Multi-task cooperation and scheduling method and system applied to old-age service robot

Abstract

The application provides a multi-task cooperation and scheduling method and system applied to an old-age service robot, relates to the technical field of intelligent old-age service, acquires an old-age service demand instruction sequence and an old-age environment perception information flow; service scene analysis processing is performed on the two, a service object state feature set containing a service object position and a behavior feature and a service resource state feature set containing a service robot position and a load feature are obtained, task response association construction is performed from the two, a task resource pairing relationship graph representing a multi-dimensional adaptive association strength and a task execution priority sequence representing an execution urgency degree order are generated; multi-task cooperation and scheduling are performed according to the pairing relationship graph and the priority sequence, a service robot action instruction set containing a robot identifier, a target position, an action type and an execution time sequence is generated and is distributed to a corresponding control terminal. The application realizes dynamic adaptation and urgency degree priority cooperation and scheduling of multi-robot tasks in an old-age scene.

Description

Technical Field

[0001] This invention relates to the field of smart elderly care service technology, and more specifically, to a multi-task collaboration and scheduling method and system for elderly care service robots. Background Technology

[0002] As an important technological means to address population aging, elderly care service robots undertake various service tasks in scenarios such as home care for elderly people living alone and auxiliary services in nursing homes, including meal delivery, medication reminders, fall detection, and walking assistance. Since multiple service robots and multiple elderly people requiring services often exist simultaneously in elderly care scenarios, how to rationally allocate tasks among multiple robots and how to coordinate and schedule them according to the urgency of tasks and resource availability have become key issues affecting the quality and efficiency of elderly care services.

[0003] In existing technologies, task scheduling for elderly care service robots often employs simple first-come, first-served queues or priority queuing mechanisms based on a single rule. For example, some solutions allocate tasks sequentially based on the time of instruction receipt, assigning the earliest arriving instruction to the nearest robot. Other solutions use static priority lists based on task type, such as always placing emergency call tasks with the highest priority, while executing routine tasks like meal delivery and companionship in a fixed order. Still other solutions centrally assign tasks to all robots through a central server, but their matching logic typically only considers the geometric distance between the robot and the service recipient or whether the task type matches, without jointly modeling the real-time behavioral state of the service recipient and the real-time load capacity of the robot.

[0004] However, the elderly care scenario is highly dynamic and uncertain. The behavioral state of the elderly (such as whether they are walking, sleeping, or at risk of falling) and the load state of the robot (such as whether it is currently performing other tasks, remaining battery power, and the occupancy status of the robotic arm) change rapidly over time. Existing scheduling methods based on static distance or static priority cannot dynamically perceive this multi-dimensional state coupling relationship, resulting in a serious mismatch between task allocation and actual service needs. For example, assigning meal delivery tasks to robots that are performing emergency care and are already fully loaded, or prioritizing low-urgency tasks over high-urgency tasks, seriously affects the timeliness and safety of elderly care services. Summary of the Invention

[0005] In view of the aforementioned problems, and in conjunction with the first aspect of the present invention, embodiments of the present invention provide a multi-task collaboration and scheduling method applied to elderly care service robots, the method comprising:

[0006] The system acquires a sequence of elderly care service demand instructions and an information flow for elderly care environment perception. The sequence of elderly care service demand instructions includes multiple elderly care service demand instruction units, each of which corresponds to a service demand type identifier and a service demand trigger timestamp. The information flow for elderly care environment perception includes a multi-dimensional environmental status perception data flow collected by multiple perception nodes within the elderly care environment.

[0007] The elderly care service demand instruction sequence and the elderly care environment perception information flow are processed by service scenario parsing to obtain a service object status feature set and a service resource status feature set. The service object status feature set includes service object location features and service object behavior features. The service resource status feature set includes service robot location features and service robot load features.

[0008] Based on the service object status feature set and the service resource status feature set, task response association construction processing is performed to generate a task resource pairing relationship graph and a task execution priority sequence. The task resource pairing relationship graph is used to characterize the multidimensional adaptation association strength between each elderly care service demand instruction unit and the available service robot, and the task execution priority sequence is used to characterize the order of execution urgency of each elderly care service demand instruction unit.

[0009] Based on the task resource pairing relationship diagram and the task execution priority sequence, multi-task collaborative scheduling is performed to generate a service robot action instruction set. The service robot action instruction set is then distributed to the corresponding service robot control terminal to trigger service execution operations. The service robot action instruction set contains multiple service robot action instruction units, each of which corresponds to a service robot identifier, a target service object location, a service action type, and a service action execution sequence.

[0010] Furthermore, embodiments of the present invention also provide a multi-task collaboration and scheduling system for elderly care service robots, comprising:

[0011] A processor; a machine-readable storage medium for storing machine-executable instructions of the processor; wherein the processor is configured to execute the above-described multi-task coordination and scheduling method for elderly care service robots by executing the machine-executable instructions.

[0012] Based on the above, by analyzing the service scenario of the elderly care service demand instruction sequence and the elderly care environment perception information flow, it is possible to simultaneously extract the location and behavioral characteristics of service objects and the location and load characteristics of service robots from the multi-dimensional environmental state perception data flow. This allows subsequent task scheduling decisions to be based on a comprehensive perception of the real-time status of both service supply and demand sides, rather than relying on static distance or type matching. Furthermore, by constructing task response associations based on the service object state feature set and the service resource state feature set, the generated task resource pairing relationship graph can quantitatively characterize the comprehensive matching degree between each elderly care service demand instruction unit and each available service robot in a multi-dimensional adaptation association strength manner. Simultaneously, the task execution priority sequence can reflect the urgency of each demand. The timing importance of instructions is determined, and together these two factors constitute a dual-dimensional scheduling basis that balances adaptability and urgency. Furthermore, multi-task collaborative scheduling is performed based on the task resource pairing relationship diagram and the task execution priority sequence. Each action instruction unit in the generated service robot action instruction set clearly includes the service robot identifier, the location of the target service object, the service action type, and the service action execution sequence. This allows multiple service robots to execute non-conflicting and highly efficient service operations within the same time window based on the globally optimal pairing relationship and priority order. This achieves a unified dynamic adaptability, execution urgency, and resource load balancing in multi-task allocation in elderly care service scenarios, effectively avoiding the task mismatch and service delay problems caused by ignoring the coupling relationship between the real-time behavior state of the service object and the real-time load state of the robot in existing technologies. Attached Figure Description

[0013] Figure 1 This is a schematic diagram of the execution flow of the multi-task collaboration and scheduling method for elderly care service robots provided in an embodiment of the present invention.

[0014] Figure 2 This is one of the front-end interface diagrams of the multi-task collaboration and scheduling method for elderly care service robots provided in the embodiments of the present invention;

[0015] Figure 3 This is the second schematic diagram of the front-end interface of the multi-task collaboration and scheduling method for elderly care service robots provided in this embodiment of the invention. Detailed Implementation

[0016] Figure 1This is a flowchart illustrating a multi-task collaboration and scheduling method for elderly care service robots according to an embodiment of the present invention. This embodiment provides a specific implementation of the multi-task collaboration and scheduling method for elderly care service robots, using the collaborative operation of a heterogeneous service robot cluster deployed in an elderly care community as a unified application scenario. This scenario includes residential units, public corridors, activity halls, and medical stations, and is equipped with location-aware sensor arrays, behavior-aware sensor arrays, environmental status sensor arrays, as well as delivery robots, inspection robots, health monitoring robots, and cleaning robots.

[0017] Step S110: Obtain the elderly care service demand instruction sequence and the elderly care environment perception information flow. The elderly care service demand instruction sequence contains multiple elderly care service demand instruction units. Each elderly care service demand instruction unit corresponds to a service demand type identifier and a service demand trigger timestamp. The elderly care environment perception information flow contains a multi-dimensional environmental status perception data flow collected by multiple perception nodes in the elderly care environment.

[0018] Each residential unit's smart interactive terminal, voice acquisition device, and wearable assistance device continuously receive service requests. Each service request is encapsulated as a senior care service request instruction unit, containing a service request type identifier field and a service request trigger timestamp field. The service request type identifier is an enumerated string, obtained by the natural language understanding component parsing the request content and mapping it to a service type dictionary. The service request trigger timestamp is stored as a Coordinated Universal Time (UTC) millisecond-level integer value. All senior care service request instruction units arriving within a preset aggregation window length are arranged in ascending order of their service request trigger timestamps, forming a senior care service request instruction sequence.

[0019] Each sensing node reports sensing data records at a fixed frequency. Each sensing data record includes a source sensing node identifier field, a sensing data type encoding field, and a corresponding numerical field. All sensing data records are aggregated into an elderly care environment sensing information stream indexed by timestamps. The transmission layer adopts a publish-subscribe model, with subscribers filtering and receiving data by topic. Informed consent authorization signed by the service recipient or their legal guardian has been obtained before data collection. The original depth vision frames are processed in real-time on the edge computing units of the sensing nodes to blur facial regions and desensitize skeletal key point coordinates, and only the desensitized pose feature numerical sequence is uploaded.

[0020] Step S120: Perform service scenario parsing processing on the elderly care service demand instruction sequence and the elderly care environment perception information flow to obtain a service object status feature set and a service resource status feature set. The service object status feature set includes service object location features and service object behavior features, and the service resource status feature set includes service robot location features and service robot load features.

[0021] The system performs service scenario parsing and processing on the sequence of instructions for elderly care service demands and the information flow of elderly care environment perception, and extracts a structured set of features from heterogeneous data.

[0022] Step S121: Perform demand attribute parsing processing on the elderly care service demand instruction sequence, extract the service demand type identifier and service demand trigger timestamp from each elderly care service demand instruction unit, map the service demand type identifier to demand intent category label and demand urgency label, and map the service demand trigger timestamp to demand trigger time node.

[0023] Iterate through each elderly care service demand instruction unit in the sequence, extracting its service demand type identifier and service demand trigger timestamp. Query a pre-stored mapping table. The row structure of this table includes service demand type identifier values, demand intent category label values, and demand urgency label values. Perform a full match query using the service demand type identifier as the search key to retrieve the corresponding demand intent category label and demand urgency label. The demand intent category label is an enumerated string. The demand urgency label is an ordered enumerated string, with values ​​representing regular, priority, and urgency. The mapping relationship is pre-defined based on the objective timeliness requirements of the service content corresponding to each service demand type identifier. The service demand trigger timestamp is converted to the local time zone and rounded to the nearest second as the demand trigger time node.

[0024] Step S122: Perform sensing data splitting processing on the elderly care environment sensing information flow. Based on the source sensing node identifier of each sensing data in the elderly care environment sensing information flow, divide the elderly care environment sensing information flow into a service object sensing data subset and a service resource sensing data subset.

[0025] The elderly care environment perception information flow is a hybrid data flow. Each perception data record has a pre-configured monitoring category attribute for its source perception node identifier. The monitoring category attribute value includes both categories representing service object monitoring and service resource monitoring. The perception data records in the elderly care environment perception information flow are read one by one, their source perception node identifiers are extracted, and the corresponding monitoring category attribute is queried. If the monitoring category attribute is for service object monitoring, the perception data record is categorized into the service object perception data subset; if the monitoring category attribute is for service resource monitoring, the perception data record is categorized into the service resource perception data subset. The separated service object perception data subset and service resource perception data subset are then processed separately.

[0026] Step S123: Extract service object location perception data sequence and service object behavior perception data sequence from the service object perception data subset; perform location trajectory tracking processing on the service object location perception data sequence to generate a continuous location change trajectory for each service object; and perform behavior pattern recognition processing on the service object behavior perception data sequence to generate a behavior state description vector for each service object.

[0027] The service object perception data subset is grouped by service object identifier, and each group is sorted in ascending order by timestamp. From the sorted grouped data, location-type records and behavior-type records are filtered out according to the perception data type encoding, which respectively constitute the service object location perception data sequence and the service object behavior perception data sequence.

[0028] Each record in the service object location-aware data sequence contains a timestamp and the two-dimensional planar coordinates reported by the sensing node at that moment. Position trajectory tracking is implemented using a state estimator based on a Kalman filter framework. The state vector of the state estimator consists of four scalars: two-dimensional position coordinates and two-dimensional velocity components. The state transition matrix is ​​constructed based on the sampling interval, with the transition coefficients between the position and velocity components in the state transition matrix being the sampling interval. The observation vector consists of measured coordinates, and the observation matrix maps the position components in the state vector to the observation space. Recursive operations of prediction and update steps are performed frame-by-frame on the service object location-aware data sequence. The prediction step calculates the prior state estimate and prior error covariance, while the update step calculates the Kalman gain matrix and then corrects the posterior state estimate and posterior error covariance. The position components in the posterior state estimate of each frame are taken as smoothed position coordinates and concatenated temporally to obtain the continuous position change trajectory of the service object.

[0029] The service object behavior perception data sequence consists of depth image frames acquired by a depth vision sensor arranged in chronological order. Each depth image frame has H rows and W columns of pixels, with each pixel value representing a depth distance. Behavior pattern recognition processing is performed using a pre-trained 3D convolutional neural network (CNNN) model. The input to the CNNN model is a four-dimensional tensor formed by stacking consecutively sampled depth image frames within a time window. The feature extraction part of the CNNN model consists of five sets of alternating 3D convolutional layers and 3D max-pooling layers. In each set, the 3D convolutional layers use 3D convolutional kernels to simultaneously perform sliding convolution along the time, height, and width dimensions, generating multi-channel 3D feature maps. The 3D max-pooling layer downsamples by taking the maximum value within the local spatiotemporal neighborhood. At the end of the feature extraction part, a 3D global average pooling layer is connected to calculate the average of the 3D feature maps for each channel across all time, height, and width dimensions, outputting a feature vector. The feature vector is fed into a fully connected classification layer. The number of output neurons in the fully connected classification layer equals the total number of preset behavior categories, including lying down to rest, sitting to read, walking slowly, standing to talk, eating, and falling. The output value of the fully connected classification layer is processed by a normalized exponential function to obtain a probability distribution vector, which is the vector describing the behavior state of the service object within the time window.

[0030] Step S124: Extract service robot position perception data sequence and service robot load perception data sequence from the service resource perception data subset; perform position trajectory tracking processing on the service robot position perception data sequence to generate a continuous position change trajectory for each service robot; and perform load state parsing processing on the service robot load perception data sequence to generate a load state description vector for each service robot.

[0031] The service resource perception data subset is grouped by service robot identifier, and each group is sorted in ascending order by timestamp. From the sorted grouped data, location-type records and load-type records are selected according to the perception data type encoding to form the service robot location perception data sequence and the service robot load perception data sequence, respectively.

[0032] Each record in the service robot's position awareness data sequence includes a timestamp and the two-dimensional plane coordinates calculated jointly by the robot chassis odometry and inertial measurement unit at that moment. The service robot's position awareness data sequence is then tracked using the same Kalman filter framework state estimator as in step S123, outputting the continuous position change trajectory of each service robot.

[0033] Each record in the service robot's load perception data sequence includes a timestamp, the remaining energy value at that moment, the current load weight, the number of currently assigned tasks, and a list of currently assigned task types. Load status parsing integrates these multi-source load indicators into a single load status description vector. This vector is a fixed-length array with dimensions equal to the number of selected load indicator types. The first dimension of the vector is the remaining energy value divided by the battery's rated total energy; the second dimension is the current load weight divided by the maximum safe load weight; the third dimension is the number of currently assigned tasks divided by the maximum concurrent task capacity; and the fourth dimension is the sum of the preset complexity weights for each task type in the current assigned task type list divided by a normalization constant. After standardization, each original load indicator is mapped from its original value range to a dimensionless value between 0 and 1.

[0034] Step S125: Combine the demand intent category label, the demand urgency label, and the demand trigger time node to form a service demand description vector, and associate the service demand description vector with the corresponding service object identifier for storage.

[0035] For each elderly care service demand instruction unit, a service demand description vector is constructed. This vector consists of an enumerated string, an ordered enumerated string, and an integer value. The service demand description vector and its corresponding service object identifier are combined as a record and written into a key-value mapping structure in memory, using the service object identifier as the key and the service demand description vector as the value, to complete the associative storage.

[0036] Step S126: Combine the continuous position change trajectory of each service object and the behavior state description vector of each service object to form the service object state feature set; combine the continuous position change trajectory of each service robot and the load state description vector of each service robot to form the service resource state feature set.

[0037] For each service object identifier, the continuous location change trajectory corresponding to that identifier is used as the service object's location feature, and the behavioral state description vector corresponding to that identifier is used as the service object's behavioral feature. Together, these constitute the service object's state feature entry. The state feature entries corresponding to all service object identifiers are then aggregated to obtain the service object's state feature set.

[0038] For each service robot identifier, the continuous position change trajectory corresponding to that identifier is used as the service robot's position feature, and the load state description vector corresponding to that identifier is used as the service robot's load feature. Together, these constitute the service robot's state feature entry. The state feature entries corresponding to all service robot identifiers are then aggregated to obtain the service resource state feature set.

[0039] Step S130: Based on the service object status feature set and the service resource status feature set, perform task response association construction processing to generate a task resource pairing relationship graph and a task execution priority sequence. The task resource pairing relationship graph is used to characterize the multidimensional adaptation association strength between each elderly care service demand instruction unit and the available service robot. The task execution priority sequence is used to characterize the order of execution urgency of each elderly care service demand instruction unit.

[0040] After the service object status feature set and the service resource status feature set have been generated, the task response association construction process is carried out.

[0041] Step S131: Calculate the current location region of each service object based on the location change trajectory of each service object in the service object status feature set, and spatiotemporally bind the current location region of each service object with the demand triggering time node corresponding to the service object in the elderly care service demand instruction sequence to generate a service demand entry with spatial location marker.

[0042] The system iterates through the continuous location change trajectory corresponding to each service object identifier in the service object status feature set, extracts the smoothed location coordinates of the latest time step in the continuous location change trajectory, and performs point-to-surface inclusion determination on the smoothed location coordinates and pre-divided grid areas to determine the grid area number into which the smoothed location coordinates fall. It then searches for the elderly care service demand instruction unit associated with the service object identifier in the elderly care service demand instruction sequence and obtains the demand trigger time node of the elderly care service demand instruction unit. The grid area number and demand trigger time node are added as additional attributes and merged with the original fields of the elderly care service demand instruction unit to form a service demand entry with spatial location markings.

[0043] Step S132: Calculate the current location region of each service robot based on the position change trajectory of each service robot in the service resource status feature set, bind the current location region of each service robot with the load status description vector of each service robot, and generate a service resource entry with spatial location marker and load marker.

[0044] The system iterates through the continuous position change trajectory corresponding to each service robot identifier in the service resource status feature set, extracts the smoothed position coordinates of the latest time step in the continuous position change trajectory, and determines the grid region number into which the smoothed position coordinates fall by using point-surface inclusion determination. It then extracts the load status description vector corresponding to the service robot identifier from the service resource status feature set. The grid region number and the load status description vector are added as additional attributes and merged with the service robot identifier to form a service resource entry with spatial location and load markers.

[0045] Step S133: For each service demand item with a spatial location marker, calculate the spatial distance between the service demand item and all service resource items with spatial location markers. Based on the spatial distance, the mobility parameters of the service robot, and the path accessibility in the elderly care environment, determine the accessibility parameters of the service robot.

[0046] For each service request item, iterate through all service resource items. Extract the center coordinates corresponding to the grid area number of the service request item and the center coordinates corresponding to the grid area number of the service resource item. The spatial distance is equal to the square root of the sum of the squares of the differences between the two center coordinates. Obtain the preset average movement speed of the service robot corresponding to the service resource item and the path accessibility coefficient of the elderly care environment. The path accessibility coefficient is a dimensionless value between 0 and 1, determined by the proportion of accessible paths in the environment and the real-time congestion level. The service robot accessibility parameter is equal to the path accessibility coefficient multiplied by the average movement speed, then divided by the sum of the spatial distance and a preset positive decimal.

[0047] Step S134: For each service demand entry with a spatial location marker, calculate the task load matching degree between the demand intent category label of the service demand entry and the load markers of all service resource entries, and determine the service robot task adaptability parameters based on the task load matching degree.

[0048] For each service request item, extract its request intent category label and iterate through all service resource items. Extract the load status description vector for each service resource item. A pre-stored task load matching matrix is ​​used, where rows correspond to the values ​​of the request intent category labels, columns correspond to the dimension indices of the load status description vectors, and each element in the task load matching matrix represents the request weight of that request intent category label for that load dimension. The task load matching degree is equal to the sum of the element-wise products of the row vector corresponding to the request intent category label in the task load matching matrix and the load status description vector. The service robot task adaptability parameter is obtained by transforming the task load matching degree using a preset monotonically increasing mapping function; the service robot task adaptability parameter takes values ​​between 0 and 1.

[0049] Step S135: The service robot reachability parameters, the service robot task adaptability parameters, and the urgency label of the service requirement item are fused to generate a multi-dimensional adaptation association strength between each service requirement item and each service resource item. The task resource pairing relationship graph is constructed based on the multi-dimensional adaptation association strength. Each node in the task resource pairing relationship graph is a service requirement item or a service resource item, and the weight of each edge is the multi-dimensional adaptation association strength.

[0050] For each pair of service request items and service resource items, obtain the corresponding service robot accessibility parameters, service robot task adaptability parameters, and the urgency quantification value corresponding to the urgency label of the service request item. The urgency quantification value is obtained by querying a preset mapping table from urgency to numerical values. The multidimensional adaptation association strength equals the service robot accessibility parameter multiplied by a first weight coefficient, plus the service robot task adaptability parameter multiplied by a second weight coefficient, plus the urgency quantification value multiplied by a third weight coefficient. The first, second, and third weight coefficients are all preset values, all between 0 and 1, and their sum is 1. Based on the multidimensional adaptation association strength values ​​of all pairs of items, construct a task-resource pairing relationship graph. The task-resource pairing relationship graph is a bipartite graph structure, containing two types of nodes: one type of node is the service request item node, with each service request item node corresponding to one service request item; the other type of node is the service resource item node, with each service resource item node corresponding to one service resource item. If a service requirement item and a service resource item are paired, an edge is connected between them, and the weight of the edge is the corresponding multidimensional adaptation association strength value.

[0051] Step S136: Extract the urgency level tags and trigger time nodes of all service request items, determine the urgency weight of each service request item based on the urgency level tags, and determine the waiting time length of each service request item based on the trigger time nodes.

[0052] Extract the urgency level tags and trigger timestamps for all service request items. Based on the urgency level tags, determine the urgency weight of each service request item using a pre-defined mapping table. The urgency weight is set to preset values ​​corresponding to high, medium, and low urgency levels. The waiting time is equal to the current system time minus the request trigger time, converted to seconds.

[0053] Step S137: Convert the waiting time length of each service request item into a time urgency quantification value by combining it with a preset unit time urgency conversion coefficient; perform a comprehensive urgency ranking calculation on each service request item according to the urgency weight of each service request item and the time urgency quantification value to generate the task execution priority sequence, where each element in the task execution priority sequence corresponds to a service request item and its comprehensive urgency ranking value.

[0054] For each service request item, the time urgency metric is equal to the waiting time multiplied by a preset unit time urgency conversion factor. The overall urgency ranking value is equal to the urgency weight plus the time urgency metric. All service request items are sorted in descending order of their overall urgency ranking values ​​to generate a task execution priority sequence. Each element in the task execution priority sequence contains the identifier of a service request item and its overall urgency ranking value.

[0055] Step S140: Perform multi-task collaborative scheduling processing based on the task resource pairing relationship diagram and the task execution priority sequence to generate a service robot action instruction set, and distribute the service robot action instruction set to the corresponding service robot control terminal to trigger service execution operations. The service robot action instruction set contains multiple service robot action instruction units, each service robot action instruction unit corresponding to a service robot identifier, a target service object location, a service action type, and a service action execution sequence.

[0056] After the task resource pairing diagram and task execution priority sequence have been generated, multi-task collaborative scheduling is performed.

[0057] Step S141: Extract service requirement items one by one from the task execution priority sequence in descending order of comprehensive urgency value, and take the currently extracted service requirement item as the current task to be scheduled.

[0058] The task execution priority sequence is traversed from beginning to end. Each time, the service request item with the highest overall urgency ranking value is extracted as the current task to be scheduled. After extraction, the service request item is marked as processed from the task execution priority sequence.

[0059] Step S142: Search for all service resource entries connected to the current task to be scheduled in the task resource pairing relationship graph, obtain the multidimensional adaptation association strength of each connected service resource entry, and generate a candidate service robot list in descending order of the multidimensional adaptation association strength.

[0060] In the task resource pairing graph, starting from the service requirement node corresponding to the currently scheduled task, find all service resource node nodes connected to that service requirement node by an edge. Read the weight value of each edge. Sort all connected service resource node nodes in descending order of their edge weight values ​​to generate a candidate service robot list. Each element in the candidate service robot list represents a service resource item and its multidimensional adaptation association strength.

[0061] Step S143: for each candidate service robot in the candidate service robot list, extract a load state description vector of the candidate service robot, calculate a current remaining load capacity of the candidate service robot, compare the current remaining load capacity with an expected task load corresponding to a demand intention category label of the current to-be-scheduled task, filter out candidate service robots whose remaining load capacity is less than the expected task load, and generate an available service robot list.

[0062] For each candidate service robot in the candidate service robot list, a load state description vector is extracted from its service resource entry. Let the first dimension of the load state description vector be L1, the second dimension be L2, and the third dimension be L3, then the current remaining load capacity Crem = L1*(1-L2)*(1-L3). A preset expected task load threshold Lth corresponding to the demand intention category label of the current to-be-scheduled task is obtained. If Crem < Lth, the candidate service robot is filtered out; if Crem ≥ Lth, the candidate service robot is retained. All retained candidate service robots form the available service robot list.

[0063] Step S144: for each available service robot in the available service robot list, extract a current position area of the available service robot and a spatial position marker of the current to-be-scheduled task, and calculate an estimated movement time for the available service robot to move from the current position area to a service object position corresponding to the spatial position marker.

[0064] For each available service robot in the available service robot list, extract the center coordinate corresponding to the grid area number from its service resource entry, and extract the center coordinate corresponding to the grid area number from the service demand entry of the current to-be-scheduled task. Let the two center coordinates be (x1, y1) and (x2, y2) respectively, then the spatial distance D = ((x1-x2)^2+(y1-y2)^2)^(1 / 2). Let the path passage coefficient be P and the average movement speed be Vavg, then the estimated movement time Tmove = D / (P*Vavg).

[0065] Step S145: determine a service arrival time of the available service robot for executing the current to-be-scheduled task according to the estimated movement time and a demand trigger time node of the current to-be-scheduled task, compare the service arrival time with a preset latest arrival time of the current to-be-scheduled task, filter out available service robots whose service arrival time is later than the preset latest arrival time, and generate a compliant service robot list.

[0066] The service arrival time equals the current system time plus the estimated travel time. The system retrieves the preset latest arrival time for the currently scheduled task. If the service arrival time is greater than the preset latest arrival time, the available service robot is filtered out; if the service arrival time is less than or equal to the preset latest arrival time, the available service robot is retained. All retained available service robots constitute the compliant service robot list.

[0067] Step S146: Select the compliance service robot with the highest multidimensional adaptation association strength and the shortest estimated movement time from the compliance service robot list as the optimal assigned service robot, bind the current task to be scheduled with the optimal assigned service robot, and generate a task assignment relationship record.

[0068] For each compliant service robot in the compliant service robot list, obtain its multi-dimensional adaptation association strength and estimated movement time. Select the compliant service robot with the highest multi-dimensional adaptation association strength and the lowest estimated movement time as the optimal assigned service robot. Bind the currently scheduled task to the optimal assigned service robot to generate a task assignment relationship record, which includes the identifier of the currently scheduled task and the identifier of the optimal assigned service robot.

[0069] Step S147: Use the optimal assigned service robot identifier in the task assignment relationship record as the service robot identifier, use the spatial location mark of the current task to be scheduled as the target service object location, map the demand intent category label of the current task to be scheduled as the service action type, use the service arrival time as the service action execution sequence, and combine them to generate a service robot action instruction unit.

[0070] Extract the optimal assigned service robot identifier from the task assignment relationship record. Extract the center coordinates corresponding to the grid area number from the service request entry of the current task to be scheduled as the target service object location. Map the requirement intent category label of the current task to be scheduled to a service action type using a preset service action type mapping table. Use the service arrival time as the service action execution sequence. Combine the service robot identifier, target service object location, service action type, and service action execution sequence to generate a service robot action instruction unit.

[0071] Step S148: Add the service robot action instruction unit to the service robot action instruction set, remove the currently scheduled task from the task execution priority sequence, remove the connection edge related to the currently scheduled task from the task resource pairing relationship graph, update the load state description vector of the optimal assigned service robot, and repeat the above steps until the task execution priority sequence is empty to obtain the complete service robot action instruction set.

[0072] Add the service robot action instruction unit to the service robot action instruction set. Remove the currently scheduled task from the task execution priority sequence. Remove all edges connected to the node corresponding to the currently scheduled task from the task resource pairing graph. Let the third dimension of the original load state description vector of the optimal assigned service robot be L3, the original number of assigned tasks be Ntask, and the maximum concurrent task capacity be Nmax. After the update, the value of the third dimension is (Ntask+1) / Nmax. If a new task type is added, the fourth dimension is recalculated as the sum of the preset complexity weights of each task type in the current assigned task type list divided by the normalization constant. Repeat steps S141 to S148 until the task execution priority sequence is empty, obtaining a complete service robot action instruction set. Distribute each service robot action instruction unit in the service robot action instruction set to the service robot control terminal with the corresponding service robot identifier through the wireless local area network, triggering the service robot to perform a service operation.

[0073] Step S210: After distributing the set of service robot action instructions to the corresponding service robot control terminals, receive the service execution status feedback information periodically reported by each service robot control terminal. The service execution status feedback information includes the service robot identifier, the type of the currently executed action, the current execution progress status, and the current location coordinates.

[0074] After receiving the service robot action command unit, each service robot control terminal reports service execution status feedback information to the data processing server at fixed intervals. The service execution status feedback information includes the service robot identifier, the type of the current action being performed, the current execution progress status, and the current location coordinates. The current execution progress status is an integer percentage from 0 to 100, and the current location coordinates are two-dimensional planar coordinates.

[0075] Step S220: Group and aggregate all received service execution status feedback information according to the service robot identifier to generate a continuous status feedback sequence for each service robot. Perform state change analysis processing on the continuous status feedback sequence of each service robot to extract the execution progress change curve and position movement trajectory curve of each service robot.

[0076] Service execution status feedback information is grouped and aggregated according to service robot identifiers. Each group is sorted in ascending order by the reported timestamp, generating a continuous status feedback sequence for each service robot identifier. A sequence consisting of timestamps and the current execution progress status is extracted from the continuous status feedback sequence to generate an execution progress change curve. Finally, a sequence consisting of timestamps and the current location coordinates is extracted from the continuous status feedback sequence to generate a location movement trajectory curve.

[0077] Step S230: Compare the execution progress change curve of each service robot with the expected execution progress corresponding to the demand intent category label of the service demand item bound to the service robot, calculate the execution progress deviation, and compare the position movement trajectory curve of each service robot with the service action execution sequence planned in the action instruction unit of the service robot, calculate the position movement trajectory deviation.

[0078] For each service robot identifier, a preset expected execution progress curve is obtained based on the demand intent category label of the service demand item it is bound to. The execution progress deviation is equal to the absolute value of the difference between the current execution progress state and the expected execution progress curve at the corresponding time offset. The planned path sequence is obtained from the service robot action instruction unit corresponding to the service robot identifier. The position movement trajectory deviation is equal to the arithmetic mean of the Euclidean distances between each sampling point on the position movement trajectory curve and the corresponding point in the planned path sequence. Let the coordinates of the i-th sampling point on the position movement trajectory curve be (xi, yi), the coordinates of the i-th corresponding point in the planned path sequence be (xpi, ypi), and the total number of sampling points be M. The position movement trajectory deviation Dtraj = (1 / M)*Σ((xi-xpi)^2+(yi-ypi)^2)^(1 / 2), where i is summed from 1 to M.

[0079] Step S240: Compare the execution progress deviation with a preset execution progress deviation threshold to obtain an abnormal progress state; compare the position movement trajectory deviation with a preset position trajectory deviation threshold to obtain an abnormal trajectory state; based on the abnormal progress state and the abnormal trajectory state, comprehensively determine the task execution abnormality index of each service robot; compare the task execution abnormality index with a preset abnormality judgment condition, and filter out the abnormal service robot identifier and its associated abnormal task information where the task execution abnormality index exceeds the preset abnormality judgment condition.

[0080] Let the execution progress deviation be Dprog, and the preset execution progress deviation threshold be Throg. If Dprog > Throg, the progress anomaly status Aprog = 1; otherwise, Aprog = 0. Let the position movement trajectory deviation be Dtraj, and the preset position trajectory deviation threshold be Throg. If Dtraj > Throg, the trajectory anomaly status Atraj = 1; otherwise, Atraj = 0. Let the first anomaly weight coefficient be β1, the second anomaly weight coefficient be β2, and the task execution anomaly index Iabn = β1 * Aprog + β2 * Atraj. Let the preset anomaly judgment threshold be Thabn. If Iabn > Thabn, the service robot identifier is selected as the anomaly service robot identifier, and its associated anomaly task information is extracted.

[0081] Step S250: Extract the urgency label and intent category label of the original service request entry corresponding to the abnormal task from the abnormal task information, and determine the rescheduling urgency of the abnormal task based on the urgency label and intent category label.

[0082] Extract the urgency and intent category labels of the original service request entries from the abnormal task information. Let the numerical mapping value corresponding to the urgency label be Umap, the numerical mapping value corresponding to the intent category label be Imap, the first urgency weight coefficient be γ1, the second urgency weight coefficient be γ2, and the rescheduling urgency Uresched = γ1 * Umap + γ2 * Imap.

[0083] Step S260: Search the task resource pairing relationship diagram for all service resource entries connected to the abnormal task, exclude the service resource entries corresponding to the abnormal service robot identifier, obtain a set of alternative service resource entries, reassign the abnormal task according to the rescheduling urgency and the multidimensional adaptation association strength of the set of alternative service resource entries, generate a compensation service robot action instruction unit, add it to the service robot action instruction set, and distribute the updated service robot action instruction set to the corresponding service robot control terminal to trigger the compensation service execution operation.

[0084] In the task resource pairing relationship graph, find all service resource item nodes connected to the service requirement item node corresponding to the abnormal task. Exclude the service resource item nodes corresponding to the abnormal service robot identifier. The remaining service resource item nodes constitute the candidate service resource item set. For each candidate service resource item in the candidate service resource item set, obtain its multi-dimensional adaptation association strength S. Let the first scheduling weight coefficient be δ1 and the second scheduling weight coefficient be δ2. The rescheduling comprehensive score Sresched = δ1*Uresched + δ2*S. Select the candidate service resource item with the largest Sresched as the compensation service robot. Generate the compensation service robot action instruction unit according to step S147, add the compensation service robot action instruction unit to the service robot action instruction set, and distribute the updated service robot action instruction set to the service robot control terminal corresponding to the compensation service robot to trigger the compensation service execution operation.

[0085] Step S310: Perform type correlation analysis on the service demand type identifier of each elderly care service demand instruction unit in the elderly care service demand instruction sequence, and extract the co-occurrence frequency and temporal correlation of demand intent category tags between different elderly care service demand instruction units.

[0086] In this embodiment, all elderly care service demand instruction units in the elderly care service demand instruction sequence are statistically combined pairwise according to their corresponding demand intent category labels. A preset time window length is used as a sliding window, and the elderly care service demand instruction sequence is scanned along the time axis. Within each sliding window, combinations of different demand intent category labels that appear simultaneously are recorded. Let the total number of sliding steps of the sliding window be K. In the k-th sliding window, if demand intent category label A and demand intent category label B appear simultaneously, their co-occurrence frequency count is incremented by 1. After traversing all K sliding windows, the co-occurrence frequency C(A, B) of any two demand intent category labels is obtained. The temporal correlation is extracted as follows: for any two demand intent category labels, the number of times demand intent category label A appears before demand intent category label B and the time interval is within a preset temporal correlation window is counted, which is used as the temporal correlation strength from A to B. Through co-occurrence frequency and temporal correlation, a description of the correlation between service demand types is constructed.

[0087] Step S320: Construct a service demand type association graph based on the co-occurrence frequency and temporal correlation. Each node in the service demand type association graph is a service demand type identifier, and each edge indicates that two connected service demand type identifiers have an association triggering relationship within a preset time window. The strength of the edge is determined by the co-occurrence frequency.

[0088] In this embodiment, an undirected weighted graph is constructed as a service demand type association graph. Each node in the graph represents a service demand type identifier. For any two nodes A and B, if their co-occurrence frequency C(A, B) is greater than a preset co-occurrence frequency threshold, an edge is established between nodes A and B. The weight of this edge, Wedge(A, B), is determined by the co-occurrence frequency: Wedge(A, B) = C(A, B) / Cmax, where Cmax is the maximum value among all co-occurrence frequencies. This normalization operation maps the edge weight to the interval between 0 and 1. Simultaneously, a directed attribute is attached to each edge, recording the temporal association strength from A to B and from B to A. The service demand type association graph is stored in memory as an adjacency list. Each node entry contains a node identifier and a list of adjacent edges, and each adjacent edge contains the neighbor node identifier, edge weight, and temporal association strength in both directions.

[0089] Step S330: When a new elderly care service demand instruction unit enters the elderly care service demand instruction sequence, extract the service demand type identifier and demand trigger timestamp of the new elderly care service demand instruction unit, and search for a set of associated service demand type identifiers that have an associated trigger relationship with the service demand type identifier in the service demand type association graph.

[0090] In this embodiment, when a new elderly care service demand instruction unit is added to the elderly care service demand instruction sequence, the service demand type identifier STnew and the demand trigger timestamp TSnew of the new elderly care service demand instruction unit are extracted. In the service demand type association graph, starting from the node corresponding to STnew, all its adjacent edges are traversed. For each adjacent edge, the service demand type identifier corresponding to the neighboring node and the temporal association strength from STnew to that neighboring node are obtained. Neighboring nodes with a temporal association strength greater than a preset temporal association strength threshold are selected to form a set of associated service demand type identifiers. Each service demand type identifier in this set of associated service demand type identifiers has a significant temporal sequence association relationship with STnew.

[0091] Step S340: Generate a predicted service requirement entry based on the set of associated service requirement type identifiers and the requirement trigger timestamp, insert the predicted service requirement entry into the task execution priority sequence, and adjust the overall urgency ranking value of the predicted service requirement entry in the task execution priority sequence according to the predicted association strength parameter. The predicted service requirement entry includes a predicted service requirement type identifier, a predicted association strength parameter, and a predicted trigger time window.

[0092] In this embodiment, for each associated service demand type identifier STrel in the associated service demand type identifier set, a predicted service demand entry is generated. The predicted service demand type identifier of this predicted service demand entry is STrel. The predicted association strength parameter Pral is equal to the weighted average of the temporal association strength from STnew to STrel and the edge weights between STnew and STrel, i.e., Pral = λ1 * Wdir + λ2 * Wedge, where Wdir is the normalized value of the temporal association strength from STnew to STrel, Wedge is the edge weight, and λ1 and λ2 are preset weight coefficients, with λ1 + λ2 = 1. The start time of the prediction trigger time window is TSnew plus the minimum value of the historical average time interval from STnew to STrel, and the end time is TSnew plus the maximum value of the historical average time interval from STnew to STrel. The predicted service demand item is inserted into the task execution priority sequence. Its initial comprehensive urgency ranking value is jointly determined by the prediction correlation strength parameter Prel and the start time of the prediction trigger time window. Specifically, the comprehensive urgency ranking value is equal to the basic urgency plus Prel multiplied by the urgency adjustment coefficient. The basic urgency is determined by the time difference between the start time of the prediction trigger time window and the current system time.

[0093] Step S350: In the task resource pairing relationship diagram, a node is pre-created for the predicted service demand item and the pre-allocation multidimensional adaptation association strength between the predicted service demand item and each service resource item is calculated. A pre-occupancy mark is applied to the pre-allocation multidimensional adaptation association strength of the predicted service demand item. The pre-occupancy mark of the predicted service demand item is jointly analyzed with the load state description vector of the service resource item. The task load reservation of the corresponding service robot is adjusted to reduce the task response delay when the actual service demand is triggered.

[0094] In this embodiment, a predicted service demand entry node is created for each predicted service demand entry in the task resource pairing relationship graph. For each predicted service demand entry, its pre-allocation multidimensional adaptation association strength with all service resource entries is calculated according to steps S133 to S135. A pre-occupancy flag is applied to the calculated pre-allocation multidimensional adaptation association strength. The pre-occupancy flag is a flag indicating that the pairing relationship is a predictive allocation rather than a deterministic allocation. Let the pre-allocation multidimensional adaptation association strength of the predicted service demand entry be Spre, and the pre-occupancy flag be Flagpre. Flagpre=1 indicates that the pairing relationship has been pre-occupied. For each service resource entry, the third dimension L3, i.e., the proportion of currently assigned tasks, is extracted from its load status description vector. All pairing relationships with pre-occupancy flags are traversed, and the number of times the service resource entry has been pre-occupied, Npre, is accumulated. The adjusted task load reservation amount L3adj=L3+Npre / Nmax. In subsequent actual task scheduling, L3adj is used to replace the original L3 to calculate the current remaining load capacity, so that the service robot can reserve some load capacity in advance for predicted service demand, thereby enabling a rapid response when the predicted service demand is actually triggered.

[0095] Step S410: Perform environmental event detection processing on the elderly care environment perception information flow to identify abnormal environmental events occurring in the elderly care environment. The abnormal environmental events include an environmental abnormal event type identifier, the coordinates of the location where the environmental abnormal event occurs, and the timestamp of the environmental abnormal event.

[0096] In this embodiment, the information flow of the elderly care environment is scanned frame by frame, and a preset set of environmental anomaly detection rules is used for matching. The environmental anomaly detection rule set consists of several rule entries, each containing a trigger condition, an anomaly event type identifier, and an anomaly level. The trigger condition is defined by the sensing data type encoding and a threshold condition. For example, when the value recorded by the smoke sensor exceeds a preset smoke concentration threshold, the trigger condition is met, and the corresponding anomaly event type identifier is smoke alarm. When the millimeter-wave radar sensor detects a moving target in a restricted area and the current time period is during nighttime restricted hours, the trigger condition is met, and the corresponding anomaly event type identifier is restricted area intrusion. When the temperature value reported by the environmental temperature and humidity sensor exceeds a preset high temperature threshold for a duration exceeding a preset duration, the trigger condition is met, and the corresponding anomaly event type identifier is high temperature anomaly. When any trigger condition is met, the corresponding anomaly event type identifier is extracted, the deployment location coordinates of the sensing node associated with the sensing data record that triggered the condition are extracted as the location coordinates of the environmental anomaly event, and the current system time is extracted as the timestamp of the environmental anomaly event. These three are then encapsulated into a single environmental anomaly event.

[0097] Step S420: Determine the type of emergency service requirement associated with the environmental anomaly event based on the environmental anomaly event type identifier, determine the emergency service target area based on the location coordinates of the environmental anomaly event, and determine the emergency service trigger time based on the timestamp of the environmental anomaly event.

[0098] In this embodiment, a mapping table is pre-stored between environmental anomaly event type identifiers and emergency service request types. The row structure of this mapping table contains two entries: environmental anomaly event type identifier and emergency service request type. A complete match query is performed using the environmental anomaly event type identifier as the search key to retrieve the corresponding emergency service request type. For example, if the environmental anomaly event type identifier is a smoke alarm, the emergency service request type is emergency fire extinguishing; if the environmental anomaly event type identifier is a restricted area intrusion, the emergency service request type is security patrol intervention; if the environmental anomaly event type identifier is a high temperature anomaly, the emergency service request type is environmental temperature control. The location coordinates of the environmental anomaly event are used to determine the grid area number into which it falls using a point-to-surface inclusion determination method. This grid area number, along with a circular coverage area centered on the location coordinates and with a preset emergency radiation radius, is used as the emergency service target area. The timestamp of the environmental anomaly event is used as the emergency service trigger time.

[0099] Step S430: Generate an emergency elderly care service demand instruction unit, use the emergency service demand type as the service demand type identifier of the emergency elderly care service demand instruction unit, use the emergency service trigger time as the service demand trigger timestamp of the emergency elderly care service demand instruction unit, and insert the emergency elderly care service demand instruction unit at the top of the elderly care service demand instruction sequence.

[0100] In this embodiment, an emergency elderly care service request instruction unit is constructed. The service request type identifier field of this unit is assigned the value of an emergency service request type, and the service request trigger timestamp field is assigned the value of the emergency service trigger time. This emergency elderly care service request instruction unit is inserted at the head of the elderly care service request instruction sequence, making it the first elderly care service request instruction unit to be processed in the sequence. Simultaneously, the urgency label of this emergency elderly care service request instruction unit is directly set to a label indicating urgency.

[0101] Step S440: In the task execution priority sequence, the comprehensive urgency ranking value corresponding to the emergency elderly care service demand instruction unit is set to the highest priority, and the task resource pairing relationship with a priority lower than that of the emergency elderly care service demand instruction unit in the currently executing task resource pairing relationship diagram is interrupted.

[0102] In this embodiment, within the task execution priority sequence, a preset maximum overall urgency ranking value is assigned to the service request item corresponding to the emergency elderly care service request instruction unit. This maximum overall urgency ranking value is greater than the overall urgency ranking values ​​of all other service request items. All established task resource pairing relationships in the task resource pairing relationship graph are traversed, and it is checked whether the overall urgency ranking value of the service request item corresponding to each pairing relationship is less than the maximum overall urgency ranking value. If so, the pairing relationship is marked as interrupted, and the pairing information before the interruption is recorded, including the service request item identifier, the service resource item identifier, and the edge weight value.

[0103] Step S450: Release the occupied service resource entries from the interrupted task resource pairing relationship, add the released service resource entries back to the task resource pairing relationship graph, and re-execute the task response association construction process for the emergency elderly care service demand instruction unit to generate an emergency task resource pairing relationship.

[0104] In this embodiment, all service resource entry nodes involved in pairing relationships marked as interrupted are released. The specific release process involves: deleting the edges corresponding to the interrupted pairing relationships from the task resource pairing relationship graph; restoring the load state description vector of the service resource entry node to its state before the pairing relationship was established, i.e., restoring the third dimension value to Ntask / Nmax instead of (Ntask+1) / Nmax, and correspondingly reverting the fourth dimension value. The released service resource entry nodes reappear as available nodes in the task resource pairing relationship graph. For the emergency elderly care service demand instruction unit, the task response association construction process is re-executed according to steps S131 to S135. The multi-dimensional adaptation association strength between the service demand entry corresponding to the emergency elderly care service demand instruction unit and all available service resource entries is calculated. Edges are established between the emergency elderly care service demand instruction unit node and each service resource entry node in the task resource pairing relationship graph, generating emergency task resource pairing relationships.

[0105] Step S460: Based on the emergency task resource pairing relationship and the highest priority marker in the task execution priority sequence, perform emergency scheduling processing on the emergency elderly care service demand instruction unit to generate an emergency service robot action instruction unit. Add the emergency service robot action instruction unit to the service robot action instruction set. Distribute the updated service robot action instruction set to the corresponding service robot control terminal to trigger the emergency service execution operation. After the emergency service execution operation is completed, restore the interrupted task resource pairing relationship to the state before the interruption and re-perform multi-task collaborative scheduling processing.

[0106] In this embodiment, the service resource item with the strongest multi-dimensional adaptation association strength is selected as the emergency assigned service robot from the emergency task resource pairing relationship. Following step S147, the identifier of the emergency assigned service robot is used as the service robot identifier, the center coordinates of the emergency service target area are used as the target service object location, the emergency service demand type is mapped to a service action type through a preset service action type mapping table, and the current system time plus the estimated travel time is used as the service action execution sequence. These are combined to generate an emergency service robot action instruction unit. This emergency service robot action instruction unit is inserted at the head of the service robot action instruction set. The updated service robot action instruction set is distributed to the service robot control terminal corresponding to the emergency assigned service robot, triggering the emergency service execution operation. After the emergency service execution operation is completed, task completion feedback information reported by the emergency assigned service robot control terminal is received. Based on the pairing information recorded in step S440 before the interruption, the interrupted pairing relationships are restored one by one to the task resource pairing relationship diagram, and the multi-task collaborative scheduling processing of steps S141 to S148 is re-executed.

[0107] Step S510: Perform spatial conflict analysis on the service robot action instruction set, extract the target service object position and service action execution sequence from all service robot action instruction units, and detect the degree of spatial position overlap between different service robots in the same time segment.

[0108] In this embodiment, for all service robot action instruction units in the service robot action instruction set, the target service object position and service action execution sequence are extracted. The time axis is divided into continuous time segments according to a preset time slice granularity. For each time segment, all service robot action instruction units in the execution state within that time segment are counted. For any two service robot action instruction units within the same time segment, let their target service object positions be P1 and P2, and the estimated movement path of the service robot within that time segment be R1 and R2, respectively. The degree of spatial position overlap is determined by calculating the minimum distance between the two position intervals. Let the geometric center of R1 be C1 and the geometric center of R2 be C2, and the minimum distance Dmin_path = ((x_C1-x_C2)^2+(y_C1-y_C2)^2)^(1 / 2)-(r1+r2), where r1 and r2 are the radii of the envelope circles of R1 and R2, respectively. The degree of spatial position overlap is Overlap = 1 / (1+Dmin_path), and the larger this value, the higher the degree of overlap.

[0109] Step S520: When service robot action command units with spatial overlap exceeding a preset conflict threshold are detected in the same time segment, the conflicting service robot set and conflicting time segment are identified.

[0110] In this embodiment, a preset conflict threshold is set as Thoverlap. For each time segment, if the spatial overlap of any two service robot action command units exceeds Thoverlap, then the time segment is marked as a conflict time segment. The service robot identifiers corresponding to all service robot action command units with spatial overlap exceeding Thoverlap within the conflict time segment are aggregated, deduplicated, and then a conflicting service robot set is formed.

[0111] Step S530: For each conflict service robot in the conflict service robot set, extract its target service object location and current location area, calculate the service path trajectory of the conflict service robot, and generate a service path trajectory description containing a sequence of path nodes.

[0112] In this embodiment, for each conflict service robot in the conflict service robot set, the center coordinates corresponding to the location of its target service object and its current location area are extracted. Using the center coordinates of the current location area as the path start point and the location of the target service object as the path end point, the A* path search algorithm is run on a grid map of the traversable area of ​​the elderly care environment to generate a service path trajectory consisting of a continuous sequence of grid area numbers. During the path search process, the movement cost between grid nodes is determined by the inter-grid distance and the path traversability coefficient. The generated service path trajectory description includes a path node sequence, where each element is a grid area number and its corresponding estimated arrival time.

[0113] Step S540: Perform path intersection analysis on the service path trajectory descriptions of every two conflicting service robots in the conflicting service robot set, and calculate the location coordinates and intersection time points of the service path trajectory intersection points.

[0114] In this embodiment, for every two conflicting service robots in the conflicting service robot set, their service path trajectory descriptions are extracted. The path node sequences of the two service path trajectory descriptions are expanded chronologically, and it is checked whether they have the same grid area number. If the same grid area number exists, the estimated arrival time difference between the two conflicting service robots reaching the same grid area number is further checked. If the absolute value of the time difference is less than a preset time difference threshold, the center coordinates of the same grid area number are used as the intersection coordinates of the service path trajectories, and the average of the two estimated arrival times is used as the intersection time point.

[0115] Step S550: Based on the coordinates of the intersection point of the service path trajectory and the intersection time point, the timing of the service actions of the conflicting service robots is adjusted by delaying the execution timing of the service actions of one of the conflicting service robots by a conflict avoidance time interval, and generating the timing-adjusted conflicting service robot action instruction unit.

[0116] In this embodiment, for two conflicting service robots that intersect, the priority passage order is determined based on the urgency labels of their corresponding service request entries. The urgency label indicates that the higher-priority conflicting service robot has priority. The execution sequence of the service actions of the conflicting service robot with the lower priority passage order is delayed by a conflict avoidance time interval. The conflict avoidance time interval is equal to the sum of the estimated time required for the two conflicting service robots to pass through the intersection plus a preset safety time margin. The execution sequence of the service actions of the conflicting service robot is modified to the original sequence plus the conflict avoidance time interval, generating a conflicting service robot action instruction unit with adjusted timing.

[0117] Step S560: During the conflict time segment, perform path replanning processing on the conflict service robots in the conflict service robot set, generate alternative service path trajectories to avoid intersections based on the passable area information in the elderly care environment, and generate spatially adjusted conflict service robot action command units.

[0118] In this embodiment, for conflict service robots requiring path replanning, the grid area numbers corresponding to the intersection points are marked as temporary obstacle nodes on the passable area grid map of the elderly care environment. The A* path search algorithm is re-run, using the current location center coordinates of the conflict service robot as the path start point and the target service object location as the path end point, generating an alternative service path trajectory on the grid map excluding temporary obstacle nodes. If an alternative service path trajectory exists and its total path length increment does not exceed a preset maximum detour length ratio, the alternative service path trajectory is used as the spatially adjusted service path trajectory. If an alternative service path trajectory does not exist or the total path length increment exceeds a preset maximum detour length ratio, the process reverts to using only the timing adjustment method in step S550. The target service object location and service action execution timing corresponding to the spatially adjusted service path trajectory are updated to the original service robot action command unit, generating a spatially adjusted conflict service robot action command unit.

[0119] Step S570: Replace the original service robot action instruction unit in the service robot action instruction set with the timing-adjusted conflict service robot action instruction unit and the spatially adjusted conflict service robot action instruction unit to obtain the conflict-resolved service robot action instruction set.

[0120] In this embodiment, the original service robot action instruction unit is located in the service robot action instruction set by using the conflicting service robot identifier and the original service action execution sequence as search criteria. The original service robot action instruction unit is then replaced with the conflicting service robot action instruction unit after timing or spatial adjustment. After the replacement is completed, the service robot action instruction set after conflict resolution is obtained.

[0121] Step S610: Obtain a set of historical service execution records of the elderly care service robot. The set of historical service execution records contains multiple historical service execution records. Each historical service execution record contains a historical service demand type identifier, a historical service object location area, a historical service execution start time, a historical service execution duration, and a historical service robot identifier.

[0122] In this embodiment, all historical service execution records within a preset historical time range are read from the persistent storage of the data processing server. Each historical service execution record corresponds to a service request entry that has been completed. The record fields include a historical service request type identifier, a historical service object location region, a historical service execution start time, a historical service execution duration, and a historical service robot identifier. The historical service object location region is the grid region number where the service object was located when the service request was triggered. The historical service execution start time is the timestamp of the timestamp when the service robot actually started executing the service operation. The historical service execution duration is the total duration from the start of the service to its completion. The historical service robot identifier is a unique identifier for the service robot that executed the service.

[0123] Step S620: Perform service pattern mining processing on the historical service execution record set, extract the frequency distribution of various service demand type identifiers in different time periods based on the historical service demand type identifier and the historical service execution start time, and generate a periodic distribution pattern of service demand.

[0124] In this embodiment, all records in the historical service execution record set are expanded along the time dimension. Multiple time period granularities are preset, including hourly, daily, and weekly periods. For each time period granularity, the frequency of occurrence of each historical service demand type identifier in each time slot within that period is counted. For example, in the hourly period, the number of triggers of each historical service demand type identifier within each hourly time period is counted. In the daily period, the average number of triggers of each historical service demand type identifier within each hour of the day is counted. In the weekly period, the number of triggers of each historical service demand type identifier per day of the week is counted. The above multi-period statistical results are summarized into a service demand periodic distribution pattern, which is a three-dimensional frequency array, with the three dimensions being the time period type, the time slot index, and the service demand type identifier, respectively.

[0125] Step S630: Extract the association distribution between different location areas and different service demand type identifiers based on the historical service object location area and historical service demand type identifier, and generate a service demand spatial distribution pattern.

[0126] In this embodiment, for each record in the historical service execution record set, a location type pair is formed by extracting the historical service object location region and the historical service demand type identifier. The frequency of occurrence of each location type pair is counted to construct a two-dimensional frequency matrix. The rows of the matrix correspond to grid region numbers, the columns correspond to historical service demand type identifiers, and the matrix elements represent the total number of times that historical service demand type identifier is triggered within that grid region number. This two-dimensional frequency matrix is ​​then normalized row-wise so that the sum of the elements in each row is 1, resulting in a service demand spatial distribution pattern. Each element in this service demand spatial distribution pattern represents a conditional probability estimate of triggering a specific historical service demand type identifier under a given grid region number.

[0127] Step S640: Extract the execution duration distribution features of different service requirement type identifiers based on the historical service execution duration and historical service requirement type identifiers, and generate a service execution duration reference range.

[0128] In this embodiment, the historical service execution record set is grouped according to the historical service demand type identifier. For each group, the execution duration of all historical services is extracted, and the mean μdur and standard deviation σdur of the execution duration corresponding to that historical service demand type identifier are calculated. The service execution duration reference interval is defined as [μdur-z*σdur, μdur+z*σdur], where z is a preset standard deviation multiple parameter. For historical service demand type identifiers with a small number of historical service execution records that are insufficient to support statistical estimation, a preset default execution duration reference interval is used.

[0129] Step S650: Based on the periodic distribution pattern of service demand, the spatial distribution pattern of service demand, and the reference interval of service execution time, the probability of service demand occurrence in each time period and each location area within a preset time period is estimated, and a service demand prediction heatmap is generated.

[0130] In this embodiment, a start and end time for a future preset time period is set. This future preset time period is divided into several prediction time slots according to the time period granularity. For each prediction time slot, the historical frequency distribution of each service demand type identifier corresponding to the time slot is extracted from the periodic distribution pattern of service demand. For each grid area number, the conditional probability distribution of each service demand type identifier corresponding to the grid area number is extracted from the spatial distribution pattern of service demand. The estimated value of the probability of service demand occurrence is obtained by multiplying the periodic frequency distribution and the spatial conditional probability distribution element-wise and then normalizing. The estimated value of the probability of service demand occurrence for each grid area number is bound to the center coordinate of the grid area number to generate a service demand prediction heatmap. In the service demand prediction heatmap, each grid area number corresponds to a probability value, which represents the comprehensive probability of any type of service demand occurring in the grid area number within the future preset time period.

[0131] Step S660: Based on the service demand prediction heatmap, adjust the standby location distribution of service robots in advance at the current moment, dispatch idle service robots to areas with high predicted service demand to stay and wait, generate a set of service robot pre-deployment instructions, and distribute them to the corresponding service robot control terminal to trigger the pre-deployment movement operation.

[0132] In this embodiment, the task status of all service robots is monitored in real time, and service robots that are currently idle and not pre-occupied are selected to form an idle service robot list. In the service demand prediction heatmap, grid area numbers are sorted from high to low probability values, and the top K grid area numbers with the highest probability values ​​are selected as the predicted high-demand areas, where K is the number of idle service robots. Each idle service robot in the idle service robot list is matched one-to-one with the predicted high-demand areas, with the matching principle being the minimum total estimated movement distance. For each matching pair, a service robot pre-deployment instruction unit is generated, which includes the idle service robot identifier, the center coordinates of the predicted high-demand area, and the pre-deployment action type. All service robot pre-deployment instruction units are aggregated into a service robot pre-deployment instruction set and distributed to the corresponding service robot control terminal, triggering the idle service robots to move in advance to the predicted high-demand areas to wait for service.

[0133] For example, step S710: Obtain a set of service robot energy consumption status information during the collaborative process of elderly care service tasks. The set of service robot energy consumption status information includes the current remaining energy value and energy consumption rate per unit distance of movement for each service robot.

[0134] In this embodiment, each service robot control terminal reports energy consumption status information at fixed intervals. This information includes the service robot identifier, the current remaining energy value, and the energy consumption rate per unit distance traveled. The current remaining energy value is provided by the service robot's battery management unit and is measured in watt-hours (Wh). The energy consumption rate per unit distance traveled is calculated by the service robot's chassis drive unit based on historical energy consumption data and is measured in Wh per meter. All reported energy consumption status information is aggregated into a service robot energy consumption status information set.

[0135] Step S720: When performing multi-task collaborative scheduling on the task resource pairing relationship graph, for each service requirement item to be scheduled, extract the current remaining energy value and energy consumption rate per unit movement distance of each available service robot in the list of available service robots that match the service requirement item.

[0136] In this embodiment, after the list of available service robots is generated in step S143, for each available service robot in the list of available service robots, the current remaining energy value Eremain and energy consumption rate per unit moving distance rconsume are searched from the energy consumption state information set of service robots.

[0137] Step S730: calculating the estimated energy consumption of each available service robot moving from the current position area to the position of the service object corresponding to the service demand entry, comparing the estimated energy consumption with the current remaining energy value, and determining the estimated remaining energy value of the service robot after completing the service demand entry.

[0138] In this embodiment, for each available service robot in the list of available service robots, let the estimated moving distance from the current position area of the available service robot to the position of the service object corresponding to the service demand entry be Dmove, then the estimated energy consumption Ecost=Dmove*rconsume. Let the current remaining energy value of the available service robot be Eremain, then the estimated remaining energy value Eremain_after=Eremain-Ecost after the service robot completes the service demand entry. The static energy consumption during service execution is also considered: let the upper limit of the reference interval of service execution duration corresponding to the service demand entry be Tmax, and the static energy consumption per unit time be rstatic, then the total estimated energy consumption Etotal=Ecost+Tmax*rstatic, and Eremain_after=Eremain-Etotal.

[0139] Step S740: comparing the estimated remaining energy value with a preset energy safety threshold, filtering out available service robots whose estimated remaining energy value is lower than the preset energy safety threshold, and generating an energy-compliant service robot list.

[0140] In this embodiment, let the preset energy safety threshold be Esafe. For each available service robot in the list of available service robots, if Eremain_after<Esafe, the available service robot is filtered out; if Eremain_after≥Esafe, the available service robot is retained. All retained available service robots constitute the energy-compliant service robot list.

[0141] Step S750: selecting the energy-compliant service robot with the highest multi-dimensional adaptive association strength and the minimum estimated energy consumption from the energy-compliant service robot list as the energy-saving assigned service robot, and binding the service demand entry with the energy-saving assigned service robot for task assignment.

[0142] In this embodiment, for each energy-compliant service robot in the energy-compliant service robot list, its multi-dimensional adaptation correlation strength S and total estimated energy consumption Etotal are obtained. An energy-saving assignment score Senergy = α*S + β*(1-Etotal / Emax) is constructed, where α and β are preset weighting coefficients and α+β=1, and Emax is the rated total battery energy. This formula normalizes the energy consumption and converts it into an energy-saving tendency before weighting and fusing it with the adaptation strength. The energy-compliant service robot with the largest Senergy is selected as the energy-saving assignment service robot, and service request items are task-bound to this energy-saving assignment service robot.

[0143] Step S760: After completing the task binding of all service requirement items, perform energy consumption balance analysis on the task binding sequence of each service robot, calculate the cumulative estimated energy consumption of each service robot, detect the balance of cumulative estimated energy consumption among the service robots, and when the balance of cumulative estimated energy consumption among the service robots exceeds the preset imbalance threshold, transfer the service requirement item with the lowest multidimensional adaptation association strength in the task binding sequence of the high-energy-consuming service robot to the task binding sequence of the low-energy-consuming service robot to balance the overall energy consumption of each service robot.

[0144] In this embodiment, after all service request items have completed initial task binding, for each service robot, the total estimated energy consumption of all service request items in its task binding sequence is accumulated to obtain the cumulative estimated energy consumption Eacc of that service robot. The mean μE and standard deviation σE of Eacc for all service robots are calculated. The degree of energy balance is measured by the coefficient of variation CV = σE / μE. If CV is greater than the preset imbalance threshold CVth, energy balance adjustment is initiated. The service robot with the largest Eacc is selected as the high-energy-consuming service robot, and the service request item with the lowest multidimensional adaptation association strength S is selected from its task binding sequence as the item to be transferred. The service robot with the smallest Eacc is selected as the low-energy-consuming service robot. The item to be transferred is removed from the task binding sequence of the high-energy-consuming service robot and added to the task binding sequence of the low-energy-consuming service robot. The cumulative estimated energy consumption of the two service robots is updated, and CV is recalculated. The above balance adjustment process is repeated until CV ≤ CVth or the preset maximum adjustment round is reached.

[0145] Step S770: Update the service robot action instruction set according to the task binding sequence after energy consumption balancing adjustment, and distribute the updated service robot action instruction set to the corresponding service robot control terminal to trigger service execution operation.

[0146] In this embodiment, based on the task binding sequence adjusted for energy balance, the service robot action instruction unit for each service robot is regenerated, and the service robot action instruction set is updated. The updated service robot action instruction set is then distributed to the corresponding service robot control terminal to trigger service execution operations.

[0147] Step S810: Perform demand intent association parsing on the elderly care service demand instruction sequence, extract the time interval of the service demand trigger timestamps between different elderly care service demand instruction units, construct a service demand co-occurrence frequency distribution in units of time windows, identify service demand type identifier combinations with time-series correlation based on the service demand co-occurrence frequency distribution, and mark service demand type identifier combinations that appear more frequently than a preset co-occurrence frequency threshold within the same time window as service demand association pairs.

[0148] In this embodiment, the service demand type identifier and demand trigger time node of each elderly care service demand instruction unit are extracted from the elderly care service demand instruction sequence. A preset time window length is used as the statistical window, sliding along the time axis with a preset step size. Within each statistical window, all service demand type identifiers appearing in that window are recorded, and the number of times any two service demand type identifiers appear simultaneously within that window is counted. After traversing all statistical windows, a two-dimensional co-occurrence frequency matrix is ​​constructed. The rows and columns of the matrix correspond to the values ​​of the service demand type identifiers, and the matrix elements are the cumulative co-occurrence counts of the corresponding service demand type identifiers in all statistical windows. Let Fco(A, B) be the cumulative co-occurrence count of service demand type identifier A and service demand type identifier B in the co-occurrence frequency matrix, and let Fth be the preset co-occurrence frequency threshold. If Fco(A, B) is greater than Fth, then the combination of service demand type identifier A and service demand type identifier B is marked as a service demand association pair. Simultaneously, the time interval characteristics of the service demand association pair are calculated, that is, the demand triggering time nodes corresponding to A and B are extracted from all statistical windows containing A and B, and the average value of the absolute value of the difference between the demand triggering time node of A and the demand triggering time node of B is calculated as the average time interval of the service demand association pair.

[0149] Step S820: Perform service action coordination analysis on the two service request type identifiers in each service request association pair, extract the service robot motion trajectory and service operation area involved in the service action type corresponding to each service request type identifier, and calculate the degree of spatial overlap and temporal parallelism of the two service action types.

[0150] In this embodiment, for each service request type identifier in a service request pair, the corresponding service action type is obtained through a preset service action type mapping table. For each service action type, all historical execution records of that service action type are extracted from the historical service execution record set. Based on the location trajectory data in the historical execution records, a typical motion trajectory envelope and a typical operation area for that service action type are generated. The typical motion trajectory envelope is the convex hull polygon of the historical trajectory point set, and the typical operation area is the density estimation area of ​​the historical operation dwell points. The degree of spatial overlap is obtained by calculating the ratio of the intersection area to the union area of ​​the two typical operation areas. Let the area of ​​the typical operation area of ​​service action type A be AreaA, the area of ​​the typical operation area of ​​service action type B be AreaB, the intersection area be AreaAB_intersect, and the union area be AreaAB_union. The degree of spatial overlap is Soverlap = AreaAB_intersect / AreaAB_union. The degree of parallelism in time sequence is determined by analyzing the overlapping execution ratio of two service action types on the time axis. Let the average execution time of the two service action types be TA and TB, respectively, and the length of the time segment that can be executed in parallel be Tparallel. Then the degree of parallelism Pparallel = Tparallel / max(TA, TB), where the length of the time segment that can be executed in parallel is obtained by analyzing the maximum overlap time of the two service action types without causing spatial resource conflicts during execution.

[0151] Step S830: Determine the collaborative service feasibility level of the service request association pair based on the degree of spatial overlap and the degree of temporal parallelism, and mark the service request association pair whose collaborative service feasibility level meets the preset collaborative conditions as a collaboratively executable task combination.

[0152] In this embodiment, for each service request pair, the collaborative service feasibility level Gradeco is determined by a combination of spatial overlap and parallelism: Gradeco = w1 * (1 - Soverlap) + w2 * Pparallel, where w1 and w2 are preset weight coefficients and w1 + w2 = 1, and 1 - Soverlap represents the degree of spatial non-overlap; the less spatial overlap, the higher the collaborative feasibility. The preset collaboration condition is that Gradeco is greater than the preset collaborative feasibility threshold Gth. If Gradeco is greater than Gth, the service request pair is marked as a collaboratively executable task combination.

[0153] Step S840: When multiple elderly care service demand instruction units belonging to the same collaboratively executable task group exist simultaneously in the elderly care service demand instruction sequence, the multiple elderly care service demand instruction units are extracted from the task execution priority sequence to construct a collaborative task execution group. The collaborative task execution group includes the service demand type identifier, service object location characteristics, and service demand trigger timestamp of all elderly care service demand instruction units in the group.

[0154] In this embodiment, the sequence of elderly care service demand instructions is monitored in real time. When multiple elderly care service demand instruction units belonging to the same collaboratively executable task group are detected in the sequence, these multiple elderly care service demand instruction units are extracted from the task execution priority sequence all at once. A collaborative task execution group is constructed based on the extracted set of elderly care service demand instruction units. This collaborative task execution group includes the service demand type identifier, the corresponding service object location feature, and the service demand trigger timestamp for each elderly care service demand instruction unit in the group. The service object location feature is the current location area grid number of the service object in the service object status feature set.

[0155] Step S850: In the task resource pairing relationship diagram, the collaborative task execution group is treated as a whole task unit and paired with all service resource items for collaborative task resource pairing. The overall multidimensional adaptation association strength of each service resource item for the collaborative task execution group is calculated. The overall multidimensional adaptation association strength is generated by weighted fusion of the independent multidimensional adaptation association strength of each elderly care service demand instruction unit in the group and the collaborative service feasibility level.

[0156] In this embodiment, the collaborative task execution group is added as a whole task unit node to the task resource pairing relationship graph. For each service resource item, the independent multidimensional adaptation association strength between the service resource item and the service demand item corresponding to each elderly care service demand instruction unit in the collaborative task execution group is calculated according to steps S133 to S135. Let there be M elderly care service demand instruction units in the group, and the independent multidimensional adaptation association strength corresponding to the i-th elderly care service demand instruction unit is Si. The overall multidimensional adaptation association strength Sgroup = λ1*(1 / M)*ΣSi + λ2*Gradeco, where λ1 and λ2 are preset weight coefficients and λ1 + λ2 = 1, ΣSi represents the summation of all independent multidimensional adaptation association strengths in the group, and Gradeco is the collaborative service feasibility level of the collaboratively executable task combination.

[0157] Step S860: Select the optimal collaborative service robot for the collaborative task execution group based on the overall multidimensional adaptation association strength, and generate a collaborative service robot action instruction unit combination. The collaborative service robot action instruction unit combination includes service robot action instruction units for each elderly care service demand instruction unit in the group, and the execution sequence of service actions of each service robot action instruction unit in the group is synchronously arranged according to the degree of parallelism.

[0158] In this embodiment, the service resource item with the highest overall multidimensional adaptation correlation strength (Sgroup) among all service resource items is selected as the optimal collaborative service robot. For each elderly care service demand instruction unit within the collaborative task execution group, a service robot action instruction unit is generated for this optimal collaborative service robot, using the same generation method as in step S147. The execution sequence of the service actions of each service robot action instruction unit within the group is synchronously arranged according to the parallelism level (Pparallel). Specifically, the arrangement method is as follows: the execution start time of the two service action types with the highest parallelism level is set to be the same, and the execution start times of the remaining service action types are arranged closely on the time axis in descending order of parallelism level. If there is no spatial overlap between two service action types, their execution time windows are completely overlapped. If there is partial spatial overlap, their execution time windows are staggered by the time interval required for the overlap.

[0159] Step S870: Add the collaborative service robot action instruction unit combination to the service robot action instruction set, remove all elderly care service demand instruction units corresponding to the collaborative task execution group from the task execution priority sequence, and update the load state description vector of the optimal collaborative service robot.

[0160] In this embodiment, all service robot action instruction units in the generated collaborative service robot action instruction unit combination are added to the service robot action instruction set. Service requirement entries corresponding to all elderly care service requirement instruction units contained in the collaborative task execution group are removed from the task execution priority sequence. The load state description vector of the optimal collaborative service robot is updated. Let the original number of assigned tasks be Ntask, the number of tasks in the collaborative task execution group be Mgroup, and the maximum concurrent task capacity be Nmax. After the update, the third dimension is (Ntask + Mgroup) / Nmax, and the fourth dimension is recalculated based on the newly added task type list.

[0161] Step S910: During the process of multi-task collaborative scheduling based on the task resource pairing relationship diagram and the task execution priority sequence, global service path optimization is performed on all service robots currently in task-bound state.

[0162] In this embodiment, during the multi-task collaborative scheduling process in steps S141 to S148, after each round of complete task assignment and generation of a set of service robot action instructions, a global optimization process for the service paths of all service robots currently in a task-bound state is triggered.

[0163] Step S920: Extract the target service object location and current location region of each service robot in the task-bound state, and generate the initial service path sequence of the service robot. The initial service path sequence contains ordered path nodes from the current location region to each target service object location in sequence.

[0164] In this embodiment, for each service robot in a task-bound state, the target service object locations of all service request entries in its current task-bound sequence are extracted and arranged in ascending order according to the service action execution sequence. The center coordinates of the service robot's current location area are used as the path start point, the first target service object location as the first intermediate node, the second target service object location as the second intermediate node, and so on, with the last target service object location as the path end point. Between each adjacent node pair, the A* path search algorithm is run on the traversable area grid map to generate a path node sequence between nodes. The path node sequences between all adjacent node pairs are concatenated end-to-end to obtain the initial service path sequence for the service robot.

[0165] Step S930: Perform cross-analysis on the initial service path sequences of all service robots in the task-bound state to identify overlapping path segments and path intersection nodes between the initial service path sequences of different service robots within the same time period.

[0166] In this embodiment, the initial service path sequences of all service robots are unfolded in a unified time-space coordinate system. For each service robot's initial service path sequence, an estimated arrival time is added to each path node according to the execution sequence of its service actions. For any two different service robot's initial service path sequences, it is checked along the time axis whether there is a grid area number that is the same, and the time difference between the estimated arrival times of the two service robots to the same grid area number is less than a preset path conflict time threshold. If the above condition is met, the path segment corresponding to the same grid area number is marked as a path overlap segment. If a node in the path node sequence of one service robot and a node in the path node sequence of another service robot share the same grid area number, and the time difference between the estimated arrival times of the two service robots to that node is less than the preset path conflict time threshold, but their path directions after that node are different, then that node is marked as a path intersection node.

[0167] Step S940: For each overlapping path segment and intersection node, calculate the estimated arrival time of each service robot on the overlapping path segment or intersection node. When the time difference between the estimated arrival times of different service robots is less than a preset time conflict threshold, mark the overlapping path segment or intersection node as a path conflict area. For each path conflict area, extract the service demand intention category label and demand urgency label of all service robots that will pass through the path conflict area. Determine the priority passage order of each service robot through the path conflict area based on the demand urgency label.

[0168] In this embodiment, for each overlapping path segment or intersection node, all service robots that will pass through that location and their estimated arrival times are obtained. If the absolute value of the difference between the estimated arrival times of any two service robots is less than a preset time conflict threshold, the overlapping path segment or intersection node is marked as a path conflict area. For this path conflict area, the urgency level tag of the service request entry bound to each service robot that will pass through is extracted. The priority passage order is determined according to the value of the urgency level tag, with tags indicating urgency taking precedence over tags indicating priority, and tags indicating priority taking precedence over tags indicating normal. For service robots with the same urgency level tag, their request trigger time nodes are compared, and those with earlier trigger times pass first.

[0169] Step S950: According to the priority passage order, the initial service path sequence of each service robot is adjusted segment by segment. For service robots with a lower priority passage order, a deceleration and waiting instruction is inserted at the path node before reaching the path conflict area to generate the service path sequence after speed adjustment. The path deviation feasibility analysis is performed on the service path sequence after speed adjustment, the passable area information in the elderly care environment perception information flow is detected, and it is determined whether there are alternative path branches that can bypass the path conflict area.

[0170] In this embodiment, for path conflict areas, the initial service path sequence of the service robot with the highest priority remains unchanged according to the priority passage order. For service robots with lower priority passage orders, a deceleration and waiting instruction is inserted at the path node before the path conflict area. The deceleration and waiting instruction delays the arrival time of the service robot in the path conflict area by a waiting time, which is equal to the estimated time required for the service robot with a higher priority passage order to pass through the path conflict area plus a preset safety time margin. The path sequence after inserting the deceleration and waiting instruction is used as the speed-adjusted service path sequence. A path offset feasibility analysis is performed on the speed-adjusted service path sequence: on the passable area grid map, with the path conflict area as the center, it is searched for whether there is an alternative path branch that bypasses the path conflict area from a path node before the path conflict area to a path node after the path conflict area. The alternative path branch must satisfy the requirement that the increase in the total path length does not exceed the preset maximum detour ratio of the original path conflict area segment length, and does not introduce new path conflicts.

[0171] Step S960: When the alternative path branch exists, the service path sequence of the service robot with the later priority passage order is switched from the original path conflict area to the alternative path branch to generate the service path sequence after detour adjustment. When the alternative path branch does not exist, the service path sequence after speed adjustment is retained.

[0172] In this embodiment, if an alternative path branch that meets the constraints is found in step S950, the path segment corresponding to the original path conflict area in the service path sequence of the service robot with a lower priority passage order is replaced with the alternative path branch, the corresponding path nodes and their expected arrival times are updated, and a service path sequence after detour adjustment is generated. If no alternative path branch exists, the service path sequence after speed adjustment is retained as the final adjustment result.

[0173] Step S970: Bind the service path sequence of each service robot after path conflict resolution to the corresponding service robot action instruction unit, and update the service action execution sequence and the path planning description corresponding to the service action type in each service robot action instruction unit.

[0174] In this embodiment, the service path sequence of each service robot, after speed adjustment or detour adjustment, is bound to the service robot action instruction unit with the corresponding service robot identifier. The service action execution timing in each service robot action instruction unit is updated to the adjusted estimated arrival time, and the path planning description corresponding to the service action type is updated to the adjusted service path sequence.

[0175] Step S1010: Perform service demand intent evolution analysis on the elderly care service demand instruction sequence, extract multiple elderly care service demand instruction units generated by the same service object within a continuous time period, and construct a service object demand evolution sequence. The service object demand evolution sequence includes service demand type identifiers arranged according to the service demand trigger timestamp.

[0176] In this embodiment, the sequence of elderly care service demand instructions is grouped according to the service recipient identifier. For each group of service recipient identifiers, all elderly care service demand instruction units generated within a preset continuous time period are extracted and sorted in ascending order by the service demand trigger timestamp. The service demand type identifier of each sorted elderly care service demand instruction unit is extracted sequentially to form an ordered list of service demand type identifiers. This ordered list is the service recipient demand evolution sequence for that service recipient.

[0177] Step S1020: Perform demand state transition pattern mining on the service object demand evolution sequence, count the transition frequency of sequential transitions between different service demand type identifiers, calculate the transition probability distribution of each service demand type identifier to other service demand type identifiers, and construct a service object demand state transition topology graph based on the transition probability distribution. In the service object demand state transition topology graph, each node represents a service demand type identifier, and each directed edge represents the transition relationship from the source service demand type identifier to the target service demand type identifier. The weight of the directed edge is the corresponding transition probability.

[0178] In this embodiment, the service object demand evolution sequences of all service objects are aggregated into a transition frequency statistics set. For each service object demand evolution sequence, for each pair of adjacent service demand type identifiers (A, B), where A is the predecessor service demand type identifier and B is the successor service demand type identifier, the transition frequency count from A to B is incremented by 1. After the statistics are completed, for each service demand type identifier A, the transition probability from A to each other service demand type identifier B is calculated. The transition probability is equal to the transition frequency from A to B divided by the total frequency of transitions from A to all service demand type identifiers. A directed weighted graph is constructed as the service object demand state transition topology graph. Each node in the graph corresponds to a service demand type identifier. If the transition probability from A to B is greater than 0, a directed edge from A to B is added, with the edge weight being the corresponding transition probability.

[0179] Step S1030: When a new elderly care service demand instruction unit enters the elderly care service demand instruction sequence, extract the service object identifier and service demand type identifier corresponding to the new elderly care service demand instruction unit, and match the personalized demand state transition path corresponding to the service object identifier in the service object demand state transition topology map.

[0180] In this embodiment, when a new elderly care service demand instruction unit enters the elderly care service demand instruction sequence, its service object identifier OIDnew and service demand type identifier STnew are extracted. In the service object demand state transition topology diagram, the node corresponding to STnew is taken as the starting point. Based on the historical data of the service object demand evolution sequence of OIDnew, the personalized demand state transition path of OIDnew is extracted from the general service object demand state transition topology diagram. That is, the frequency distribution of transition records starting from STnew in the historical evolution sequence of OIDnew. If historical data for OIDnew is lacking, the global transition probability is used as the default value.

[0181] Step S1040: Based on the personalized demand state transition path and the current service demand type identifier, traverse the service object demand state transition topology graph along the directed edge direction, and extract the set of subsequent service demand type identifiers whose transition probability exceeds the preset transition probability threshold, as the set of predicted subsequent service demand type identifiers.

[0182] In this embodiment, starting from the current service demand type identifier node in the service object demand state transition topology graph, the process traverses along all outgoing edges. For each neighbor node pointed to by an outgoing edge, the weight of that edge, i.e., the transition probability, is obtained. Assuming a preset transition probability threshold Pth, neighbor nodes with transition probabilities greater than Pth are selected, and the service demand type identifiers corresponding to these neighbor nodes form a set for predicting subsequent service demand type identifiers.

[0183] Step S1050: For each predicted subsequent service demand type identifier in the predicted subsequent service demand type identifier set, calculate the average time interval from the current service demand type identifier to the predicted subsequent service demand type identifier based on the historical data of the service object demand evolution sequence; and estimate the predicted trigger time window based on the demand trigger timestamp and the average time interval.

[0184] In this embodiment, for each predicted subsequent service demand type identifier STpred in the predicted subsequent service demand type identifier set, all instances that transitioned from STnew to STpred are found in the historical service object demand evolution sequence of that service object. The average time interval between two triggers in these instances is calculated as the average time interval ΔTavg. Let the demand trigger timestamp of the new elderly care service demand instruction unit be TSnew, the center time of the predicted trigger time window be TSnew+ΔTavg, and the window radius be a preset scaling factor multiplied by the standard deviation of ΔTavg. If the standard deviation is not available, a preset default radius is used.

[0185] Step S1060: Based on the attribute definition of the predicted subsequent service demand type identifier itself, or the urgency statistical characteristics marked in the historical service execution record, determine its predicted urgency label, and pre-insert the predictive elderly care service demand instruction unit into the task execution priority sequence.

[0186] In this embodiment, for each STpred, its default urgency label is obtained by querying a pre-stored mapping table. If there is a statistical record in the historical service execution records of the STpred that is marked with urgency, then the urgency label with the highest frequency in the statistical record is used as the predicted urgency label. A predictive elderly care service demand instruction unit is constructed, whose service demand type is identified as STpred, whose service demand trigger timestamp is the center moment of the predicted trigger time window, and whose demand urgency label is the predicted urgency label. The predictive elderly care service demand instruction unit generates corresponding predicted service demand entries according to steps S131 to S137 and inserts them into the task execution priority sequence. In the calculation of its comprehensive urgency ranking value, the urgency weight is the weight corresponding to the predicted urgency label multiplied by a prediction discount coefficient less than 1.

[0187] Step S1070: In the task resource pairing relationship diagram, a task resource pairing relationship is pre-constructed for the predictive elderly care service demand instruction unit. According to the predicted urgency label, a pre-occupancy priority weight is assigned to the pre-constructed task resource pairing relationship. In subsequent actual task scheduling, the pairing relationship whose pre-occupancy priority weight meets the preset weight threshold is given priority in response.

[0188] In this embodiment, a node is created in the task resource pairing relationship graph for the predicted service demand item corresponding to the predictive elderly care service demand instruction unit. The multidimensional adaptation association strength with all service resource items is calculated according to steps S133 to S135 to establish a pre-built task resource pairing relationship. Let the value corresponding to the prediction urgency label be Uppred, and the pre-occupancy priority weight Wpre = Uppred * S, where S is the multidimensional adaptation association strength. The pre-occupancy priority weight is appended to the edge of the pre-built task resource pairing relationship. In subsequent actual task scheduling, the pre-occupancy priority weight of each pre-built pairing relationship is compared with a preset weight threshold. If the pre-occupancy priority weight is greater than the preset weight threshold, the pre-built pairing relationship is directly converted into a deterministic pairing relationship, skipping the filtering process in steps S142 to S145, and quickly generating the service robot action instruction unit.

[0189] Step S1110: After the set of service robot action instructions is distributed to the corresponding service robot control terminal, the group collaborative operation adjustment process is performed on the multiple service robots that are performing service actions.

[0190] In this embodiment, after the service robot action instruction set is distributed, the group collaborative operation adjustment process is performed on multiple service robots that are performing service actions at fixed intervals.

[0191] Step S1120: Receive service execution progress status information and current location coordinate information reported by the control terminals of each service robot in real time, and mark multiple service robots in the same service object or the same service area as a collaborative operation group.

[0192] In this embodiment, the system continuously receives service execution progress status information and current location coordinates reported by the control terminals of each service robot. Based on the location of the target service object in the service request entries bound to each service robot, service robots whose target service object locations are within the same grid area number or adjacent grid area numbers are grouped into a collaborative operation group. Simultaneously, multiple service robots serving the same service object identifier are also marked as belonging to the same collaborative operation group.

[0193] Step S1130: Extract the current action type and service action execution sequence of each service robot in the collaborative operation group, and detect whether there is a service operation dependency relationship between the current action types of each service robot. The service operation dependency relationship means that the completion of the service operation of one service robot is a prerequisite for the start of the service operation of another service robot.

[0194] In this embodiment, for each service robot within the collaborative work group, its current execution action type and expected completion time are extracted. A service operation dependency table is pre-stored. The row structure of this table includes preceding service action types and subsequent service action types, indicating that the completion of the preceding service action type is a prerequisite for the start of the subsequent service action type. The current execution action type combinations of every two service robots within the collaborative work group are traversed, and the service operation dependency table is queried. If the current execution action type of one robot is a preceding service action type of the current execution action type of the other robot, then a service operation dependency relationship exists between them; the former is the preceding service robot, and the latter is the subsequent service robot.

[0195] Step S1140: When an execution progress deviation is detected between two service robots with service operation dependencies, the time interval between the actual completion time of the preceding service robot and the planned start time of the following service robot is calculated. When the time interval exceeds a preset coordination tolerance threshold, a dependency coordination adjustment requirement is generated.

[0196] In this embodiment, for a front-end service robot and a back-end service robot with service operation dependencies, let the actual completion time of the front-end service robot be Tfinish_pre, and the planned start time of the back-end service robot be Tstart_post. The time interval ΔTdep is calculated as Tstart_post - Tfinish_pre. If ΔTdep is less than a preset coordination tolerance threshold, it indicates that the planned start time of the back-end service robot is too close to the actual completion time of the front-end service robot, posing a risk that a delay in the front-end service robot will cause the back-end service robot to wait. In this case, a dependency coordination adjustment request is generated.

[0197] Step S1150: Dynamically postpone the execution timing of the service actions of the rear service robot, shifting the execution timing of the service actions backward by the time interval, and simultaneously detecting whether the postponement of the execution timing of the service actions of the rear service robot causes a new timing conflict with the execution timing of the service actions of other service robots in the same collaborative operation group. When a new timing conflict is detected, cascade offset adjustment is performed on the execution timing of the service actions of all affected service robots in the collaborative operation group. Based on the actual completion time of the front service robot, the updated execution timing of the service actions of each service robot is recalculated in sequence to generate a cascaded adjusted set of service robot action instruction units.

[0198] In this embodiment, the execution timing of the service actions of the subsequent service robots is shifted backward by an adjustment amount equal to a preset coordination tolerance threshold minus ΔTdep, so that the adjusted Tstart_post_new - Tfinish_pre equals the preset coordination tolerance threshold. It is then detected whether the delayed execution timing of the subsequent service robots causes spatial overlap with the execution timing of the service actions of other service robots in the collaborative work group, exceeding a preset conflict threshold, within the same time segment. If a new timing conflict occurs, the updated execution timing of each subsequent service robot is recalculated based on the actual completion time of the earliest preceding service robot in the collaborative work group, according to the topological order of the service operation dependency graph, ensuring that the time interval of each dependency pair meets the preset coordination tolerance threshold. The updated service action execution timing is then written into the corresponding service robot action instruction unit, generating a cascaded adjusted set of service robot action instruction units.

[0199] Step S1160: Replace the corresponding service robot action instruction units in the original service robot action instruction set with the cascaded adjusted service robot action instruction unit set, distribute the updated service robot action instruction set to the control terminals of each service robot in the collaborative operation group, and simultaneously extract the idle time segments in the execution sequence of the updated service actions of each service robot in the collaborative operation group. In the task execution priority sequence, search for the pending elderly care service demand instruction units whose comprehensive urgency ranking value exceeds the preset emergency threshold, determine whether the pending elderly care service demand instruction units can be inserted into the idle time segments for parallel execution, and generate dynamic insertion scheduling instructions.

[0200] In this embodiment, the original service robot action instruction units in the service robot action instruction set are replaced with the cascaded adjusted set of service robot action instruction units. The updated set of service robot action instruction units is distributed to the control terminals of each service robot within the collaborative work group. Simultaneously, the execution sequence of updated service actions of each service robot within the collaborative work group is analyzed to identify idle time segments for each service robot after completing all currently assigned tasks. Idle time segments are defined by a start time and an end time. The pending elderly care service demand instruction units in the task execution priority sequence whose comprehensive urgency ranking value exceeds a preset emergency threshold are traversed. For each pending elderly care service demand instruction unit, it is determined whether its service target location is within the serviceable range of the service robot with an idle time segment, and whether its estimated execution time is less than or equal to the length of the idle time segment. If both conditions are met, the pending elderly care service demand instruction unit is inserted into the idle time segment, generating a dynamic insertion scheduling instruction. The dynamic insertion scheduling instruction includes the service robot identifier, the target service target location of the service demand entry corresponding to the pending elderly care service demand instruction unit, the service action type, and the dynamically adjusted service action execution sequence. Add dynamic insertion temperature control commands to the service robot action command set and distribute them.

[0201] Combination Figure 2 Content, Figure 2This is one of the front-end interface diagrams of the multi-task collaboration and scheduling method for elderly care service robots provided in this embodiment of the invention, demonstrating the overall layout of the elderly care service robot scheduling and monitoring interface. The top of the interface is a title bar displaying "Elderly Care Service Robot Scheduling and Monitoring Interface" and "System Status: Online". The left panel is a "Real-time Task List," containing three horizontally arranged task cards. The first card is highlighted, labeled "Emergency," and includes the tag "Task T003: R101 Medicine Delivery," with a progress bar showing 80% completion. The second and third cards are labeled "High" and "Medium," respectively. The central panel is an "Environment Map and Scheduling View," depicting a simplified floor plan with thin solid lines, including rectangular rooms labeled "101" and "102" and corridors. The map displays the robot icon "Bot-02" and the elderly person icon "R101," with a blue dashed arrow pointing from the robot icon to the elderly person icon, indicating the planned path. The right panel displays "Robot Resource Status," with a vertical list showing the status of each robot, including "Bot-01" (85% battery, in progress), "Bot-02" (delivering medicine), and "Bot-03" (low battery, standby), all marked in red. The diagram also includes connections and annotations: a dashed line with an arrow leads from "Task T003" on the left task card to the "R101" elderly person's location on the central map; a dashed line with an arrow leads from "Bot-02" on the right resource panel to the robot icon on the central map. The intersection of these two dashed lines is marked with "Task-Resource Pairing Relationship." The bottom legend box explains: "Solid lines: Interface border; Dashed lines: Logical connections; Arrows: Data flow."

[0202] Combination Figure 3 Content, Figure 3This is the second front-end interface diagram of the multi-task collaboration and scheduling method for elderly care service robots provided in this embodiment of the invention, showing the interactive interface of the anomaly alarm and rescheduling confirmation pop-up. This pop-up is overlaid on the scheduling and monitoring interface, using a thick red dashed line to highlight the anomaly and intervention process. The pop-up title bar is filled in red (or represented by a red diagonal line), and displays "Anomaly Alarm and Rescheduling Confirmation" inside. The anomaly information area contains a prominent red exclamation mark icon, along with three lines of text: "Anomaly Type: Path Blockage", "Affected Task: R101 Medicine Delivery (T003)", and "Affected Robot: Bot-02". The scheme comparison area is divided into two sub-boxes: the left box, "Original Path (Interrupted)," displays a thumbnail of the central map, showing a solid red line from Bot-02 to R101 interrupted by a red "X" at a certain point in the corridor; the right box, "Suggested New Path," displays the same thumbnail map, showing the Bot-02 icon bypassing another corridor via a green dashed line to reach R101, with the label "Bot-02 Detour" next to the dashed line. The decision information area displays "System suggestion: Task T003 will continue execution by Bot-02, estimated delay +2 minutes." The operation button area contains three side-by-side buttons: "Approve Plan" (green border), "View Details" (blue border), and "Cancel" (gray border). The diagram also includes patented process annotations: a red dashed arrow leads from the "Path Blocked" text in the error information area to the red "X" in the left frame, and a green dashed arrow leads from the "Bot-02 Detour" annotation in the right frame to the "Approve Plan" button at the bottom. Outside the pop-up window, there is a... Figure 2 The arrow leading from the "Bot-02" status bar in the interface points to this pop-up window, with the label "Interaction triggered by abnormal status" next to it.

[0203] In an exemplary embodiment, a multi-task collaboration and scheduling system for elderly care service robots is provided. This system can be a terminal, server, etc., and its internal structure diagram may include a processor, memory, input / output interface, communication interface, display unit, and input device. The processor, memory, and input / output interface are connected via a system bus, and the communication interface, display unit, and input device are also connected to the system bus via the input / output interface. The processor provides computing and control capabilities. The memory includes a non-volatile storage medium and internal memory. The non-volatile storage medium stores an operating system and computer programs. The internal memory provides an environment for the operation of the operating system and computer programs in the non-volatile storage medium. The input / output interface is used for exchanging information between the processor and external devices. The communication interface is used for wired or wireless communication with external terminals; wireless communication can be achieved through Wi-Fi, mobile cellular networks, near-field communication, or other technologies. When the computer program is executed by the processor, it implements a multi-task collaboration and scheduling method for elderly care service robots. The display unit is used to form a visually visible image and can be a display screen, projection device, or virtual reality imaging device. The display screen can be an LCD screen or an e-ink screen. The input device can be a touch layer covering the display screen, or it can be a button, trackball, or touchpad set on the shell of the multi-task collaboration and scheduling system used in elderly care service robots, or it can be an external keyboard, touchpad, or mouse, etc.

[0204] It should be noted that, in order to simplify the description of the present invention and thus help to understand one or more embodiments of the invention, multiple features may sometimes be grouped into one embodiment, drawing or description thereof in the foregoing description of the embodiments of the present invention.

Claims

1. A multi-task collaboration and scheduling method applied to elderly care service robots, characterized in that, The method includes: The system acquires a sequence of elderly care service demand instructions and an information flow for elderly care environment perception. The sequence of elderly care service demand instructions includes multiple elderly care service demand instruction units, each of which corresponds to a service demand type identifier and a service demand trigger timestamp. The information flow for elderly care environment perception includes a multi-dimensional environmental status perception data flow collected by multiple perception nodes within the elderly care environment. The elderly care service demand instruction sequence and the elderly care environment perception information flow are processed by service scenario parsing to obtain a service object status feature set and a service resource status feature set. The service object status feature set includes service object location features and service object behavior features. The service resource status feature set includes service robot location features and service robot load features. Based on the service object status feature set and the service resource status feature set, task response association construction processing is performed to generate a task resource pairing relationship graph and a task execution priority sequence. The task resource pairing relationship graph is used to characterize the multidimensional adaptation association strength between each elderly care service demand instruction unit and the available service robot, and the task execution priority sequence is used to characterize the order of execution urgency of each elderly care service demand instruction unit. Based on the task resource pairing relationship diagram and the task execution priority sequence, multi-task collaborative scheduling is performed to generate a service robot action instruction set. The service robot action instruction set is then distributed to the corresponding service robot control terminal to trigger service execution operations. The service robot action instruction set contains multiple service robot action instruction units, each of which corresponds to a service robot identifier, a target service object location, a service action type, and a service action execution sequence.

2. The multi-task collaboration and scheduling method for elderly care service robots according to claim 1, characterized in that, The process of parsing the elderly care service demand instruction sequence and the elderly care environment perception information flow yields a service object status feature set and a service resource status feature set, including: The elderly care service demand instruction sequence is parsed to extract the service demand type identifier and service demand trigger timestamp from each elderly care service demand instruction unit. The service demand type identifier is mapped to a demand intent category label and a demand urgency label, and the service demand trigger timestamp is mapped to a demand trigger time node. The elderly care environment perception information flow is subjected to perception data diversion processing. Based on the source perception node identifier of each perception data in the elderly care environment perception information flow, the elderly care environment perception information flow is divided into a service object perception data subset and a service resource perception data subset. Extract service object location perception data sequences and service object behavior perception data sequences from the service object perception data subset, perform location trajectory tracking processing on the service object location perception data sequences to generate continuous location change trajectories for each service object, and perform behavior pattern recognition processing on the service object behavior perception data sequences to generate behavior state description vectors for each service object. Extract service robot position perception data sequences and service robot load perception data sequences from the service resource perception data subset; perform position trajectory tracking processing on the service robot position perception data sequences to generate continuous position change trajectories for each service robot; and perform load state parsing processing on the service robot load perception data sequences to generate load state description vectors for each service robot. The service requirement description vector is formed by combining the requirement intent category label, the requirement urgency label, and the requirement trigger time node, and the service requirement description vector is associated with and stored with the corresponding service object identifier. The continuous position change trajectory of each service object and the behavior state description vector of each service object are combined to form the service object state feature set, and the continuous position change trajectory of each service robot and the load state description vector of each service robot are combined to form the service resource state feature set.

3. The multi-task collaboration and scheduling method for elderly care service robots according to claim 1, characterized in that, The step of constructing a task response association based on the service object state feature set and the service resource state feature set, generating a task resource pairing relationship graph and a task execution priority sequence, includes: Calculate the current location region of each service object based on the location change trajectory of each service object in the service object status feature set, and then spatiotemporally bind the current location region of each service object with the demand triggering time node corresponding to the service object in the elderly care service demand instruction sequence to generate a service demand entry with spatial location marker. The current location region of each service robot is calculated based on the position change trajectory of each service robot in the service resource status feature set. The current location region of each service robot is then bound to the load status description vector of each service robot to generate a service resource entry with spatial location markers and load markers. For each service demand item with a spatial location marker, calculate the spatial distance between the service demand item and all service resource items with spatial location markers. Based on the spatial distance, the mobility parameters of the service robot, and the path accessibility in the elderly care environment, determine the accessibility parameters of the service robot. For each service request entry with a spatial location marker, calculate the task load matching degree between the demand intent category label of the service request entry and the load markers of all service resource entries, and determine the service robot task adaptability parameters based on the task load matching degree. The service robot reachability parameters, service robot task adaptability parameters, and the urgency label of the service request item are fused together to generate a multi-dimensional adaptation association strength between each service request item and each service resource item. The task resource pairing relationship graph is constructed based on the multi-dimensional adaptation association strength. Each node in the task resource pairing relationship graph is a service request item or a service resource item, and the weight of each edge is the multi-dimensional adaptation association strength. Extract the urgency level tags and trigger time nodes of all service request items, determine the urgency weight of each service request item based on the urgency level tags, and determine the waiting time length of each service request item based on the trigger time nodes. The waiting time length of each service request item is converted into a time urgency quantification value by combining it with a preset unit time urgency conversion factor. Based on the urgency weight of each service request item and the time urgency quantification value, a comprehensive urgency ranking calculation is performed on each service request item to generate the task execution priority sequence. Each element in the task execution priority sequence corresponds to a service request item and its comprehensive urgency ranking value.

4. The multi-task collaboration and scheduling method for elderly care service robots according to claim 1, characterized in that, The step of performing multi-task collaborative scheduling based on the task resource pairing relationship diagram and the task execution priority sequence to generate a service robot action instruction set includes: Service request items are extracted one by one from the task execution priority sequence in descending order of comprehensive urgency value, and the currently extracted service request item is taken as the current task to be scheduled. In the task resource pairing relationship graph, find all service resource entries connected to the current task to be scheduled, obtain the multidimensional adaptation association strength of each connected service resource entry, and generate a candidate service robot list in descending order of the multidimensional adaptation association strength. For each candidate service robot in the candidate service robot list, extract the load status description vector of the candidate service robot, calculate the current remaining load capacity of the candidate service robot, compare the current remaining load capacity with the expected task load corresponding to the demand intent category label of the current task to be scheduled, filter out candidate service robots whose remaining load capacity is less than the expected task load, and generate a list of available service robots. For each available service robot in the list of available service robots, extract the current location region of the available service robot and the spatial location marker of the current task to be scheduled, and calculate the estimated movement time for the available service robot to move from the current location region to the service object location corresponding to the spatial location marker; Based on the estimated travel time and the demand trigger time of the current task to be scheduled, the service arrival time of the available service robot to execute the current task to be scheduled is determined. The service arrival time is compared with the preset latest arrival time of the current task to be scheduled. Available service robots with service arrival times later than the preset latest arrival time are filtered out, and a list of compliant service robots is generated. Select the compliance service robot with the highest multidimensional adaptation association strength and the shortest estimated movement time from the list of compliance service robots as the optimal assigned service robot, bind the current task to be scheduled with the optimal assigned service robot, and generate a task assignment relationship record. The service robot identifier in the task assignment relationship record is used as the service robot identifier, the spatial location mark of the current task to be scheduled is used as the target service object location, the demand intent category label of the current task to be scheduled is mapped to the service action type, and the service arrival time is used as the service action execution sequence. These are combined to generate a service robot action instruction unit. The service robot action instruction unit is added to the service robot action instruction set, and the currently scheduled task is removed from the task execution priority sequence. The connection edge related to the currently scheduled task is removed from the task resource pairing relationship graph. The load state description vector of the optimal assigned service robot is updated. The above steps are repeated until the task execution priority sequence is empty, and the complete service robot action instruction set is obtained.

5. The multi-task collaboration and scheduling method for elderly care service robots according to claim 1, characterized in that, The method further includes: After distributing the set of service robot action instructions to the corresponding service robot control terminals, the system receives service execution status feedback information periodically reported by each service robot control terminal. The service execution status feedback information includes the service robot identifier, the type of current action, the current execution progress status, and the current location coordinates. All received service execution status feedback information is grouped and aggregated according to the service robot identifier to generate a continuous status feedback sequence for each service robot. The continuous status feedback sequence of each service robot is then processed for status change analysis to extract the execution progress change curve and position movement trajectory curve of each service robot. The execution progress change curve of each service robot is compared with the expected execution progress corresponding to the demand intent category label of the service demand item bound to the service robot, and the execution progress deviation is calculated. The position movement trajectory curve of each service robot is compared with the service action execution sequence planned in the action instruction unit of the service robot, and the position movement trajectory deviation is calculated. The execution progress deviation is compared with a preset execution progress deviation threshold to obtain an abnormal progress state. The position movement trajectory deviation is compared with a preset position trajectory deviation threshold to obtain an abnormal trajectory state. Based on the abnormal progress state and the abnormal trajectory state, the task execution abnormality index of each service robot is comprehensively determined. The task execution abnormality index is compared with a preset abnormality judgment condition to filter out the abnormal service robot identifier and its associated abnormal task information that the task execution abnormality index exceeds the preset abnormality judgment condition. Extract the urgency label and intent category label of the original service request entry corresponding to the abnormal task from the abnormal task information, and determine the rescheduling urgency of the abnormal task based on the urgency label and intent category label. In the task resource pairing relationship graph, all service resource entries connected to the abnormal task are searched, and the service resource entries corresponding to the abnormal service robot identifier are excluded. A set of alternative service resource entries is obtained. Based on the urgency of rescheduling and the multidimensional adaptation association strength of the set of alternative service resource entries, the abnormal task is reassigned, a compensation service robot action instruction unit is generated, and it is added to the service robot action instruction set. The updated service robot action instruction set is distributed to the corresponding service robot control terminal to trigger the compensation service execution operation.

6. The multi-task collaboration and scheduling method for elderly care service robots according to claim 1, characterized in that, The method further includes: A type correlation analysis is performed on the service demand type identifier of each elderly care service demand instruction unit in the elderly care service demand instruction sequence, and the co-occurrence frequency and temporal correlation of demand intent category tags among different elderly care service demand instruction units are extracted. A service demand type association graph is constructed based on the co-occurrence frequency and temporal correlation. Each node in the service demand type association graph is a service demand type identifier, and each edge indicates that two connected service demand type identifiers have an association triggering relationship within a preset time window. The strength of the edge is determined by the co-occurrence frequency. When a new elderly care service demand instruction unit enters the elderly care service demand instruction sequence, the service demand type identifier and demand trigger timestamp of the new elderly care service demand instruction unit are extracted, and the set of associated service demand type identifiers that have an associated trigger relationship with the service demand type identifier is searched in the service demand type association graph. Based on the set of associated service demand type identifiers and the demand trigger timestamp, a predicted service demand entry is generated. The predicted service demand entry is inserted into the task execution priority sequence. The overall urgency ranking value of the predicted service demand entry in the task execution priority sequence is adjusted according to the predicted association strength parameter. The predicted service demand entry includes a predicted service demand type identifier, a predicted association strength parameter, and a predicted trigger time window. In the task resource pairing relationship diagram, nodes are pre-created for the predicted service demand items, and the pre-allocation multidimensional adaptation association strength between the predicted service demand item and each service resource item is calculated. A pre-occupancy mark is applied to the pre-allocation multidimensional adaptation association strength of the predicted service demand items. The pre-occupancy mark of the predicted service demand items is jointly analyzed with the load state description vector of the service resource items, and the task load reservation amount of the corresponding service robot is adjusted to reduce the task response delay when the actual service demand is triggered.

7. The multi-task collaboration and scheduling method for elderly care service robots according to claim 1, characterized in that, The method further includes: Environmental event detection processing is performed on the elderly care environment perception information flow to identify abnormal environmental events occurring in the elderly care environment. The abnormal environmental events include an environmental abnormal event type identifier, the location coordinates of the environmental abnormal event, and the timestamp of the environmental abnormal event. The type of emergency service requirement associated with the environmental anomaly event is determined based on the environmental anomaly event type identifier; the target area for emergency service is determined based on the location coordinates of the environmental anomaly event; and the trigger time for emergency service is determined based on the timestamp of the environmental anomaly event. An emergency elderly care service demand instruction unit is generated, the type of emergency service demand is used as the service demand type identifier of the emergency elderly care service demand instruction unit, the emergency service trigger time is used as the service demand trigger timestamp of the emergency elderly care service demand instruction unit, and the emergency elderly care service demand instruction unit is inserted at the top of the elderly care service demand instruction sequence. In the task execution priority sequence, the comprehensive urgency ranking value corresponding to the emergency elderly care service demand instruction unit is set to the highest priority, and the task resource pairing relationship with a priority lower than that of the emergency elderly care service demand instruction unit in the currently executing task resource pairing relationship diagram is interrupted. Release the occupied service resource entries from the interrupted task resource pairing relationship, add the released service resource entries back to the task resource pairing relationship graph, and re-execute the task response association construction process for the emergency elderly care service demand instruction unit to generate an emergency task resource pairing relationship. Based on the emergency task resource pairing relationship and the highest priority marker in the task execution priority sequence, the emergency elderly care service demand instruction unit is urgently scheduled and processed to generate an emergency service robot action instruction unit. The emergency service robot action instruction unit is added to the service robot action instruction set, and the updated service robot action instruction set is distributed to the corresponding service robot control terminal to trigger the emergency service execution operation. After the emergency service execution operation is completed, the interrupted task resource pairing relationship is restored to the state before the interruption, and multi-task collaborative scheduling is re-performed.

8. The multi-task collaboration and scheduling method for elderly care service robots according to claim 1, characterized in that, The method further includes: Spatial conflict analysis is performed on the service robot action instruction set to extract the target service object position and service action execution sequence in all service robot action instruction units, and to detect the degree of spatial position overlap between different service robots in the same time segment. When service robot action command units with spatial overlap exceeding a preset conflict threshold are detected in the same time segment, the conflicting service robot set and conflicting time segment are identified. For each conflict service robot in the conflict service robot set, extract its target service object location and current location area, calculate the service path trajectory of the conflict service robot, and generate a service path trajectory description containing a sequence of path nodes; Perform path intersection analysis on the service path trajectory descriptions of every two conflicting service robots in the set of conflicting service robots, and calculate the location coordinates and intersection time points of the service path trajectory intersections. Based on the coordinates of the intersection point of the service path trajectory and the intersection time point, the timing of the service actions of the conflicting service robots is adjusted by delaying the timing of the service actions of one of the conflicting service robots by a conflict avoidance time interval, and generating a timing-adjusted conflicting service robot action instruction unit. During the conflict time segment, path replanning is performed on the conflict service robots in the conflict service robot set. Based on the information of passable areas in the elderly care environment, alternative service path trajectories that avoid intersections are generated, and spatially adjusted conflict service robot action command units are generated. The conflict-resolved service robot action instruction set is obtained by replacing the original service robot action instruction unit in the service robot action instruction set with the time-adjusted conflict service robot action instruction unit and the space-adjusted conflict service robot action instruction unit.

9. The multi-task collaboration and scheduling method for elderly care service robots according to claim 1, characterized in that, The method further includes: Obtain a set of historical service execution records for elderly care service robots. The set of historical service execution records contains multiple historical service execution records. Each historical service execution record contains a historical service demand type identifier, a historical service object location region, a historical service execution start time, a historical service execution duration, and a historical service robot identifier. The historical service execution record set is processed for service pattern mining. Based on the historical service demand type identifier and the historical service execution start time, the frequency distribution of various service demand type identifiers in different time periods is extracted to generate a periodic distribution pattern of service demand. Based on the historical service object location area and historical service demand type identifier, extract the association distribution between different location areas and different service demand type identifiers to generate a service demand spatial distribution pattern; Based on historical service execution duration and historical service demand type identifiers, extract the execution duration distribution characteristics of different service demand type identifiers, and generate a service execution duration reference range; Based on the periodic distribution pattern of service demand, the spatial distribution pattern of service demand, and the reference interval of service execution time, the probability of service demand occurring in each time period and each location area within a preset time period is estimated, and a service demand prediction heatmap is generated. Based on the service demand prediction heatmap, the standby location distribution of service robots is adjusted in advance at the current moment. Service robots in an idle state are dispatched to areas with high predicted service demand to stay and wait. A set of service robot pre-deployment instructions is generated and distributed to the corresponding service robot control terminal to trigger the pre-deployment movement operation.

10. A multi-task collaboration and scheduling system for elderly care service robots, characterized in that, include: processor; A machine-readable storage medium for storing machine-executable instructions of the processor; The processor is configured to execute the multi-task collaboration and scheduling method for elderly care service robots according to any one of claims 1 to 9 by executing the machine-executable instructions.

Images