message passing. Once the dynamic handover anchor point is selected, its geographical coordinates are linked in real time with the handover navigation route planning system and highlighted on the handover electronic dashboard interface. If abnormal fluctuations in values ​​exceeding three standard deviations are detected during normalization, a data review process is triggered to re-acquire the original parameter values.

[0107] S5. Real-time tracking of the spatiotemporal value decay rate of dynamic handover anchor points. When the spatiotemporal value decay rate exceeds the decay rate threshold, the second-highest value task point is updated as the dynamic handover anchor point. The specific implementation is as follows:

[0108] When establishing the spatiotemporal value time series of dynamic handover anchor points, the specific implementation method for time series recording is as follows: Using a fixed sampling frequency of 30 seconds, the spatiotemporal value values ​​of the dynamic handover anchor points are acquired in real time through the shift handover navigation system; each sampling record includes a timestamp and a pair of spatiotemporal value values; the time series data is stored in a circular buffer with a capacity of 20 sampling points. When the number of sampling points exceeds the capacity limit, the oldest historical record is automatically overwritten; the 30-second sampling frequency is based on the monitoring data of the movement speed of hospital nursing staff, which can effectively capture the typical value fluctuation cycle of the ward area. For example, within a 10-minute time period, the system will record 20 sets of data points corresponding to timestamps and values.

[0109] When fitting the decay slope of the spatiotemporal value time series of the handover using a linear regression model, the following steps are performed: First, extract the data from the 10 most recent sampling points in the time series buffer; second, convert the timestamps into offsets in seconds relative to the current time as the time variable; then, use the spatiotemporal value value as the target variable; finally, calculate the linear regression coefficients using the least squares method, and the resulting slope is the spatiotemporal value decay rate. The calculation process includes an outlier filtering mechanism; when the absolute value of the difference between a sampling point and its adjacent points exceeds three times the standard deviation, it is considered an outlier and excluded. For example, if the slope calculated for 10 valid data points is -0.015 units per second, it means that the value decays by 0.015 units per second.

[0110] When comparing the spatiotemporal value decay rate with a preset decay rate threshold in real time, the preset decay rate threshold is determined as follows: by analyzing the historical handover task database, the distribution of the maximum allowable decay rate in 100 successful handover cases is statistically analyzed, and the fifth percentile value is taken as the threshold; this threshold is set to -0.01 per second, indicating that when the actual decay rate exceeds this absolute value, the update mechanism is triggered; the comparison operation is performed every 30 seconds, and when the spatiotemporal value decay rate is detected to be less than the preset decay rate threshold, the subsequent update process is immediately triggered.

[0111] When the spatiotemporal value decay rate is detected to be greater than the decay rate threshold, the current second-highest value task point is obtained from the spatiotemporal value ranking table. The method for generating the spatiotemporal value ranking table is as follows: based on the spatiotemporal value values ​​calculated in step S4, all task points are arranged in descending order of value value to form an ordered list. The second-highest value task point is defined as the next valid entry in the ranking table after the current dynamic handover anchor point. If there are multiple task points with the same value in the ranking table, the task point closest to the geographic coordinates of the current dynamic handover anchor point is selected. For example, when the value of the dynamic handover anchor point is 0.92, the second-highest value task point may correspond to a task point with a value of 0.85.

[0112] When updating a dynamic handover anchor point to the second-highest value task point, the update operation includes three simultaneous actions: replacing the dynamic handover anchor point identifier with the second-highest value task point identifier in the handover navigation system; marking the original anchor point status as decayed in the handover spatiotemporal value ranking table; and updating the visual prompts on the handover electronic dashboard. Task points marked as decayed will be excluded from monitoring in subsequent sampling periods.

[0113] Verify the validity of the updated dynamic handover anchor point geographic coordinates in the set of tasks to be handed over. The validity determination includes dual verification: geographic coordinate verification is performed by calculating the planar Euclidean distance between the updated anchor point coordinates and the coordinates of each task point in the set of tasks to be handed over. If there is a matching point less than 2 meters, the coordinates are considered valid. Task point status verification ensures that the task point is not marked as decayed or completed in the handover spatiotemporal value ranking table. If either verification fails, the exception handling process is triggered: the graph neural network calculation in step S4 is re-executed to generate the latest ranking table.

[0114] In the above process, the linear regression model fitting is achieved using the least squares method; the buffer capacity of 20 is set to balance the requirements of real-time calculation and data stability; the three-standard-deviation threshold for outlier detection is derived from the normal distribution characteristics; the decay rate threshold of -0.01 seconds is determined by taking the fifth percentile of the decay rate distribution of historical successful cases; the geographic coordinate verification threshold of 2 meters matches the positioning accuracy of hospital Bluetooth beacons. When network latency causes data loss during sampling, linear interpolation is used to fill in the missing values; before executing the update operation, it is necessary to confirm that the second-highest value task point is still in an incomplete state; after each update, the time series buffer is reset and sampling records are restarted. The handover spatiotemporal value ranking table is fully updated every hour to ensure that the ranking results reflect the latest task status; the anomaly handling process includes the function of recording error logs and notifying the system administrator. The circular buffer adopts a first-in, first-out management mechanism; the time offset unit in the least squares calculation is seconds, and the value value is retained to three decimal places; the query range of the historical handover task database is set to valid records within the past 90 days; the fifth percentile value is calculated using linear interpolation; task points marked as decayed will not participate in the candidate list for the next 2 hours. In the verification process, the planar Euclidean distance calculation adopts the standard distance formula under the Cartesian coordinate system; after the exception is triggered, the system will recalculate and generate an alarm log within 5 seconds.

[0115] S6. Generate an initial optimized path for the successor based on the current dynamic handover anchor point, and replan the path based on the real-time position of the dynamic handover anchor point when a new emergency task is added. The specific implementation is as follows:

[0116] When obtaining the real-time geographic coordinates of the current dynamic handover anchor point as the starting point of the path, the specific implementation method is as follows: the coordinate data of the location of the dynamic handover anchor point is collected in real time through the Bluetooth beacon network of the hospital's indoor positioning system, with a positioning accuracy of 1.5 meters; the coordinate data format is a combination of X and Y coordinate values ​​in a plane coordinate system, with the unit being meters; the starting point of the path is taken as the arithmetic mean of 5 consecutively collected positioning coordinates within the last 30 seconds. For example, when the coordinate points (10.2, 25.3), (10.3, 25.2), (10.1, 25.4), (10.2, 25.3), and (10.3, 25.2) are collected consecutively, the average X coordinate of 10.22 and the average Y coordinate of 25.28 are calculated as the starting point coordinates.

[0117] When constructing the initial optimized path based on the moving decision tree generation mechanism, the following operations are performed: The moving decision tree complexity evaluation model defined in step S2 is called, with input parameters including the coordinates of the path's starting point and the coordinates of all incomplete task points in the set of tasks to be handed over; the model output is an ordered sequence of nodes connecting the starting point to each task point; the path optimization objective is to minimize the sum of the moving decision tree complexity and the travel time, where the travel time is calculated in real-time based on the channel type weight, for example, the weight coefficient for a regular corridor is 1.0, and the weight coefficient for a stairwell is 1.8; the path generation algorithm uses an improved A* search algorithm, with the heuristic function weight set to 0.7.

[0118] When performing path conflict detection after the initial optimized path is generated, the detection mechanism includes dual verification: fixed obstacle detection identifies sections where the minimum distance between the path segment and obstacles such as walls and fixed equipment is less than 0.1 meters by calling the geometric database of the hospital's building information model; temporary closed passage detection polls the passage status interface of the hospital's facility management system in real time, and marks it as a conflict when the passage identifier corresponding to the path segment returns a disabled status code; the detection frequency is set to be executed once every 15 seconds, and local path replanning is automatically triggered after a conflict segment is found.

[0119] When a new emergency task is detected, the new task identification mechanism is as follows: listen to the task push interface of the hospital nursing scheduling system in real time, and when a task instruction marked as urgent is received, extract the geographical coordinates and task attributes of the task point; at the same time, obtain the real-time location of the dynamic handover anchor point as the starting point of replanning. The location update method is the same as the path starting point acquisition method, that is, take the average value of the positioning coordinates within the last 30 seconds.

[0120] When using a multi-objective optimization algorithm to generate replanning paths, the algorithm implementation includes the following elements: the optimization objective function simultaneously considers minimizing the total travel distance, minimizing the emergency task response time, and minimizing the path complexity; constraints include avoiding all conflict zones and ensuring that the single travel distance does not exceed 500 meters; the solution process uses a genetic algorithm with an elite retention strategy, with an initial population size of 50 paths, 100 iterations, a crossover probability of 0.8, and a mutation probability of 0.1; the genetic algorithm parameters are determined through grid search on historical data; the final output scheme selects the path with the shortest response time from the set of candidate paths that satisfy the multi-objective balance.

[0121] Verify the reachability of the replanned path to the geographic coordinates of all task points in the set of tasks to be handed over. The reachability determination rule is as follows: traverse the path node sequence and calculate the planar Euclidean distance from the coordinates of each task point to the nearest path node; when all distance values ​​are less than 2 meters, it is determined to be reachable; if there is a task point whose distance exceeds the threshold, the segment completion mechanism is triggered: generate a local path loop with a radius of 5 meters with the task point as the center, and find the nearest path node to connect to it.

[0122] In the above process, the Bluetooth beacon positioning data sampling interval is set to 200 milliseconds; the building information model database is automatically updated every morning; the conflict detection threshold of 0.1 meters is set based on the actual safety margin of bed transfer; the emergency task priority label is derived from the emergency level label defined in the hospital's nursing grading system; the elite retention ratio of the genetic algorithm is set to 10% of the population size. Unaffected sections of the original path are retained during path replanning; the A* algorithm is used to calculate the closest path point for local path loop generation; all paths must be manually confirmed through the nursing supervisor's terminal interface before output. The execution status of the original path is frozen when a new emergency task is added; a hard time limit of 5 minutes is set for response time optimization; the path complexity calculation uses the moving decision tree complexity algorithm of step S2. The maximum number of replanning attempts for conflict sections is 3; exceeding this limit triggers a manual intervention process; verified path data is synchronized to the mobile nursing terminal navigation interface in real time. The building geometry database query uses a ray collision detection algorithm; the channel status polling uses an HTTP long polling mechanism to reduce network latency; the path node sequence is stored as a latitude and longitude coordinate linked list structure. The genetic algorithm uses path node index sequences for chromosome encoding. The fitness function weights are set as follows: travel distance 0.4, response time 0.5, and complexity 0.1. These weights are determined based on a survey of nursing experts. Accessibility verification includes slope detection, using elevation data obtained from the hospital's building information model. Paths with a slope exceeding 8 degrees are marked as unreachable. New path segments generated by the segment completion mechanism require re-performing conflict detection. When a replanned path fails to meet all constraints, the system automatically downgrades to a manually specified path mode and generates an alert. If a positioning signal is lost during the average calculation of location coordinates, the nearest valid coordinate is used as a substitute. The multi-objective optimization algorithm is executed within 3 seconds. The minimum spacing between path nodes is 1.5 meters to avoid path overlap.

[0123] Taking the handover process in the internal medicine ward of a hospital as an example, the specific implementation process is as follows: The system first extracts the geographical coordinates and urgency attributes of the 5 unfinished nursing tasks (including 3 infusion tasks and 2 vital sign monitoring tasks) from the hospital information system to generate a set of tasks to be handed over. Then, the A* algorithm is used to generate the optimal path decision tree from the nurses' station to each task point, and the complexity index of each path (such as the number of path nodes, the number of turns, etc.) is calculated. The complexity of the longest path is 0.62 after normalization. Through association rule mining, a strong temporal correlation is found between the infusion tasks, and the time entropy of its task chain is calculated to be 0.8 bits. Based on the weighted calculation of complexity and time entropy, the handover spatiotemporal value of the infusion task point in ward 315 is determined to be the highest (0.86), and it is set as the dynamic handover anchor point. During the handover process, real-time monitoring shows that the value decay rate of this anchor point reaches -0.015 / second, exceeding the threshold of -0.01 / second. The system automatically updates the anchor point to the second highest value monitoring task point in ward 328. When a new emergency blood draw task suddenly occurs, the system immediately replans the route based on the real-time location of ward 328, generates the optimal route sequence that includes the new task points, and verifies the reachability of all task points.

[0124] Example 2: Figure 2 A schematic diagram of a collaborative management system for dynamic scheduling and path optimization of nursing tasks according to the present invention is provided. The collaborative management system for dynamic scheduling and path optimization of nursing tasks includes:

[0125] The task acquisition module is used to obtain the geographical coordinates and task attributes of all unfinished task points in the current shift and generate a set of tasks to be handed over.

[0126] The path decision module is used to generate the optimal path movement decision tree from the successor to each task point in the set of tasks to be handed over and to evaluate the complexity of the movement decision tree.

[0127] The entropy calculation module is used to identify cross-shift task chains in the set of tasks to be handed over through association rule mining, calculate the time interval distribution entropy of adjacent tasks in the cross-shift task chain, and obtain the time-sensitivity entropy of the task chain.

[0128] The value generation module is used to generate the handover spatiotemporal value based on the moving decision tree complexity and task chain time entropy of each task point, and selects the task point with the highest value as the dynamic handover anchor point.

[0129] The anchor point update module is used to track the spatiotemporal value decay rate of dynamic handover anchor points in real time. When the spatiotemporal value decay rate is greater than the decay rate threshold, the second highest value task point is updated as the dynamic handover anchor point.

[0130] The path planning module is used to generate an initial optimized path for the successor based on the current dynamic handover anchor point, and to replan the path based on the real-time location of the dynamic handover anchor point when a new emergency task is added.

[0131] All calculations involved in the embodiments are dimensionless numerical calculations, and the preset parameters and thresholds in the calculations are set by those skilled in the art according to the actual situation.

[0132] It should be noted that this invention can be deployed on the device itself to realize embedded applications, or it can run on a PC or other terminal with a user interface, thereby meeting various hardware environments and usage requirements.

[0133] The above embodiments can be implemented, in whole or in part, by software, hardware, firmware, or any other combination thereof. When implemented using software, the above embodiments can be implemented, in whole or in part, as a computer program product. The computer program product includes one or more computer instructions or computer programs. When the computer instructions or computer programs are loaded or executed on a computer, all or part of the processes or functions described in the embodiments of this application are generated. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions can be stored in a computer-readable storage medium or transmitted from one computer-readable storage medium to another. For example, the computer instructions can be transmitted from one website, computer, server, or data center to another website, computer, server, or data center via wireless or wired transmission; wired transmission methods include optical fiber, twisted pair, coaxial cable, etc.; wireless transmission includes infrared, microwave, etc. The computer-readable storage medium can be any available medium that a computer can access or a data storage device such as a server or data center containing one or more sets of available media. The available medium can be a magnetic medium (e.g., floppy disk, hard disk, magnetic tape), an optical medium (e.g., DVD), or a semiconductor medium. A semiconductor medium can be a solid-state drive.

[0134] Those skilled in the art will understand that, for the sake of convenience and brevity, the specific working processes of the systems, devices, and modules described above can be referred to the corresponding processes in the foregoing method embodiments, and will not be repeated here.

[0135] In the several embodiments provided in this application, it should be understood that the disclosed systems, apparatuses, and methods can be implemented in other ways. For example, the apparatus embodiments described above are merely illustrative; for instance, the division of modules is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple modules or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be through some interfaces; the indirect coupling or communication connection between apparatuses or modules may be electrical, mechanical, or other forms.

[0136] The modules described as separate components may or may not be physically separate. The components shown as modules may or may not be physical modules; they may be located in one place or distributed across multiple network modules. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs.

[0137] In addition, the functional modules in the various embodiments of this application can be integrated into one processing module, or each module can exist physically separately, or two or more modules can be integrated into one module.

[0138] If the aforementioned functions are implemented as software functional modules and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, or a portion of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of this application. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0139] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

[0140] In conclusion, the above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A collaborative management method for dynamic scheduling and path optimization of nursing tasks, characterized in that, include: S1. Obtain the geographic coordinates and task attributes of all unfinished task points in the current shift, and generate a set of tasks to be handed over. S2. Generate the optimal path movement decision tree from the successor to each task point in the set of tasks to be handed over and evaluate the complexity of the movement decision tree; S3. Identify cross-shift task chains in the set of tasks to be handed over by mining association rules, calculate the time interval distribution entropy value of adjacent tasks in the cross-shift task chain, and obtain the time efficiency entropy of the task chain. S4. Based on the complexity of the moving decision tree of each task point and the time entropy of the task chain, generate the handover spatiotemporal value, and select the task point with the highest value as the dynamic handover anchor point. S5. Real-time tracking of the spatiotemporal value decay rate of dynamic handover anchor points. When the spatiotemporal value decay rate is greater than the decay rate threshold, the second highest value task point is updated as the dynamic handover anchor point. S6. Generate the initial optimized path for the successor based on the current dynamic handover anchor point, and replan the path based on the real-time position of the dynamic handover anchor point when a new emergency task is added.

2. The collaborative management method for dynamic scheduling and path optimization of nursing tasks according to claim 1, characterized in that, Obtain the geographic coordinates and task attributes of all incomplete tasks for the current shift, and generate a set of tasks to be handed over, including: Obtain the geographic coordinates and task attributes of all unfinished tasks in the current shift, including task type, task urgency, and task dependency; identify task points with cross-shift continuous execution requirements based on task dependency and mark them as mandatory handover task points; Based on the urgency of the tasks, select non-mandatory handover task points whose urgency level is higher than a preset urgency threshold; merge mandatory handover task points with the selected non-mandatory handover task points to generate a set of tasks to be handed over.

3. The collaborative management method for dynamic scheduling and path optimization of nursing tasks according to claim 2, characterized in that, Generate the optimal path movement decision tree from the successor to each task point in the set of tasks to be handed over, and evaluate the complexity of the movement decision tree, including: A weighted path network was constructed based on the hospital's indoor navigation map, including weights for passage length, access permit status, and elevator waiting time. The weighted path network is input into the Monte Carlo tree search algorithm to simulate the process of generating the optimal path movement decision tree from the current position of the successor to each task point in the set of tasks to be handed over. During the generation of the optimal path movement decision tree, environmental constraints such as temporary closed channels and sudden changes in pedestrian density are dynamically injected. Extract the maximum depth and average number of branches of all nodes in the optimal path movement decision tree; The complexity of the moving decision tree is calculated by multiplying the maximum depth by the average number of branches.

4. The collaborative management method for dynamic scheduling and path optimization of nursing tasks according to claim 3, characterized in that, By identifying cross-shift task chains in the set of tasks to be handed over through association rule mining, and calculating the time interval distribution entropy of adjacent tasks in the cross-shift task chain, the time-sensitivity entropy of the task chain is obtained, including: Analyze the task dependency attributes of all task points in the set of tasks to be handed over, identify the combination of task points that have the requirement of continuous execution across shifts, and define it as a cross-shift task chain. Obtain historical execution timestamp data for each task point in the cross-shift task chain from the hospital information system; Calculate the historical time intervals between adjacent task points in a cross-shift task chain and generate a set of time interval distribution probability. Based on the probability set of time interval distributions, the entropy value of the time interval distribution is calculated using the Shannon entropy formula. Multiplying the time interval distribution entropy value by the environmental decay factor yields the task chain time-dependent entropy.

5. The collaborative management method for dynamic scheduling and path optimization of nursing tasks according to claim 4, characterized in that, The environmental degradation factor is derived from a weighted fusion of weather index and seasonal disease outbreak coefficient.

6. The collaborative management method for dynamic scheduling and path optimization of nursing tasks according to claim 4, characterized in that, Based on the complexity of the decision tree for each task point and the time entropy of the task chain, the spatiotemporal value of the handover is generated. The task point with the highest value is selected as the dynamic handover anchor point, including: Normalize the movement decision tree complexity and task chain time entropy of each task point to the same numerical dimension. Construct a graph structure with task points as nodes, where the node feature vectors contain the normalized moving decision tree complexity and task chain time entropy; The feature vectors of adjacent task points are aggregated through the message passing mechanism of a graph neural network. For each task point, a graph readout function is applied to generate the handover spatiotemporal value, where the graph readout function includes a multilayer perceptron and a sigmoid activation function; Compare the handover spatiotemporal value values ​​of all task points, and select the task point with the highest handover spatiotemporal value value as the dynamic handover anchor point.

7. The collaborative management method for dynamic scheduling and path optimization of nursing tasks according to claim 6, characterized in that, Real-time tracking of the spatiotemporal value decay rate of dynamic handover anchor points. When the spatiotemporal value decay rate exceeds a decay rate threshold, the second-highest value task point is updated as the dynamic handover anchor point, including: Establish a time series of spatiotemporal values ​​of dynamic handover anchor points and record the spatiotemporal value values ​​of handover at a fixed sampling frequency; A linear regression model was used to fit the decay slope of the spatiotemporal value time series of the handover, which was taken as the spatiotemporal value decay rate. The spatiotemporal value decay rate is compared with a preset decay rate threshold in real time. When the spatiotemporal value decay rate is detected to be greater than the decay rate threshold, the current second highest value task point is obtained from the spatiotemporal value sorting table. Update the dynamic handover anchor point to the second highest value task point, and mark the original anchor point's status as decayed in the handover spatiotemporal value ranking table.

8. The collaborative management method for dynamic scheduling and path optimization of nursing tasks according to claim 7, characterized in that, An initial optimized path is generated for the successor based on the current dynamic handover anchor point, and the path is replanned based on the real-time position of the dynamic handover anchor point when a new emergency task is added, including: The real-time geographic coordinates of the current dynamic handover anchor point are obtained as the starting point of the path. Based on the moving decision tree generation mechanism, an initial optimized path is constructed from the starting point of the path to the remaining task points in the set of tasks to be handed over. After the initial optimized path is generated, path conflict detection is performed to identify path segments that overlap with fixed obstacles and temporary closed passages in the hospital. When a new emergency task is detected, the real-time location of the dynamic handover anchor point is obtained as the starting point for replanning. A multi-objective optimization algorithm is used to generate a replanning path from the replanning starting point to the newly added urgent task point and the incomplete task points in the set of tasks to be handed over.

9. A collaborative management system for dynamic scheduling and path optimization of nursing tasks, used to implement the collaborative management method for dynamic scheduling and path optimization of nursing tasks as described in any one of claims 1-8, characterized in that, include: The task acquisition module is used to obtain the geographical coordinates and task attributes of all unfinished task points in the current shift and generate a set of tasks to be handed over. The path decision module is used to generate the optimal path movement decision tree from the successor to each task point in the set of tasks to be handed over and to evaluate the complexity of the movement decision tree. The entropy calculation module is used to identify cross-shift task chains in the set of tasks to be handed over through association rule mining, calculate the time interval distribution entropy of adjacent tasks in the cross-shift task chain, and obtain the time-sensitivity entropy of the task chain. The value generation module is used to generate the handover spatiotemporal value based on the moving decision tree complexity and task chain time entropy of each task point, and selects the task point with the highest value as the dynamic handover anchor point. The anchor point update module is used to track the spatiotemporal value decay rate of dynamic handover anchor points in real time. When the spatiotemporal value decay rate is greater than the decay rate threshold, the second highest value task point is updated as the dynamic handover anchor point. The path planning module is used to generate an initial optimized path for the successor based on the current dynamic handover anchor point, and to replan the path based on the real-time location of the dynamic handover anchor point when a new emergency task is added.

Images