A dynamic scheduling method, device and medium for a body-possessed charging robot

Abstract

The present application relates to the technical field of charging scheduling, in particular to a dynamic scheduling method, device and medium for a body charging robot, which is used to solve the problem that the existing technology cannot realize strong consistency state management, resource conflict coordination and abnormal closed loop processing under multi-machine cooperation. The method comprises: an edge scheduling server constructs a double state mapping relationship and obtains spatiotemporal characteristic data of each body charging robot reaching a task point; according to the spatiotemporal characteristic data and the double state mapping relationship, the occupation conflict of shared resources is determined, and a set of available body charging robots is screened according to the real-time running state of each body charging robot; a scheduling strategy is generated by taking the conflict determination result and the available robot set as constraints, an atomized change operation is executed to lock the shared resources and an execution instruction is issued; stage data reported by the robot is obtained, state rollback is triggered when an abnormality is determined based on the double state mapping relationship, the locked shared resources are released, and a scheduling update corresponding to the abnormal event is executed.

Description

Technical Field

[0001] This invention relates to a dynamic scheduling method, device, and medium for a self-contained charging robot, belonging to the field of charging scheduling technology. Background Technology

[0002] With the rapid popularization of new energy vehicles, the demand for charging in scenarios such as parks, parking lots, and warehouses is increasing day by day. Traditional fixed charging piles require vehicles to drive themselves to the charging position, which leads to problems such as irregular vehicle parking, low utilization of charging positions, and difficulty for users to find charging positions. To solve these problems, mobile-operated automatic charging robots have emerged. These robots use a mobile chassis to autonomously reach the target parking space and use dual-arm end effectors to complete the entire charging process for humans, thereby completing the charging service without changing the vehicle's parking position or reducing human intervention.

[0003] However, most current mobile automated charging robots simplify the charging task to dispatching, arrival, and completion, or focus only on the insertion and removal of the charging gun itself. They lack a full-chain state design covering arrival at the parking space, alignment / insertion / holding / confirmation, charging, removal, and return. Therefore, the platform struggles to provide interpretable management of the task lifecycle. Furthermore, charging scenarios present resource contention issues such as mutually exclusive fixed charging spots and limited passing in narrow passages. Existing scheduling methods, if based solely on distance or estimated arrival time, lack conflict resolution mechanisms for parking space pre-reservation and passageway occupancy control, easily leading to resource contention. In addition, current methods neglect the relationship between the robot's own state and continuous service capability, making it difficult to achieve optimal task allocation at similar or close distances.

[0004] Therefore, a dynamic scheduling method for mobile-operated automated charging robots is needed to achieve strong consistency state management, resource conflict coordination, and anomaly handling for multiple mobile-operated automated charging robots. Summary of the Invention

[0005] This invention addresses the shortcomings of existing technologies by providing a dynamic scheduling method, device, and medium for a self-contained charging robot.

[0006] The technical solution of the present invention to solve the above-mentioned technical problems is as follows: A dynamic scheduling method for mobile-operated automated charging robots is needed to achieve strong consistency state management, resource conflict coordination and anomaly handling of multiple mobile-operated automated charging robots.

[0007] This application provides one or more embodiments of a dynamic scheduling method for a unibody charging robot, applied to a dynamic scheduling system including a park platform, an edge scheduling server, and multiple unibody charging robots, wherein the unibody charging robot is a mobile, operable, automated unibody charging robot, and the method includes: The edge scheduling server constructs a dual-state mapping relationship between the task state machine and the robot state machine, and obtains the spatiotemporal feature data of each of the embodied charging robots arriving at the task point corresponding to the charging task; wherein, the charging task has time constraints. Based on the spatiotemporal feature data and the dual-state mapping relationship, the conflict of shared resource occupation is determined, and based on the real-time operating status of each of the embodied charging robots, a set of available embodied charging robots is obtained; wherein, the shared resources include at least: fixed charging positions and passageways; Based on the determination result of the occupancy conflict and the set of available robots, a scheduling strategy corresponding to the charging task is generated. Based on the scheduling strategy, atomic change operations are performed to lock the state of the shared resources and issue execution instructions to the corresponding embodied charging robot. If the phase data reported by the corresponding embodied charging robot is obtained, and an anomaly is determined based on the dual-state mapping relationship, a state rollback is triggered, the locked shared resources are released, and the scheduling update corresponding to the anomaly event is executed.

[0008] Optionally, in one or more embodiments of this application, a dual-state mapping relationship between the task state machine and the robot state machine is constructed, and spatiotemporal feature data of the task point corresponding to the arrival of each of the embodied charging robots at the charging task is obtained, specifically including: Based on the predefined dual-state mapping rules between the task state machine and the robot state machine, the matching correspondence between the task state and the executing robot state at any given time is determined, and the matching correspondence is used as the dual-state mapping relationship. The real-time operating data reported by each of the aforementioned embodied charging robots is acquired; wherein, the real-time operating data includes at least: the real-time position and motion vector obtained based on the built-in navigation module of the embodied charging robot; Based on the real-time operation data and the current park map data, the spatiotemporal characteristic data of each of the embodied charging robots arriving at the task point corresponding to the charging task are determined; wherein, the acquisition of the spatiotemporal characteristic data includes: the estimated arrival time and the channel traversal time window.

[0009] Optionally, in one or more embodiments of this application, determining the conflict of shared resource occupancy based on the spatiotemporal feature data and the dual-state mapping relationship specifically includes: Based on the pre-defined strong consistency principle, the mutual exclusion state and occupancy sequence of the shared resources are obtained, and a baseline state context is constructed based on the mutual exclusion state and the occupancy sequence. The spatiotemporal feature data is analyzed to generate a time-series model of resource consumption for the charging task; By comparing the baseline state context with the resource occupancy time series model, a subset of suspected conflicts with temporal overlap is obtained based on capacity and mutual exclusion constraints; Based on the dual-state mapping state, the suspected conflict subset is filtered to obtain the valid conflict subset; The shared resources located in the subset of valid conflicts are determined to have an occupancy conflict.

[0010] Optionally, in one or more embodiments of this application, a scheduling strategy corresponding to the charging task is generated based on the determination result of the occupancy conflict and the set of available robots, specifically including: Taking the embodied charging robot in the available robot set and the charging task as objects, decision variables are constructed, and with the goal of minimizing task waiting time and maximizing completion rate, an optimization objective function is constructed by combining priority weights and the robot's remaining working time. The determination result of the occupation conflict is converted into a feasibility constraint; wherein, the feasibility constraint is used to limit the same fixed charging position and the same passage to be occupied by only a single task within the same time window; Based on the dual-state mapping relationship, a consistency constraint between the task state and the robot state is determined; wherein, the consistency constraint is used to constrain the scheduling strategy to satisfy the matching consistency between the task state and the robot state. Set a reassignment prohibition constraint for tasks that have entered the execution phase; wherein, the reassignment prohibition constraint is used to prohibit non-abnormally triggered task reassignment during the execution process; Based on a rolling time-domain approach, either periodic or event-driven, the optimization objective function is solved under the constraints of feasibility, consistency, and rescheduling prohibition, and the scheduling strategy is obtained through multi-stage conflict resolution.

[0011] Optionally, in one or more embodiments of this application, an atomic change operation is performed based on the scheduling strategy to lock the state of the shared resource and issue an execution instruction to the corresponding embodied charging robot, specifically including: The scheduling strategy is parsed to obtain the shared resource identifier, the corresponding embodied charging robot identifier, and the time window corresponding to the shared resource occupation. Based on the atomic transaction method, the state of the shared resource in the global state set is updated to occupied, and the state of the corresponding embodied charging robot is migrated to the executing state based on the dual state mapping relationship. After the atomic transaction is successfully submitted, an execution instruction is generated and sent to the corresponding embodied charging robot based on the corresponding embodied charging robot identifier; If the atomic transaction commit fails, the current state change is rolled back, and the scheduling policy generation is retried.

[0012] Optionally, in one or more embodiments of this application, obtaining the phase data reported by the corresponding embodied charging robot, and triggering a state rollback to release the locked shared resources when it is determined that the embodied charging robot has an anomaly based on the dual-state mapping relationship, specifically includes: The system receives real-time stage data reported by the embodied charging robot; wherein the stage data includes at least: estimated navigation arrival time, arrival event, docking stage event, and abnormal event; By comparing the current task execution state of the stage data with the expected state transition path of the dual-state mapping relationship, the execution progress of the charging task is determined, and whether it conforms to the preset transition rules of the task state machine and the robot state machine. If not, an abnormal event occurs, triggering a state rollback. The task state machine and the robot state machine are then rolled back to their pre-abnormal state in an atomic manner, and the locked shared resources are released.

[0013] Optionally, in one or more embodiments of this application, performing the scheduling update corresponding to the abnormal event specifically includes: Identify the anomaly type corresponding to the abnormal event, and determine the set of affected scheduled tasks based on the anomaly type; Based on the released shared resources, the set of tasks to be scheduled, and the set of currently available robots, a local rescheduling is performed to obtain an updated scheduling strategy. Based on the updated scheduling policy, return to execute the atomic change operation to perform the scheduling update corresponding to the abnormal event; The method further includes: writing the cause code corresponding to the exception type into the global state, and returning the cause code to the park platform.

[0014] Optionally, in one or more embodiments of this application, before the edge scheduling server constructs a dual-state mapping relationship between the task state machine and the robot state machine, and obtains the spatiotemporal feature data of each of the embodied charging robots arriving at the task point corresponding to the charging task, the method includes: The edge scheduling server receives charging tasks issued by the park platform based on an event-driven interface; Perform a field completeness check on the charging task to determine whether the charging task has any required fields; If not, the charging task will be rejected and a verification failure response will be returned to the park platform; If so, the spatiotemporal feature data is acquired in response to the charging task.

[0015] This application provides one or more embodiments of a dynamic scheduling device for a unibody charging robot, applied to a dynamic scheduling system including a park platform, an edge scheduling server, and multiple unibody charging robots. The unibody charging robot is a mobile, automated unibody charging robot. The device includes: At least one processor; and, A memory communicatively connected to the at least one processor; wherein, The memory stores instructions that can be executed by the at least one processor to enable the at least one processor to perform any of the methods described above.

[0016] One or more embodiments of this application provide a non-volatile computer storage medium storing computer-executable instructions, wherein the computer-executable instructions are configured to execute any of the methods described above.

[0017] The beneficial effects of this invention are: By constructing a dual-state mapping relationship between the task state machine and the robot state machine, and employing atomic change operations for state submission, the consistency between the task state and the executing robot state at any given time is ensured. Automatic rollback when a state update fails also avoids state inconsistencies caused by partial updates. Predicting time-dimensional conflicts based on spatiotemporal feature data, combined with further state consistency verification using the dual-state mapping relationship, ensures the rationality and feasibility of resource allocation. Based on the real-time operating status of each embodied charging robot, a set of available embodied charging robots is selected, solving the optimal choice problem under competition at the same distance and significantly improving the on-time task completion rate. By receiving real-time stage data reported by the robots and dynamically comparing the execution progress with the expected state transition path based on the dual-state mapping relationship, various anomalies can be identified, triggering atomic state rollback to release resources, sending standardized reason codes back to the park platform, and quickly restoring system order through local rescheduling, thus improving system maintainability. Attached Figure Description

[0018] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments recorded in this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort. In the drawings: Figure 1 A schematic flowchart of a dynamic scheduling method for an embodied charging robot provided in an embodiment of this application; Figure 2 A general architecture diagram of a dynamic scheduling system provided in this application embodiment. Figure 3 This application provides a schematic diagram of a business object and data model in an application scenario. Figure 4 A schematic diagram of a dual-state machine in an application scenario provided by an embodiment of this application; Figure 5 A schematic diagram of a strong consistency state management and atomic commit timing diagram in an application scenario provided by an embodiment of this application; Figure 6 This application provides a schematic diagram of multi-machine scheduling modeling and constraints in an embodiment of the present application. Figure 7 A flowchart of rolling optimization and conflict resolution in an application scenario provided in this application embodiment; Figure 8 This is a schematic diagram illustrating the execution monitoring and anomaly reversal closed loop in one application scenario of this application embodiment; Figure 9 A schematic diagram of the structure of a dynamic scheduling device for an embodied charging robot provided in an embodiment of this application; Figure 10 This is a schematic diagram of the structure of a non-volatile storage medium provided in an embodiment of this application. Detailed Implementation

[0019] This application provides a dynamic scheduling method, device, and medium for a self-contained charging robot.

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

[0021] As described in the background section, the large-scale popularization of new energy vehicles has driven the rapid construction of charging infrastructure. Currently, the main infrastructure consists of fixed charging piles and fixed charging spaces, supplemented by measures such as reservation queuing, traffic guidance and diversion, and operation management to improve utilization. However, in semi-enclosed scenarios such as parking lots, parks, and warehouses, charging demand exhibits significant spatiotemporal fluctuations and hotspot migration characteristics. The fixed charging pile system has a rigid supply location, which easily leads to problems such as the coexistence of local queues and idleness, and high costs for car owners to find charging piles and manual operation.

[0022] To address the aforementioned issues, mobile-operated automatic charging robots have emerged. These robots use a mobile chassis to autonomously reach the target parking space and employ dual-arm end effectors to complete the entire charging process for humans. This includes tasks such as opening the vehicle's charging port cover, retrieving the charging gun from the charging station, aligning and inserting the gun, unplugging the gun after charging is complete, resetting the cover, and returning the charging gun to the charging station. This allows the charging service to be completed without changing the vehicle's parking location or reducing human intervention.

[0023] As applications move from single-machine pilots to multi-machine deployments, some problems still exist with the current approach, which are analyzed in detail below: In current technologies for automated charging and robotic plug-in / plug-out guns aimed at replacing human operation, some solutions have been researched and engineered to address aspects such as vehicle charging port positioning, charging gun grasping and plugging / unplugging, and cover opening and closing. These typically employ robotic arms and end effectors to perform alignment, grasping, plugging, and confirmation actions, coupled with process control measures such as vision, force control, and status confirmation. However, while these technologies primarily enable the operation, further systematic integration is often required for large-scale multi-machine deployment, cross-parking space mobile services, and closed-loop platform task scheduling.

[0024] In autonomous navigation mobile robot technology, mobile robots are a type of system that couples autonomous navigation capabilities with robotic arm manipulation capabilities. They can autonomously reach target locations and perform operational tasks in environments such as parking lots, parks, and warehouses. Existing solutions typically implement navigation and obstacle avoidance, as well as task execution, in layers. The navigation module provides the destination and path, while the operation module completes local docking and action sequences. This type of technology provides the hardware and basic capabilities for moving to parking spaces and performing charging operations for people, but application-level task lifecycle management, resource mutual exclusion, and multi-robot collaborative scheduling still require specialized design.

[0025] In multi-robot task allocation and dynamic scheduling technologies, such as those used in warehouse AGVs, inspection robots, and delivery robots, common multi-robot scheduling methods include: greedy dispatching based on nearest distance and shortest time, distributed allocation based on auction, centralized allocation based on bipartite graph matching, and rolling optimization based on mixed integer programming. Although some solutions consider time windows or capacity constraints, they are mostly based on task models where completion is achieved upon reaching the target point, and they do not adequately cover the characteristics of mobile automatic charging applications, such as long service durations, strong resource mutual exclusion, fine-grained operation stages, and closed-loop reasons for cancellation or withdrawal.

[0026] In edge computing and platform-robot collaborative control technologies, park-type systems often employ edge servers to handle scheduling decisions and data aggregation, while robots periodically report their status and execute commands. Existing systems mostly adopt a task distribution and execution reporting model, which can meet general latency requirements. However, when it is necessary to maintain consistency among multiple objects such as tasks, robots, fixed parking spaces, and passageways, and to support anomaly compensation and traceable auditing, a more stringent state closed-loop and consistency design is often required.

[0027] As mentioned above, most existing technologies simplify charging tasks to dispatching, arrival, and completion, or focus only on the insertion and removal of the charging gun itself. They lack a comprehensive design covering the entire charging process, including arrival at the parking space, alignment / insertion / holding / confirmation, charging, removal, and return. This makes it difficult for the platform to provide interpretable management of task progress, failure points, and recovery paths, and also hinders the standardized closed-loop feedback of abnormal withdrawal reason codes. Furthermore, charging scenarios present resource contention issues such as mutually exclusive fixed charging spots and limited vehicle access in narrow passages. Existing scheduling methods, if based solely on distance or estimated arrival time, lack strategies for pre-reserving parking spots, controlling passage occupancy, and resolving conflicts, easily leading to resource contention, passage blockage, and even deadlocks, reducing overall system throughput. In addition, existing technologies often neglect robot battery thresholds, recharge strategies, and multi-task consolidation capabilities, making it difficult to achieve optimal allocation of tasks at similar or close distances under hard time window constraints. Moreover, they lack a unified compensation system for abnormal scenarios such as vehicle departure or insufficient space.

[0028] To address the challenges of existing technologies in multi-robot deployment scenarios for mobile, operable automatic charging robots, such as balancing fine-grained operational process state loops, feasibility guarantees under hard time windows, multi-robot conflict and mutual exclusion resource constraints, and engineering requirements like cancellation, withdrawal, and cause code feedback, this application proposes a dynamic scheduling method for embodied charging robots. This method is used in a park platform, edge scheduling server, and dynamic scheduling system for multiple embodied charging robots to support large-scale, stable, and high-completion-rate proactive services. Figure 1 As shown in the diagram, this application provides a schematic flowchart of a dynamic scheduling method for a self-contained charging robot. Figure 1 As can be seen, in one or more embodiments of this application, a dynamic scheduling method for a body-mounted charging robot is applied to, for example... Figure 2 The system shown includes a park platform, an edge scheduling server, and a dynamic scheduling system for multiple embodied charging robots. The embodied charging robots are mobile, automated embodied charging robots. The method includes: S101: The edge scheduling server constructs a dual-state mapping relationship between the task state machine and the robot state machine, and obtains the spatiotemporal feature data of each of the embodied charging robots arriving at the task point corresponding to the charging task; wherein, the charging task has time constraints.

[0029] This method is applied to, for example Figure 2 The system shown comprises a park platform, an edge scheduling server, and multiple embodied charging robots, which are mobile, automated embodied charging robots. Specifically, the park platform serves as a unified task entry point, responsible for generating charging tasks and distributing them to the edge scheduling server. Each task includes at least: target vehicle identifier, fixed charging location identifier, vehicle location, task trigger time, charging mode and expected duration, priority level, and task validity period or service time limit (i.e., hard time window).

[0030] The edge scheduling server, acting as the scheduling and state management center, includes the following modules: Task access and verification module: verifies the completeness of task fields and performs reachability prediction based on the ETA provided by the navigation module; Global state management module: maintains a strong consistency state for objects such as tasks, robots, charging positions, and narrow passages; Dual state machine engine module: defines the task state machine and robot state machine, responsible for state transitions and compensation actions; Resource pre-allocation and conflict coordination module: performs mutual exclusion pre-allocation and occupancy control for resources such as fixed charging positions and narrow passages; Multi-robot optimized scheduling module: solves multi-robot allocation and sequence decision-making in the rolling time domain and outputs dispatch and pre-allocation plans; Execution monitoring and anomaly withdrawal module: monitors execution progress under second-level state synchronization, handles anomalies such as vehicle departure, cancellation, timeout, and insufficient space, and returns cause codes.

[0031] The mobile, operable, automated charging robot, as the execution unit, includes at least: an autonomous mobile navigation module, a docking and operation module, and a communication and status reporting module. The docking and operation module supports fine-grained action stages such as alignment / insertion / holding / confirmation, and performs operations such as removing the charging gun, returning it, and resetting the cover plate after charging is complete.

[0032] And such as Figure 2 and Figure 3 The modules shown above are connected via event-driven mechanisms. For example, after the park platform pushes a charging task to the edge server, the edge server sends an execution command to the robot. The robot reports navigation ETA, arrival events, docking phase events, and abnormal events to the edge server at a frequency of seconds. The edge server then makes decisions and sends back information based on the strongly consistent state storage.

[0033] In process S101, the edge scheduling server predefines the task state machine and the robot state machine, and establishes state mapping rules between them. These state mapping rules constrain the matching and correspondence between the task state and the executing robot state at any given time.

[0034] Specifically, such as Figure 4 As shown in the diagram of the two-state machine, Figure 4The task state machine in the model is used to characterize the entire lifecycle of a charging task from creation to completion, and includes at least the following state nodes: Created, Assigned, EnRoute, Docking-Align, Docking-Insert, Docking-Grip, Docking-Verify, Charging, Completed, and exception / withdrawal related states: Paused, Cancelled, Expired, and Failed. Paused is used for recoverable situations such as insufficient space or channel blockage.

[0035] Figure 4 The robot state machine in the model is used to characterize the working state of the embodied charging robot, including at least the following state nodes: Idle, Serving, ReturningHome, SelfCharging, and exception-related states: Paused, Degraded, and Fault.

[0036] Since the state mapping relationship is used to constrain the matching correspondence between the task state and the executing robot state at any given time, when the task state machine is in the execution phase such as Docking-Align, Insert, Grip, Verify, or Charging, the corresponding robot state machine must be synchronously in the Serving state; when the task state machine is in an abnormal withdrawal state, the corresponding robot state machine should fall back to the Idle or ReturningHome state. Therefore, based on this state mapping relationship, it is ensured that any task can only be executed by a single robot, satisfying the state consistency constraint.

[0037] While constructing the bistate mapping relationship, or while constructing the bistate mapping relationship to stop vomiting, such as Figure 5 The edge scheduling server shown reads the spatiotemporal characteristic data of each embodied charging robot arriving at the corresponding task point for the charging task. Specifically, as... Figure 5 Each of the embodied charging robots reported real-time operating data to the edge scheduling server at a frequency of seconds, including at least: real-time position, motion vector, and current battery percentage obtained based on the embodied charging robot's built-in navigation module. and remaining work time estimate , stipulates that At this time, the robot must enter the recharge process and will not participate in order taking. The edge scheduling server calls the navigation module interface, combines it with the current park map data, and for any charging task t to be connected, calculates the distance each charging robot r can travel from its current location to the task point. Estimated arrival time:

[0038] The estimated entry time window for each narrow passage (c) along the driving route. With exit time window Channel crossing time window estimation: The above parameters are integrated and stored as the spatiotemporal feature data corresponding to this charging task.

[0039] In addition, it should be noted that, in order to achieve the above functions, such as Figure 2 or Figure 3 The embodiments shown in this application also provide a unified abstract set of four types of objects: task set. Robot Collection 1. Fixed charging position set S, 2. Narrow channel set C.

[0040] For any task Define fields: in For task identification, For fixed charging position identification, The target location is also the task location; Priority; For the requested time; This is the hard cutoff time; In charging mode; For service duration. If ,but (Hours) converted to minutes; if ,but The estimated value is provided by the platform or an empirical model. Each task is bound to a fixed charging spot via spot_id, establishing a correspondence between tasks and charging spots.

[0041] For any robot Define fields:

[0042] in, For robot identification, Current position This represents the current battery percentage. In robot state, Estimated Time Remaining is used to handle choices made at similar distances. This represents the maximum number of orders. This is a designated point for returning to the nest to replenish energy. It is stipulated that... At this time, the robot must enter the recharge process and will not participate in order taking.

[0043] For fixed charging positions Maintain mutually exclusive states: idle / pre-empted / occupied, and cannot be concurrent; for narrow channels C, Maintaining capacity constraints and the time window table Used for vehicle passing restrictions.

[0044] Narrow passage objects are capacity-constrained resources. Each passage contains at least: a passage identifier (corridor_id), a capacity cap (fixed to 1, indicating that only one robot is allowed to pass at the same time), and an occupancy time window table (reservation_table), which records the time window during which each robot occupies the passage.

[0045] Based on the above steps, the edge scheduling server possesses both a static rule framework, i.e., a dual-state machine mapping, and dynamic real-time data, i.e., spatiotemporal feature data, providing a reliable data foundation for subsequent scheduling decisions and execution.

[0046] Furthermore, in one or more embodiments of this application, before the edge scheduling server constructs a dual-state mapping relationship between the task state machine and the robot state machine, and obtains the spatiotemporal feature data of each of the embodied charging robots arriving at the task point corresponding to the charging task, in order to perform access control and legality verification on the charging tasks issued by the park platform, and to ensure that subsequent scheduling decisions can be based on complete and valid task information, the method further includes the following steps: The edge scheduling server receives charging tasks from the park platform in real time based on an event-driven interface. This event-driven interface uses a subscription-push mechanism; when the park platform generates a new task, it actively pushes the task data to the edge scheduling server to achieve low-latency task access. Then, the charging task undergoes field completeness validation to determine if it has any required fields. This validation may include: checking if the target vehicle identifier is non-empty and conforms to the format specification; checking if the fixed charging position identifier exists in the fixed charging position set S of the global state set; checking if the vehicle location has valid coordinates; and checking if the task trigger time is valid. If any fields are missing, formatted incorrectly, or otherwise fail the validation, the validation fails, the charging task is rejected, and a validation failure response is returned to the park platform. Conversely, if the validation passes, the charging task is integrated into the system, its task status is initialized to "created" in the global state set, and the subsequent process of acquiring spatiotemporal feature data is triggered. In this approach, field completeness verification ensures that tasks accessing the system have an executable basis, avoiding the waste of scheduling resources due to incomplete task information. This achieves access control for charging tasks and provides effective input for charging tasks to subsequently build dual-state mapping relationships, obtain spatiotemporal feature data, and generate scheduling strategies.

[0047] Specifically, in one or more embodiments of this application, a dual-state mapping relationship between the task state machine and the robot state machine is constructed, and spatiotemporal feature data of the task point corresponding to the arrival of each of the embodied charging robots for the charging task is obtained, specifically including: The edge scheduling server determines the matching relationship between the task state and the executing robot state at any given time based on predefined dual-state mapping rules between the task state machine and the robot state machine, and uses this matching relationship as the dual-state mapping relationship. The state mapping rules are used to constrain the matching relationship between the task state and the executing robot state at any given time, specifically including: forward matching rules (e.g., when the task state machine is in an execution phase such as alignment, insertion, gripping, confirmation, or charging, the corresponding robot state machine must be synchronously in the execution state); exclusivity constraints (e.g., any task can only be executed by a single robot); abnormal rollback rules (e.g., when the task state machine is in an abnormal withdrawal state, the corresponding robot state machine should roll back to the idle or homing state); and single-task constraints (e.g., a robot can simultaneously execute a maximum of 3 tasks, and single-task execution is only supported within a 10-meter neighborhood).

[0048] The system acquires real-time operational data reported by each android charging robot. This real-time operational data includes at least: real-time position and motion vectors obtained based on the robot's built-in navigation module, as well as battery status, remaining working time, and current status. Based on the real-time operational data and the current park map data, the spatiotemporal characteristic data of each android charging robot's arrival at the charging task point is determined. The acquisition of spatiotemporal characteristic data includes: estimated arrival time and channel traversal time window, as described above. Figure 2 or Figure 3 The navigation module calculates the distance from the current position to the task point for each charging task t to be connected. The estimated arrival time is:

[0049] And the estimated entry time window for each narrow passage c that needs to be passed through in the driving route. With exit time window The estimated channel traversal time window is: .

[0050] S102: Based on the spatiotemporal feature data and the dual-state mapping relationship, determine the conflict of shared resource occupation, and based on the real-time operating status of each of the embodied charging robots, filter to obtain a set of available embodied charging robots; wherein, the shared resources include at least: fixed charging positions and passageways.

[0051] After obtaining the dual-state mapping relationship and spatiotemporal feature data based on S101, in order to address the resource contention issues caused by the scarcity of fixed charging positions as a resource that cannot be used concurrently, and the constraints of passing vehicles in narrow channels in multi-robot collaborative operation scenarios of mobile charging robots, as well as the invalid conflicts and resource waste caused by the discrepancy between the resource pre-occupancy state and the actual execution state of the charging task or robot, this embodiment of the specification will determine the occupation conflict of shared resources based on the dual-state mapping relationship and spatiotemporal feature data obtained in S101. Simultaneously, to address the problem of unreasonable task allocation caused by selecting robots solely based on distance or estimated arrival time without considering the robot's own battery threshold and health status, this embodiment of the specification also selects a set of available embodied charging robots based on the real-time operating status of each embodied charging robot. During this conflict determination process, the edge scheduling server reads the current state of the shared resources from the global state set. For each fixed charging position, its mutually exclusive status, such as Free, Reserved, or Occupied, is obtained. For each narrow channel, its capacity constraint-corresponding occupancy time window table is obtained, where the time window table records the time interval during which the channel has been reserved or occupied. For the charging task to be connected and the candidate robot, based on the spatiotemporal feature data obtained in S101, during the resource occupancy conflict determination process, the expected arrival time and service duration can be extracted to construct the expected occupancy time period, and the status of the specified charging position within this time period is checked. If it has been reserved or occupied by other tasks and the time periods overlap, it is determined as a charging position occupancy conflict. Alternatively, the expected entry time window and exit time window can be extracted to construct the expected traversal time period, and the channel occupancy time window table within this time period can be checked for overlapping records. If there is overlap, it is determined as a channel occupancy conflict. Simultaneously with resource conflict determination, based on the real-time operating status of each embodied charging robot, a set of available embodied charging robots is obtained. Specifically, this can be based on power conditions. The robot's current battery level is above the forced homing threshold, ensuring continuous service capability; the robot is not in an unavailable state such as faulty, self-charging, or homing state; the robot's current order count is less than the maximum order count, allowing it to accept new tasks; and the reachability condition is that the estimated arrival time is earlier than the task's hard deadline minus a preset safety margin, ensuring arrival within the time window. Robots meeting these conditions are then screened to obtain a set of usable robot charging capabilities.

[0052] Specifically, in one or more embodiments of this application, the determination of shared resource occupancy conflicts based on the spatiotemporal feature data and the dual-state mapping relationship specifically includes: Based on the pre-defined strong consistency principle, the mutual exclusion state and occupancy sequence of shared resources are obtained, and a baseline state context is constructed based on the mutual exclusion state and occupancy sequence. The pre-defined strong consistency principle requires that the task and robot states must satisfy a matching relationship at any given time, for example: ; Furthermore, each task can only be performed by a single robot: .

[0053] Furthermore, the pre-defined strong consistency principle aims to achieve strong consistency at the second level, such as... Figure 4 The embodiment shown in this application employs a pre-defined strong consistency principle that combines single-write by the scheduling center with event fact sources. This pre-defined strong consistency principle includes a single-write principle, a version control principle, and an atomic commit principle. Specifically, the single-write principle means that the robot only reports factual events such as arrival, alignment completion, insertion completion, and insufficient space, while the edge server adjudicates and writes the global state. The version control principle ensures that any event and instruction carries... When versions do not match, the process either proceeds to idempotency handling or discards the state, ensuring monotonic evolution of the state. This is achieved by defining a global state set. For the first This is a snapshot of the next state version, containing the states of all objects including tasks, robots, charging stations, and channel resources. Each scheduling decision and resource pre-allocation is committed in the form of a transaction. ; in This is a set of control actions such as order dispatching / reservation / withdrawal. The robot reports a set of events; This is a state transition function with version verification. To avoid out-of-order and duplicate events and instructions, each event and instruction carries... Yes; when versions do not match, idempotent processing or discarding occurs to ensure monotonic state evolution. And as... Figure 5 The atomic commit principle shown is that state changes are committed in the form of atomic transactions to avoid intermediate states such as duplicate order dispatch, parking space being occupied, and lane collisions, thus ensuring the consistency of the state of multiple objects.

[0054] Therefore, based on the above principles, the edge scheduling server reads the shared resource state under the current state version and constructs a baseline state context. For each fixed charging position s, the baseline state context includes: mutual exclusion state. (Free, Reserved, Occupied) and the occupancy sequence, i.e., if... If a task is reserved or occupied, the identifier of the occupied task t and the occupation time period are recorded. For each narrow channel, the baseline state context includes: capacity constraints and occupation timing records, recording the entry and exit time windows for each occupation. By setting the baseline state context, it can serve as a reference system for subsequent conflict determination, ensuring that all decisions are based on a globally consistent state at the same moment.

[0055] For a charging task *t* to be connected and candidate robots *r* from the set of available embodied charging robots, spatiotemporal feature data is parsed to generate a resource occupancy time-series model for the charging task. This model includes a fixed charging position occupancy time-series and a narrow channel occupancy time-series. The fixed charging position occupancy time-series represents the fixed charging position specified by the charging task, generated based on the expected arrival time and service duration in the spatiotemporal feature data. The narrow channel occupancy time-series represents the channel that robot *r* is expected to exclusively occupy during the channel traversal time window. Specifically, for each narrow channel *c* that robot *r* needs to traverse in its path, the channel occupancy time-series is generated based on the expected entry and exit time windows in the spatiotemporal feature data.

[0056] By comparing the baseline state context with the resource occupancy time series model, a subset of suspected conflicts with temporal overlap is obtained based on capacity and mutual exclusion constraints. Specifically, for each resource occupancy time series in the resource occupancy time series model, it is checked whether the occupancy time series of fixed charging positions in the baseline state context overlaps with the fixed charging position occupancy time series in the resource occupancy time series model. If so, and the fixed charging position is either pre-occupied or occupied, it is considered a suspected conflict. Simultaneously, it is checked whether the occupancy time window table of narrow channels in the baseline state context has any records that overlap with the time series in the resource occupancy time series model. If so, it is recorded as a suspected conflict. Summarizing all suspected conflicts yields a subset of suspected conflicts.

[0057] Based on the dual-state mapping, a subset of suspected conflicts is filtered out, eliminating those caused by inconsistencies in state, thus obtaining a subset of valid conflicts. Shared resources located within the valid subset of conflicts are identified as having occupancy conflicts.

[0058] In addition, it should be noted that: Figure 7As shown, the scheduling process in this application can also be triggered at fixed intervals, and also supports event-driven triggering. Event-driven triggering scenarios include: new task arrival, vehicle departure or platform cancellation of task, robot reporting of abnormal events, and task timeout approaching. The edge scheduling server reads the current state version from the global state set to form the state baseline for this scheduling decision. This snapshot contains the latest state of all objects, including the task set, robot set, fixed charging position state, and narrow channel occupancy time window table. Based on the state snapshot X(k), the set of tasks to be scheduled and the set of available robots are obtained through filtering. The tasks to be scheduled include newly arrived tasks and tasks that need to be rescheduled due to abnormalities.

[0059] S103: Based on the determination result of the occupancy conflict and the set of available robots, generate a scheduling strategy corresponding to the charging task, perform atomic change operations based on the scheduling strategy, lock the state of the shared resources, and issue execution instructions to the corresponding embodied charging robot.

[0060] After obtaining the resource conflict determination result and the set of available robots, a scheduling strategy corresponding to the charging task is generated, using the conflict determination result and the set of available robots as constraints. Atomistic change operations are then executed according to the scheduling strategy to lock the shared resource state and issue execution instructions to the corresponding embodied charging robot. In a certain scenario, such as... Figure 6 As shown, the multi-robot optimization scheduling module solves the multi-robot allocation and sequence decision in the rolling time domain, and outputs the dispatch and pre-occupancy plan. Specifically, it uses rolling optimization to solve the multi-robot task allocation and short sequence decision, that is, assuming the current set of tasks to be scheduled is... The available robot sets are .

[0061] Considering that the robot can connect a maximum of 10m neighbors Vehicles, define a set of serializable neighborhoods: ; Decision variables: A two-stage representation of allocation and sequence is used, defining binary variables: And define the task start service time (starting from entering Docking-Align) as .

[0062] Objective function: Primarily focusing on minimizing waiting time and maximizing completion rate, a weighted minimization objective is constructed:

[0063] in The timeout penalty function is used to approximate the hard constraints. This indicates priority weights, meaning VIP / urgent / normal correspond to different weights or different penalty coefficients. Explicitly introduce the robot's remaining work time to solve the problem of competition within the same distance; ReassignCost is used to suppress frequent reassignments.

[0064] That is, as Figure 6 As shown, the input layer gathers various constraints and basic data required for scheduling decisions, mainly including the following elements: the navigation module outputs the estimated arrival time of each robot to each task point, providing a basis for subsequent time window calculations. Regarding resource constraints, fixed charging positions are defined as non-concurrent resources, which can only be occupied by one task at a time; narrow passages are defined with a capacity cap=1, allowing only one robot to pass at a time. These resource constraints will be transformed into mutual exclusion constraints in the optimization model. The robot set provides the real-time status of each robot, including the current battery percentage, estimated remaining work time, and maximum number of consecutive orders. A current battery percentage > 20% is a prerequisite for robot participation in scheduling. The task set T provides the characteristic parameters of each task, including priority (e.g., VIP / urgent / normal), hard deadline window (deadline=20min), and service duration. These parameters will affect the weight setting and feasibility determination of the optimization objective. The optimization model layer uses a rolling optimization scheduler as its core, combined with... Figure 7 The solution is obtained using a mixed-integer programming method, employing a warm-start technique that utilizes the solution from the previous cycle as an initial feasible solution to accelerate convergence. After the optimization model is solved, the output layer generates a resource reservation table, which includes the reservation time windows for each fixed charging position and the reservation time windows for each narrow channel, used for subsequent resource locking and conflict coordination. The robot service sequence generates an ordered execution sequence of 1 to 3 tasks for each robot, clearly defining the execution order of each task. The task-robot allocation table establishes the correspondence between tasks and robots, specifying which robot performs each task. Combined with... Figure 7 As shown, conflict detection mainly includes two categories: first, checking for overlapping pre-occupancy times at the same charging position; and second, checking for overlapping occupancy time windows at the same channel. If a conflict is detected, conflict resolution is initiated; otherwise, a freeze window check is performed. For scheduling solutions with conflicts, a splitting constraint is added or the task execution order is adjusted, and the solution is resolved within a limited number of iterations. During the iterative solution process, the reassignment suppression threshold ε must be met; that is, reassignment is only accepted when the improvement in benefit of the new solution relative to the old solution exceeds the threshold ε, otherwise the original solution is maintained. A freeze window check is performed on scheduling solutions that pass conflict detection, based on the following reassignment prohibition constraint method.

[0065] Specifically, in one or more embodiments of this specification, a scheduling strategy corresponding to the charging task is generated based on the determination result of the occupancy conflict and the set of available robots, specifically including: Taking the embodied charging robots in the available robot set and the charging tasks as objects, decision variables are constructed to represent which task a robot will execute in its execution sequence. Based on the constructed decision variables, the edge scheduling server constructs an optimization objective function with the core objectives of minimizing task waiting time and maximizing task completion rate, combined with priority weights and the robot's remaining working time. The design of this objective function corresponds to the multi-machine scheduling modeling described above. Then, the edge scheduling server converts the occupancy conflict determination results obtained in S102 into mathematical programming feasibility constraints, thereby limiting that within the same time window, the same fixed charging position and the same passageway are occupied by only a single task, ensuring that the generated scheduling strategy meets the capacity and mutual exclusion requirements of shared resources. This feasibility constraint is first reflected in the mutual exclusion of fixed charging positions. For each fixed charging position, at most one task can occupy it within the same time period. Suppose a task is assigned to a fixed charging position, and its occupation time period is the time interval consisting of the expected start time and service duration. Then, for any two different tasks assigned to the same fixed charging position, their occupation time periods must not overlap. Furthermore, each narrow passage has a capacity of 1, allowing at most one robot to pass through at any given time. If a robot needs to pass through a narrow passage to perform a task, and its traversal time is the interval between the expected entry and exit time windows, then for any two different tasks using the same narrow passage, their traversal time periods must not overlap. In addition, for robots and tasks identified as having occupancy conflicts in S102, the feasibility constraint directly prohibits the allocation scheme, meaning the robot cannot undertake the task. This ensures the scheduling strategy is executable at the physical resource level, preventing task failures or system deadlocks due to resource contention from the outset.

[0066] Based on the dual-state mapping relationship, consistency constraints between task states and robot states are determined to ensure that the generated scheduling strategy conforms to the state machine's transition rules. These consistency constraints ensure that the scheduling strategy matches the task and robot states consistently. In other words, if a task is assigned to a robot, the task state machine must transition from the created state to the assigned state, and simultaneously, the robot state machine must transition from the idle or executing state to the executing state. For robots whose original state is executing, it must be ensured that their current number of tasks not reaches the maximum task-to-task capacity limit. This state mapping relationship must remain unchanged during the execution of the scheduling strategy until the task enters the next stage or an anomaly triggers a state change. Furthermore, each task can only be executed by a single robot; that is, the same task cannot be assigned to multiple robots simultaneously, and the number of tasks a robot can handle at any given time cannot exceed its maximum task-to-task capacity. This avoids inconsistencies such as tasks being assigned but not executed by the robot, or the robot having departed but the task state not being updated.

[0067] A reassignment prohibition constraint is set for tasks that have entered the execution phase. This constraint prevents non-abnormally triggered task reassignment during execution. Specifically, tasks in execution phases such as alignment, insertion, holding, confirmation, and charging are considered to have entered a frozen window. For tasks within the frozen window, their corresponding allocation decisions cannot be changed in subsequent scheduling cycles; reassignment of these tasks is prohibited. For non-frozen tasks not yet in the frozen window, a reassignment cost mechanism is introduced. The reassignment cost, as a term in the optimization objective function, measures the cost of changing the original allocation scheme. A reassignment operation is only performed when the improvement in benefit from the new solution relative to the old solution exceeds a preset threshold; otherwise, the original allocation scheme is maintained. This mechanism effectively suppresses frequent reassignments caused by minor disturbances, maintaining the stability of the scheduling system.

[0068] Under the combined constraints of feasibility, consistency, and reassignment prohibition, the edge scheduling server solves the optimization objective function periodically or in an event-driven manner using a rolling time-domain approach, and obtains the scheduling strategy through multi-stage conflict resolution. Specifically, a two-stage conflict resolution strategy is adopted to improve real-time performance. The first stage is the initial solution stage, where the optimization objective function and other constraints besides channel constraints are solved to obtain an initial solution, ignoring narrow channel conflicts. The second stage is the conflict detection and resolution stage, where channel conflicts are detected in the initial solution based on a channel occupancy time window table. Conflicting decision plans are then subject to non-overlapping constraints, and the solution is re-solved within a limited number of iterations. If the conflict cannot be completely eliminated within the number of iterations, a priority preemption rule is activated, allowing lower-priority tasks to relinquish channel resources to higher-priority tasks. Furthermore, a priority handling mechanism is introduced during the solution process. For high-priority tasks such as VIP and emergency tasks, the priority weight coefficient is increased in the objective function to give them greater weight in the optimization process. Stricter built-in deadlines are set for VIP tasks, such as reserving a safety margin based on the actual hard deadline. When high-priority tasks conflict with ordinary tasks for resources, high-priority tasks are allowed to preempt the reserved resources of ordinary tasks. After solving, a complete scheduling strategy is output, including a task allocation table (i.e., the correspondence between each task and the executing robot), the task execution sequence of each robot, the estimated start time of each task, the reserved time period for each fixed charging position, and the reserved time window for each narrow passage.

[0069] Specifically, in one or more embodiments of this application, atomic change operations are performed based on a scheduling strategy to lock the state of shared resources and issue execution instructions to the corresponding embodied charging robot, specifically including the following processes: The scheduling strategy is parsed to obtain the shared resource identifiers corresponding to the scheduling strategy, namely the identifiers of the fixed charging positions and narrow channels that need to be locked, the corresponding embodied charging robot identifiers, namely the robot assigned to perform the task, and the time window corresponding to the shared resource occupation.

[0070] Based on atomic transactions, the state of shared resources in the global state set is updated to occupied. The state of the corresponding embodied charging robot, based on a two-state mapping, is then transitioned to the executing state. That is, for a fixed charging position s, its mutual exclusion state is updated from idle or pre-occupied to occupied. If the pre-occupancy time window is after the current time, it can also be updated to pre-occupied, indicating that the resource has been pre-occupied but not yet used. The state update simultaneously records the task identifier and occupation time period for occupying the resource. For narrow channels, the occupation time window table is updated, adding a new occupation record containing the robot identifier r, the entry time window, and the exit time window. The updated occupation time window table must satisfy the capacity constraint, i.e., only one record within the same time period. The robot state transition is as follows: if the robot's original state was idle, its state is transitioned to executing, indicating that the robot has started executing a task. If the robot's original state was executing and the current number of orders has not reached the capacity constraint, the executing state is maintained, and its task list is updated, adding the new task to the execution sequence. The state machine of the corresponding charging task is transitioned from created to allocated, recording the identifier r of the allocated robot. This state update is synchronized with the robot's state transition, ensuring consistency in the state mapping between the task and the robot. It should also be noted that the edge scheduling server performs version verification before executing the above state changes.

[0071] If all the above state change operations are executed successfully and the version verification passes, then as follows: Figure 5 The atomic transaction is submitted as shown. After the atomic transaction is successfully submitted, the edge scheduling server generates an execution instruction and sends it to the corresponding embodied charging robot based on the corresponding embodied charging robot identifier. The execution instruction includes at least the following: task identifier id_t, used by the robot to identify the specific task; target location, i.e., the coordinates of the vehicle; designated charging position, i.e., the fixed charging position identifier corresponding to the task; task execution sequence, if it is a sequential single-task scenario, the position and order of the current task in the sequence must be specified; and estimated arrival time, which serves as a reference target for robot navigation.

[0072] If the atomic transaction fails to commit, the current state change is rolled back, and the scheduling policy generation is triggered again, which means returning to S103 to perform the update operation.

[0073] S104: Obtain the stage data reported by the corresponding embodied charging robot. If an anomaly is detected in the embodied charging robot based on the dual-state mapping relationship, trigger state rollback, release the locked shared resources, and execute the scheduling update corresponding to the abnormal event. The normal monitoring branch is responsible for handling the normal progress of the task. When the stage data reported by the robot matches the expected transition path of the task state machine, the system advances the task state along the normal monitoring branch, sequentially completing the alignment, insertion, gripping, confirmation, and charging stages until the task is completed. The abnormal monitoring branch is responsible for detecting and handling various abnormal situations. That is, the system continuously compares the stage data reported by the robot with the expected state transition path. When a deviation is detected or an abnormal event is received, the system enters the abnormal monitoring branch. The execution monitoring branch is responsible for tracking the overall execution progress of the task, including macro-level monitoring such as whether the expected arrival time has exceeded the timeout and whether the service duration is abnormal.

[0074] based on Figure 7 After the command shown is issued, the execution monitoring phase begins. (Combined with...) Figure 8 As shown, the monitoring phase is divided into normal monitoring, abnormal monitoring, and execution monitoring.

[0075] like Figure 7 The data shown is based on monitoring and is reported by the corresponding embodied charging robot, including navigation estimated arrival time, arrival events, docking stage events, and status change notifications for fine-grained operation stages such as alignment completion, insertion completion, gripping completion, and confirmation completion, as well as notifications of abnormal events. When an abnormal event is determined to exist, that is... Figure 8 When the anomaly monitoring branch is activated and an anomaly is detected, the edge scheduling server triggers a state rollback, atomically reverting the task state machine and robot state machine to their pre-anomaly states and releasing locked shared resources. After the state rollback and resource release are complete, the edge scheduling server sends the anomaly cause code back to the park platform, forming a closed-loop feedback between task status and cause code. After anomaly handling is completed, the edge scheduling server executes the scheduling update corresponding to the anomaly event, i.e., triggering local rescheduling. Specifically, the withdrawal / compensation closed loop is entered when the following anomalies are detected: vehicle departure / platform cancellation: task transfer has been cancelled, and charging position / channel pre-occupancy is released; if the robot is already on the road, a stop or reassignment command is issued. Timeout: when predicted... If the deadline has been reached, the task will be withdrawn due to timeout, returning a timeout reason code and triggering a partial rescheduling of the remaining tasks. Insufficient workspace: When the robot detects insufficient workspace, the task will pause and wait / backoff to a cancelled or directly cancelled state, returning a space-insufficient reason code and simultaneously executing a backoff to homing / standby strategy. After releasing resources, the queued tasks will be recalculated. Channel blockage: If channel reservations cannot be fulfilled or execution exceeds the threshold for blockage, channel blockage will be triggered. The task can then backoff to homing / standby and reorder channel tokens, withdrawing and reassigning if necessary. To improve interpretability, this invention incorporates the withdrawal reason code as part of the task's final state into the global state, which is subscribed to by the park platform, forming a closed-loop feedback of task state combined with the reason code.

[0076] Specifically, in one or more embodiments of this application, obtaining the phase data reported by the corresponding embodied charging robot, and triggering a state rollback to release the locked shared resources when it is determined that the embodied charging robot has an anomaly based on the dual-state mapping relationship, specifically includes the following steps: like Figure 5 The system receives real-time stage data reported by the embodied charging robot. This stage data includes at least: navigation estimated arrival time, arrival events, docking stage events, and abnormal event reports. This mechanism ensures the scheduling system can respond to environmental and state changes in real time. Then, it is combined with... Figure 4 The dual-state machine design shown compares the current task execution state of the stage data with the expected state transition path of the dual-state mapping relationship to determine the execution progress of the charging task and whether it conforms to the preset transition rules of the task state machine and the robot state machine. If it does not conform, an abnormal event occurs, triggering a state rollback. This is done by atomically rolling back the task state machine and the robot state machine to their pre-abnormal state based on a transaction mechanism, releasing the task's lock on the fixed charging position and passageway, and releasing the shared resources back to the global resource pool. It is also necessary to ensure that the state rollback and resource release are completed within the same cycle to avoid data inconsistency issues where the state has been rolled back but the resources are still occupied. In addition, it should be noted that for some explicit and serious faults in abnormal events, such as the robot reporting insufficient space, the edge scheduling server does not need to perform the above-mentioned state transition rule verification. It directly determines the abnormality based on the reported fault identifier and does not need to verify whether the current state conforms to the transition rules. In such cases, the server still needs to record a snapshot of the task state and a snapshot of the robot state at the moment the abnormality occurs as a target reference benchmark for subsequent rollback operations.

[0077] Specifically, in one or more embodiments of this application, performing the scheduling update corresponding to the abnormal event includes: like Figure 8The system identifies the anomaly types corresponding to abnormal events, including: VehicleLeft, UserCancel, Timeout, Insufficient Workspace, CorridorBlocked, RobotLowSOC, and RobotFault. Then, based on the anomaly type, such as... Figure 7 As shown Figure 8 As shown, the set of affected tasks to be scheduled is determined. Based on the released shared resources, the set of tasks to be scheduled, and the currently available robot set, local rescheduling is performed to obtain an updated scheduling strategy. Local rescheduling adopts the same rolling time-domain optimization method as in S103 above. In the optimization objective function, tasks rescheduled due to anomalies can be assigned higher priority weights or a rescheduling penalty coefficient can be introduced to ensure that they can be processed as soon as possible and avoid excessive service delays caused by anomalies. In terms of constraint handling, local rescheduling also follows feasibility constraints, state consistency constraints, and reassignment prohibition constraints. However, considering the urgency of the anomaly scenario, the restriction of the reassignment prohibition constraint can be appropriately relaxed, allowing necessary adjustments to tasks that have not yet entered the freeze window. At the same time, in terms of the solution algorithm, local rescheduling adopts the same two-stage conflict repair strategy as S103. Since the system state changes significantly after an anomaly event, multiple iterations may be required to obtain a feasible solution. At this time, the edge scheduling server dynamically adjusts the upper limit of the number of iterations according to the urgency of the anomaly event, achieving a balance between solution quality and real-time performance. After the local rescheduling is completed, the updated scheduling strategy is output, which includes at least the new allocation scheme for the affected tasks, the adjusted execution sequence of each robot, and the re-pre-occupancy time window for each shared resource.

[0078] After obtaining the updated scheduling policy, the atomic change operation is executed according to the updated scheduling policy to perform the scheduling update corresponding to the abnormal event. That is, the updated scheduling policy is parsed, the shared resource identifier, the corresponding embodied charging robot identifier, and the resource occupancy time window are extracted; the shared resource status is updated to occupied in an atomic transaction manner, and the corresponding robot status is migrated to the executing state according to the dual-state mapping relationship; after the atomic transaction is successfully committed, the execution instruction is generated and sent to the corresponding embodied charging robot; if the transaction fails to commit, it is rolled back and the scheduling is retried.

[0079] The methods also include: Figure 5 The cause code corresponding to the exception type is written to the global status and then returned to the park platform.

[0080] like Figure 9As shown in the diagram, this application provides a structural schematic of a dynamic scheduling device for a self-contained charging robot. Figure 9 As can be seen, in one or more embodiments of this application, a dynamic scheduling device for a body-mounted charging robot includes: At least one processor; and, A memory communicatively connected to the at least one processor; wherein, The memory stores instructions that can be executed by the at least one processor to enable the at least one processor to perform any of the methods described above.

[0081] like Figure 10 As shown in the diagram, this application provides a schematic diagram of a non-volatile storage medium structure. Figure 5 It is understood that, in one or more embodiments of this application, a non-volatile storage medium stores computer-executable instructions 1001, which are capable of executing any of the methods described above.

[0082] The various embodiments in this application are described in a progressive manner. Similar or identical parts between embodiments can be referred to mutually. Each embodiment focuses on describing the differences from other embodiments. In particular, the embodiments of apparatus, devices, and non-volatile computer storage media are basically similar to the method embodiments, and therefore described more simply; relevant parts can be referred to the descriptions of the method embodiments.

[0083] The foregoing has described specific embodiments of this application. Other embodiments are within the scope of the appended claims. In some cases, the actions or steps recited in the claims may be performed in a different order than that shown in the embodiments and may still achieve the desired results. Furthermore, the processes depicted in the drawings do not necessarily require the specific or sequential order shown to achieve the desired results. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0084] The above description is merely one or more embodiments of this application and is not intended to limit this application. For those skilled in the art, various modifications and variations can be made to one or more embodiments of this application. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principle of one or more embodiments of this application should be included within the scope of the claims of this application.

Claims

1. A dynamic scheduling method for an embodied charging robot, characterized in that, A dynamic scheduling system applied to a park platform, an edge scheduling server, and multiple embodied charging robots, wherein the embodied charging robots are mobile, automated embodied charging robots, and the method includes: The edge scheduling server constructs a dual-state mapping relationship between the task state machine and the robot state machine, and obtains the spatiotemporal feature data of each of the embodied charging robots arriving at the task point corresponding to the charging task; wherein, the charging task has time constraints. Based on the spatiotemporal feature data and the dual-state mapping relationship, the conflict of shared resource occupation is determined, and based on the real-time operating status of each of the embodied charging robots, a set of available embodied charging robots is obtained; wherein, the shared resources include at least: fixed charging positions and passageways; Based on the determination result of the occupancy conflict and the set of available robots, a scheduling strategy corresponding to the charging task is generated. Based on the scheduling strategy, atomic change operations are performed to lock the state of the shared resources and issue execution instructions to the corresponding embodied charging robot. If the phase data reported by the corresponding embodied charging robot is obtained, and an anomaly is determined based on the dual-state mapping relationship, a state rollback is triggered, the locked shared resources are released, and the scheduling update corresponding to the anomaly event is executed.

2. The dynamic scheduling method for a body-mounted charging robot according to claim 1, characterized in that, A dual-state mapping relationship between the task state machine and the robot state machine is constructed, and spatiotemporal feature data of the task points corresponding to the arrival of each of the embodied charging robots at the charging task are obtained, specifically including: Based on the predefined dual-state mapping rules between the task state machine and the robot state machine, the matching correspondence between the task state and the executing robot state at any given time is determined, and the matching correspondence is used as the dual-state mapping relationship. The real-time operating data reported by each of the aforementioned embodied charging robots is obtained; wherein, the real-time operating data includes at least: the real-time position and motion vector obtained based on the built-in navigation module of the embodied charging robot; Based on the real-time operation data and the current park map data, the spatiotemporal characteristic data of each of the embodied charging robots arriving at the task point corresponding to the charging task are determined; wherein, the acquisition of the spatiotemporal characteristic data includes: the estimated arrival time and the channel traversal time window.

3. The dynamic scheduling method for a body-mounted charging robot according to claim 1, characterized in that, Based on the spatiotemporal feature data and the dual-state mapping relationship, the determination of shared resource occupancy conflicts specifically includes: Based on the pre-defined strong consistency principle, the mutual exclusion state and occupancy sequence of the shared resources are obtained, and a baseline state context is constructed based on the mutual exclusion state and the occupancy sequence. The spatiotemporal feature data is analyzed to generate a time-series model of resource consumption for the charging task; By comparing the baseline state context with the resource occupancy time series model, a subset of suspected conflicts with temporal overlap is obtained based on capacity and mutual exclusion constraints; Based on the dual-state mapping state, the suspected conflict subset is filtered to obtain the valid conflict subset; The shared resources located in the subset of valid conflicts are determined to have an occupancy conflict.

4. The dynamic scheduling method for a body-mounted charging robot according to claim 1, characterized in that, Based on the determination result of the occupancy conflict and the set of available robots as constraints, a scheduling strategy corresponding to the charging task is generated, specifically including: Taking the embodied charging robot in the available robot set and the charging task as objects, decision variables are constructed, and with the goal of minimizing task waiting time and maximizing completion rate, an optimization objective function is constructed by combining priority weights and the robot's remaining working time. The determination result of the occupation conflict is converted into a feasibility constraint; wherein, the feasibility constraint is used to limit the same fixed charging position and the same passage to be occupied by only a single task within the same time window; Based on the dual-state mapping relationship, a consistency constraint between the task state and the robot state is determined; wherein, the consistency constraint is used to constrain the scheduling strategy to satisfy the matching consistency between the task state and the robot state. Set a reassignment prohibition constraint for tasks that have entered the execution phase; wherein, the reassignment prohibition constraint is used to prohibit non-abnormally triggered task reassignment during the execution process; Based on a rolling time-domain approach, either periodic or event-driven, the optimization objective function is solved under the constraints of feasibility, consistency, and rescheduling prohibition, and the scheduling strategy is obtained through multi-stage conflict resolution.

5. The dynamic scheduling method for a body-mounted charging robot according to claim 1, characterized in that, Based on the scheduling strategy, atomic change operations are performed to lock the state of the shared resources and issue execution instructions to the corresponding embodied charging robot, specifically including: The scheduling strategy is parsed to obtain the shared resource identifier, the corresponding embodied charging robot identifier, and the time window corresponding to the shared resource occupation. Based on the atomic transaction method, the state of the shared resource in the global state set is updated to occupied, and the state of the corresponding embodied charging robot is migrated to the executing state based on the dual state mapping relationship. After the atomic transaction is successfully submitted, an execution instruction is generated and sent to the corresponding embodied charging robot based on the corresponding embodied charging robot identifier; If the atomic transaction commit fails, the current state change is rolled back, and the scheduling policy generation is retried.

6. The dynamic scheduling method for a body-mounted charging robot according to claim 1, characterized in that, Obtain the phase data reported by the corresponding embodied charging robot. Based on the dual-state mapping relationship, if an anomaly is determined in the embodied charging robot, trigger a state rollback and release the locked shared resources. Specifically, this includes: The system receives real-time stage data reported by the embodied charging robot; wherein the stage data includes at least: estimated navigation arrival time, arrival event, docking stage event, and abnormal event; By comparing the current task execution state of the stage data with the expected state transition path of the dual-state mapping relationship, the execution progress of the charging task is determined, and whether it conforms to the preset transition rules of the task state machine and the robot state machine. If not, an abnormal event occurs, triggering a state rollback. The task state machine and the robot state machine are then rolled back to their pre-abnormal state in an atomic manner, and the locked shared resources are released.

7. The dynamic scheduling method for a body-mounted charging robot according to claim 1, characterized in that, Executing the scheduling update corresponding to the aforementioned abnormal event specifically includes: Identify the anomaly type corresponding to the abnormal event, and determine the set of affected scheduled tasks based on the anomaly type; Based on the released shared resources, the set of tasks to be scheduled, and the set of currently available robots, a local rescheduling is performed to obtain an updated scheduling strategy. Based on the updated scheduling policy, return to execute the atomic change operation to perform the scheduling update corresponding to the abnormal event; The method further includes: writing the cause code corresponding to the exception type into the global state, and returning the cause code to the park platform.

8. The dynamic scheduling method for a body-mounted charging robot according to claim 1, characterized in that, Before the edge scheduling server constructs a dual-state mapping relationship between the task state machine and the robot state machine, and obtains the spatiotemporal feature data of each of the embodied charging robots arriving at the task point corresponding to the charging task, the method includes: The edge scheduling server receives charging tasks issued by the park platform based on an event-driven interface; Perform a field completeness check on the charging task to determine whether the charging task has any required fields; If not, the charging task will be rejected and a verification failure response will be returned to the park platform; If so, the spatiotemporal feature data is acquired in response to the charging task.

9. A dynamic scheduling device for an embodied charging robot, characterized in that, A dynamic scheduling system applied to a park platform, an edge scheduling server, and multiple embodied charging robots, wherein the embodied charging robots are mobile, automated embodied charging robots, and the equipment includes: At least one processor; and, A memory communicatively connected to the at least one processor; wherein, The memory stores instructions that can be executed by the at least one processor to enable the at least one processor to perform the method described in any one of claims 1-8.

10. A non-volatile storage medium storing computer-executable instructions, characterized in that, The computer-executable instructions are capable of performing the method described in any one of claims 1-8.

Images