Hospital multi-scene transportation robot hybrid scheduling method based on dynamic priority

Abstract

The invention discloses a hospital multi-scene transportation robot hybrid scheduling method based on dynamic priority, and belongs to the technical field of robot control and scheduling. The method comprises the following steps: acquiring hospital dynamic environment data; receiving and analyzing the multi-mode instruction to form task parameters; preliminarily grading according to task parameters and preset rules, and dynamically adjusting task priorities in combination with a dynamic map and the current state of the robot; based on the adjusted task priority and robot type, distributing robots and generating an initial path considering distance and people flow density; the robot is driven to track along the path trajectory, and obstacle behaviors are predicted in real time and dynamically avoided; decomposing a transportation task into task chains, and when the electric quantity is lower than a threshold value, deciding whether to insert a charging task in combination with the priority; and after the task is completed, prompting and transmitting feedback data. According to the invention, dynamic priority, behavior prediction and hybrid scheduling are adopted, intelligent distribution and path optimization of transportation tasks are realized, and the efficiency and safety of internal logistics of hospitals are improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of robot control and scheduling, in particular to a hospital multi-scene transportation robot hybrid scheduling method based on dynamic priority. BACKGROUND

[0002] The hospital multi-scene transportation robot is an intelligent mobile device applied in medical institutions for automatically transporting various materials such as medicines, specimens, meals, and bedding. This kind of robot usually integrates functions such as autonomous navigation, environment perception, and task management, aiming to reduce the non-core labor burden of medical staff and improve the efficiency and accuracy of hospital logistics support. The core technology lies in scheduling control, which is responsible for receiving transportation instructions, allocating robots, planning paths, and controlling robots to safely and efficiently complete transportation tasks in complex hospital environments.

[0003] The existing hospital transportation robot scheduling method mostly adopts a scheduling strategy based on fixed rules. When receiving a transportation task, the robot is allocated and the route is planned according to the preset priority and simple path algorithm. Although some advanced robots have basic obstacle detection and avoidance capabilities, they usually react only after encountering obstacles, and their path planning mainly relies on static building maps, which do not take into account dynamic changes in the environment such as human flow.

[0004] The above existing technology has obvious defects in actual application. The fixed priority cannot adapt to the characteristics of frequent emergency situations in hospitals, which may cause a sudden and more urgent task to wait for a long time due to priority rule restrictions. The shortest path planning based on static maps often leads the robot to enter areas with dense human flow, resulting in frequent stopping and detours, greatly reducing the actual transportation efficiency. Different types of robots are often independently scheduled, lacking coordination, causing resource waste. These defects collectively result in low overall operating efficiency and limited intelligence level of existing robots in complex and dynamic hospital scenarios. SUMMARY

[0005] To solve the above problems, the present application provides a hospital multi-scene transportation robot hybrid scheduling method based on dynamic priority, which combines multi-modal instruction analysis, dynamic priority adjustment, heterogeneous robot collaborative scheduling, and predictive path planning, and can realize intelligent transportation task response, dynamic optimization of path selection, and long-term high reliability of operation.

[0006] The above objectives can be achieved through the following scheme: The application discloses a dynamic priority-based hospital multi-scene transportation robot hybrid scheduling method, which comprises the following steps: acquiring hospital multi-scene environment information and transportation robot types, generating environment data containing a dynamic map and available robot types; receiving and analyzing multi-modal instructions, extracting task parameters containing to-be-transported materials, quantity, starting point and destination, wherein the multi-modal instructions are different modes of instructions that robots can receive and understand from voice, manual input and application programs; preliminarily classifying according to the task parameters and preset scheduling rules, obtaining an initial task priority, and dynamically adjusting the initial task priority in combination with the dynamic map and the current state of the transportation robot, to generate an adjusted task priority; based on the adjusted task priority and the available robot types, assigning a transportation robot from a robot queue, and generating an initial path that takes into account short distance and small human flow density according to the dynamic map; driving the transportation robot to track the initial path, and predicting the behavior of obstacles in the process of advancing in real time, dynamically generating and executing avoidance actions according to the prediction results to form an adjusted path; decomposing the transportation task into a task chain containing several links, and monitoring the power of the transportation robot in real time during the execution of the task chain, when the power is lower than the power threshold, deciding whether to insert a charging task in combination with the adjusted task priority of the current task, to obtain an updated task chain; after the transportation robot completes the updated task chain, prompting task completion, and collecting transportation process data to form feedback data, and transmitting the feedback data to a central scheduling unit.

[0007] Optionally, the acquiring hospital multi-scene environment information and transportation robot types, and generating environment data containing a dynamic map and available robot types comprises: collecting static area information and dynamic information of the hospital through a multi-sensor fusion technology; constructing a multi-layer structure dynamic map based on the static area information and dynamic information, wherein the dynamic map contains a static map base layer and a personnel density heat map layer and a device state layer superimposed thereon; identifying the transportation robot types, and combining the dynamic map and the identified transportation robot types to generate the environment data.

[0008] Optionally, the receiving and analyzing multi-modal instructions, and extracting task parameters containing to-be-transported materials, quantity, starting point and destination comprises: receiving multi-modal instructions from voice, manual input or application programs; analyzing the multi-modal instructions by using a natural language processing technology, and extracting keywords of to-be-transported materials, quantity, starting point and destination; performing real-time feasibility verification on the starting point and destination in the keywords by using the dynamic map, and after the verification passes, encapsulating the keywords to form the task parameters.

[0009] Optionally, the initial task priority is obtained by preliminary classification according to the task parameters and preset scheduling rules, and the initial task priority is dynamically adjusted in combination with the dynamic map and the current state of the transport robot to generate an adjusted task priority, including: matching the task parameters according to the scheduling rules for quantifying the importance of the material and the destination point to obtain the initial task priority; obtaining the current state of each transport robot from the robot queue, wherein the current state includes a power value and position information; constructing a comprehensive weighting model, taking the initial task priority, an urgency attenuation factor, the dynamic map and the current state of the transport robot as inputs, performing weighted operation, and outputting the adjusted task priority, wherein the urgency attenuation factor is a parameter for quantifying and adjusting the task priority, and the role is to prevent a certain high-priority task from occupying a high priority for a long time, ignoring other tasks that may also become urgent.

[0010] Optionally, the transport robot is allocated from the robot queue based on the adjusted task priority and the available robot type, and an initial path that takes into account short distance and small crowd density is generated according to the dynamic map, including: filtering candidate robot types from the transport robot types according to the material in the task parameters; calculating the expected travel time of the available robots of the candidate robot types to the task starting point, and selecting the robot with the shortest expected travel time as the allocated transport robot; using a path planning algorithm that weights and sums the physical length of the path and the crowd density in the dynamic map as the travel cost to calculate a path with the lowest total travel cost for the allocated transport robot as the initial path.

[0011] Optionally, the transport robot is driven to track the initial path, and the behavior of obstacles in the travel process is predicted in real time, and the adjusted path is generated and executed according to the prediction results, including: using a model predictive control algorithm to make the allocated transport robot accurately travel along the initial path, and outputting the real-time tracking state of the robot; when the sensor of the robot detects an obstacle, using a fuzzy control algorithm to fuzz the sensor input and perform fuzzy reasoning to obtain a behavior prediction result of the future movement intention of the obstacle; taking the behavior prediction result as a dynamic constraint and feeding it back to the model predictive control algorithm for trajectory re-planning to generate and execute an avoidance action, and the trajectory set of the avoidance action constitutes the adjusted path.

[0012] Optionally, the task is decomposed into a task chain containing several links, and the power of the transport robot is monitored in real time during the execution of the task chain, and when the power is lower than the power threshold, the adjusted task priority of the current task is combined to decide whether to insert a charging task, and an updated task chain is obtained, including: decomposing the complete transport task into ordered links of picking up goods, taking the elevator, transporting, and delivering goods to form a task chain; continuously monitoring the real-time power value of the transport robot and comparing it with the power threshold; when the real-time power value is lower than the power threshold, it is judged whether the adjusted task priority of the current task is high priority, if yes, a charging task is inserted after the completion of the current task chain, if no, the current task chain is paused and a charging task is immediately inserted to form an updated task chain.

[0013] Optionally, after the transport robot completes the updated task chain, the task completion is prompted, the transport process data is collected to form feedback data, and the feedback data is transmitted to the central scheduling unit, including: issuing a task completion prompt to the destination person through voice and light; integrating the path deviation, link time consumption, and abnormal events recorded during the execution of the transport task to form structured transport process data; packaging the transport process data as feedback data and sending it to the central scheduling unit for subsequent optimization of the scheduling rules.

[0014] Optionally, the method further comprises: periodically refreshing the dynamic map in the environment data with the latest sensor acquisition information to maintain the real-time accuracy of the dynamic map; according to the feedback data received by the central scheduling unit, using a machine learning algorithm to adaptively adjust the generation logic of the scheduling rules and the initial task priority; integrating environment perception, task analysis, scheduling planning, behavior prediction, and emergency handling into a closed-loop control logic of perception, decision, execution, and learning.

[0015] Based on the same inventive concept, the application also provides a dynamic priority-based hospital multi-scene transportation robot hybrid scheduling system, which comprises: an environment perception module for acquiring hospital multi-scene environment information and transportation robot types, and generating environment data containing a dynamic map and available robot types; a task analysis module for receiving and analyzing multi-modal instructions, and extracting task parameters containing transported materials, quantity, starting point and destination, wherein the multi-modal instructions are different modes of instructions that robots can receive and understand from voice, manual input and application; a priority management module for preliminarily classifying according to the task parameters and preset scheduling rules, obtaining an initial task priority, and dynamically adjusting the initial task priority in combination with the dynamic map and the current state of the transportation robot to generate an adjusted task priority; a scheduling planning module for assigning transportation robots from a robot queue based on the adjusted task priority and the available robot types, and generating an initial path that takes into account short distance and small crowd density according to the dynamic map; a trajectory tracking and intervention module for driving the transportation robot to track the initial path, and predicting the behavior of obstacles in the process of advancing in real time, dynamically generating and executing avoidance actions according to the prediction results to form an adjusted path; an emergency handling module for decomposing the transportation task into a task chain containing several links, and monitoring the power of the transportation robot in real time during the execution of the task chain, deciding whether to insert a charging task in combination with the adjusted task priority of the current task when the power is lower than a power threshold, and obtaining an updated task chain; and an information feedback module for prompting task completion after the transportation robot completes the updated task chain, collecting transportation process data to form feedback data, and transmitting the feedback data to a central scheduling unit.

[0016] Compared with the prior art, the application has the following advantages: 1. The application constructs a multi-layer dynamic map containing real-time dynamic information, and deeply couples it with task priority adjustment, robot allocation and path planning, thereby improving the intelligence and context awareness of scheduling decisions. The task execution order and robot travel route can be optimized in real time according to the changing crowd density and equipment state inside the hospital, avoiding path congestion and resource mismatch caused by information lag in traditional scheduling methods, so that the entire transportation system has strong adaptability to complex dynamic environments and high running efficiency.

[0017] 2、The application proposes a hybrid priority management mechanism combining static rules with dynamic adjustment, and introduces human-machine behavior prediction and flexible interaction strategy. Not only can the initial classification be made according to the urgency of the task itself, but also dynamic adjustment can be made in combination with task waiting time, environmental congestion and robot state, to ensure that the most critical task can always be responded in time. At the same time, by predicting the moving intention of pedestrians and actively taking measures to avoid or interact, the robot can pass through the dense area more smoothly and safely, and the collaborative work level in the human-machine integration environment is improved.

[0018] 3、The application designs a task chain decomposition and closed-loop emergency handling process, which enhances the reliability and continuity of the robot transportation system. By decomposing complex tasks into fine links for monitoring and establishing an intelligent charging decision mechanism based on task priority, the energy state of the robot can be managed, and the task interruption caused by power consumption can be effectively prevented. The data collection and feedback throughout the process further constitute a closed loop for self-learning optimization, which enables continuous improvement of the scheduling strategy from historical experience, realizes self-iteration and improvement of performance, and ensures long-term stable and efficient operation.

[0019] Other features and advantages of the present application will be set forth in the following description, and in part will become apparent to those skilled in the art from the description, or can be learned by practice of the application. The objects and other advantages of the present application can be realized and attained by the structure particularly pointed out in the description, claims and drawings. BRIEF DESCRIPTION OF DRAWINGS

[0020] In order to more clearly illustrate the technical solutions in the embodiments of the present application or the prior art, the following will briefly introduce the drawings needed to be used in the embodiment or prior art description. Obviously, the drawings in the following description are some embodiments of the present application, and for those skilled in the art, other drawings can also be obtained without creative labor on the basis of these drawings.

[0021] Figure 1 is a process schematic diagram of a hospital multi-scene transportation robot hybrid scheduling method based on dynamic priority according to an embodiment of the present application.

[0022] Figure 2 is a path planning and personnel density heat map according to an embodiment of the present application.

[0023] Figure 3 is a speed trajectory line diagram of robot obstacle avoidance according to an embodiment of the present application.

[0024] Figure 4 is a structure schematic diagram of a hospital multi-scene transportation robot hybrid scheduling system based on dynamic priority according to an embodiment of the present application. DETAILED DESCRIPTION

[0025] In order to make the objects, technical solutions and advantages of the embodiments of the present application clearer, the technical solutions in the embodiments of the present application will be described clearly and completely below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are some but not all of the embodiments of the present application. Based on the embodiments in the present application, all other embodiments obtained by a person of ordinary skill in the art without creative work fall within the protection scope of the present application.

[0026] With reference to Figure 1 One embodiment of the present application proposes a dynamic priority-based hospital multi-scene transportation robot hybrid scheduling method, which adopts a technical solution combining multi-modal instruction analysis, dynamic priority adjustment, heterogeneous robot collaborative scheduling and predictive path planning. The method can realize intelligent response of transportation tasks, dynamic optimization of path selection and long-term high reliability of operation, and improve the efficiency and accuracy of hospital internal logistics.

[0027] The method of the embodiment specifically includes: Obtaining hospital multi-scene environment information and transportation robot types, and generating environment data containing a dynamic map and available robot types; Receiving and analyzing multi-modal instructions to extract task parameters containing transported materials, quantity, starting point and destination, wherein the multi-modal instructions are different modes of instructions that robots can receive and understand from voice, manual input and application programs; Preliminary classification is performed according to the task parameters and preset scheduling rules to obtain an initial task priority, and the initial task priority is dynamically adjusted in combination with the dynamic map and the current state of the transportation robot to generate an adjusted task priority; Based on the adjusted task priority and the available robot types, a transportation robot is allocated from a robot queue, and an initial path is generated according to the dynamic map, which takes into account short distance and small human flow density; The transportation robot is driven to track the initial path, and the behavior of obstacles in the travel process is predicted in real time, and avoidance actions are dynamically generated and executed according to the prediction results to form an adjusted path; The transportation task is decomposed into a task chain containing several links, and the power of the transportation robot is monitored in real time during task chain execution. When the power is lower than a power threshold, it is decided whether to insert a charging task in combination with the adjusted task priority of the current task to obtain an updated task chain; Specifically, a real-time updated dynamic map is constructed through multi-source sensor fusion technology. Natural language processing technology is used to analyze diversified user instructions and convert them into standardized task parameters. The core innovation lies in a double-layer priority decision mechanism, which first classifies tasks preliminarily according to preset rules, and then dynamically adjusts them according to real-time dynamic information and the state of the robot itself, ensuring that the task priority accurately reflects the urgency and feasibility at the moment. According to the characteristics and priority of the task, the most suitable performer is intelligently allocated from multiple types of robots, and an initial path is planned considering the distance and dynamic environmental risk. Combining the cutting-edge algorithms of model predictive control and fuzzy control, the robot can not only accurately track the path, but also predict and flexibly respond to pedestrians and obstacles. The task chain management and emergency handling mechanism ensures the continuity and robustness of long-time tasks, and intelligently determines the charging time. The data feedback after the completion of the task forms a learning closed loop, which can self-optimize the scheduling rules according to the historical execution effect, realizing continuous performance iteration.

[0028] After the transportation robot completes the updated task chain, it prompts the completion of the task and collects the transportation process data to form feedback data, which is transmitted to the central scheduling unit.

[0029] Optionally, the acquisition of hospital multi-scene environment information and transportation robot types, and the generation of environment data including dynamic map and available robot types include: The static area information and dynamic information of the hospital are collected through multi-sensor fusion technology; Specifically, laser radar and pre-imported building blueprints are used to construct static area information including rooms, corridors, elevators and other locations and physical dimensions, forming a basic static map. Through the camera network covering each floor and Wi-Fi signal detection equipment, real-time dynamic information is collected. Among them, the camera uses computer vision algorithm to identify crowd flow, but does not identify individual identity, only counts the number of personnel in the area, so as to calculate the personnel density; Wi-Fi detection assists in judging the degree of crowd density by monitoring the strength and number of mobile device signals. For movable devices, low-power Bluetooth or RFID tags attached to them are used for positioning to obtain their real-time device location information.

[0030] Based on the static area information and dynamic information, a multi-layer structure dynamic map is constructed, wherein the dynamic map includes a static map base layer and a personnel density heat map layer and a device state layer superimposed thereon; Specifically, the collected static area information and dynamic information are fused to construct a multi-layer structure dynamic map. The dynamic map takes the static map as the basic layer, and two key dynamic data layers are superimposed thereon. The first is the personnel density heat map layer. The personnel density heat map layer renders the real-time calculated personnel density data in the form of a visual heat map on the map, and different colors represent different congestion levels, providing an obstacle avoidance basis for subsequent path planning. The second is the device status layer. The device status layer marks the real-time acquired device position information on the map and displays its current state, such as idle or in use.

[0031] Identifying the type of transport robot and combining the dynamic map with the identified type of transport robot to generate environment data.

[0032] Specifically, the type of the accessed transport robot is identified. Each robot registers its unique identifier and preset robot type with the central dispatching unit when it is started. These information is integrated to finally generate a structured environment data package. The environment data not only contains a multi-layer dynamic map, but also includes the type, quantity and state list of the currently available transport robots. Figure 2 As shown, it is described how the path planning algorithm of the present application plans an optimal initial path for the robot that takes into account both distance and passage risk according to the real-time personnel density heat map.

[0033] Optionally, the receiving and parsing of the multi-modal instruction includes extracting task parameters including the to-be-transported material, quantity, starting point and destination point. Receiving multi-modal instructions from voice, manual input or application; Specifically, through the deployment of microphones, touch screens and special applications on the hospital nurse station, mobile devices or robot bodies, transportation requests from medical staff are received, which are multi-modal instructions. For voice instructions, an automatic speech recognition engine is called to convert continuous speech signals into text strings. For manual input and application instructions, the text input by the user in the graphical interface or the selected options are directly obtained.

[0034] The multi-modal instruction is parsed using natural language processing technology to extract the keywords of the to-be-transported material, quantity, starting point and destination point; Specifically, all instruction texts from various sources are uniformly sent to natural language processing. Natural language processing is a technology that enables computers to understand and interpret human language. First, intent recognition is performed to determine whether the core purpose of the instruction is material transportation. After confirming the intent, the named entity recognition technology is used to automatically scan and extract keywords of predefined categories from the text. These keywords accurately correspond to the core elements of the transportation task, including the material to be transported, such as "blood samples" or "sterile surgical packs"; the quantity, such as "three boxes" or "five portions"; the starting point, such as "the third floor of the laboratory"; and the destination, such as "the fifth operating room".

[0035] The starting point and destination in the keywords are verified for real-time feasibility using the dynamic map, and after verification, the keywords are encapsulated to form task parameters.

[0036] Specifically, the extracted keywords do not immediately form a task, but enter a critical real-time feasibility verification link. The extracted starting point and destination information are matched and checked in real time with the constructed dynamic map. The verification process queries the dynamic map to determine whether the target location is currently accessible, such as checking whether the relevant area is temporarily closed due to disinfection, or whether the corridor leading to the area has too high a density of personnel, or even whether the key elevator equipment is in a state of repair. Only when both the starting point and the destination are confirmed to be currently accessible does the verification pass. After verification, these checked keywords are encapsulated into a standardized data structure to form the final task parameters.

[0037] Optionally, according to the task parameters and the preset scheduling rules, an initial task priority is obtained, and the initial task priority is dynamically adjusted in combination with the dynamic map and the current state of the transportation robot to generate an adjusted task priority, including: According to the scheduling rules that quantify the importance of the material and the destination, the task parameters are matched to obtain an initial task priority; Specifically, after receiving the generated task parameters, an initial task priority is assigned to the task according to a preset rule base. The rule base quantifies the nature of the material to be transported and the importance of the destination, for example, emergency blood samples transported to the operating room are defined as the highest priority P0 level; tasks for transporting regular medicines are P1 level; tasks for transporting medical records or test sheets are P2 level; and tasks for transporting clean uniforms or handling medical waste are the lowest P3 level. This stage assigns a basic, static priority to each task.

[0038] The current state of each transportation robot in the robot queue is obtained, wherein the current state includes a power value and location information; Specifically, real-time dynamic information is continuously obtained from environmental data, especially the personnel density on the target path, and the current state of each robot in the heterogeneous robot queue is polled in real time, and the key state parameters include the accurate power value and position information of the robot.

[0039] A comprehensive weighting model is constructed, and the initial task priority, urgency decay factor, dynamic map, and current state of the transport robot are taken as inputs for weighting operation, and the adjusted task priority is output, wherein the urgency decay factor is a parameter for quantifying and adjusting the task priority, and the role is to prevent a certain high-priority task from occupying a high priority for a long time, ignoring other tasks that may also become urgent.

[0040] Specifically, a comprehensive weighting model is used to dynamically calculate and update the final priority of the task. The calculation method of the model is: , wherein, is the final output of the adjusted task priority score. is the initial priority baseline score determined by the preliminary classification. is the time that the task has been waiting since it is generated, is a non-linear growth function representing the urgency decay factor, ensuring that the priority of a low-priority task that has been waiting for a longer time will be dynamically improved over time, preventing task starvation. is the average personnel density on the optimal initial path from the starting point to the destination point calculated according to the dynamic map. is a function that maps high personnel density to a higher positive adjustment value, meaning that tasks that pass through crowded areas need higher priority to ensure timely execution. is the power value of the robot that is currently most suitable for executing the task. is a function that maps low power values to negative adjustment values, i.e. when the available robot power is generally low, the priority of non-urgent tasks will be appropriately reduced, giving priority to high-power robots to perform critical tasks or charge. , is the weight coefficient of each influencing factor, which is preset by the actual operation demand of the hospital or optimized through historical data learning. Among them, , , are normalization functions that convert inputs of different physical dimensions into unified, dimensionless priority adjustment scores, thereby ensuring the mathematical logic rigor of the formula. Through this model, the dynamic score that comprehensively considers the importance of the task itself, waiting time, environmental congestion, and the state of the robot itself, i.e. the adjusted task priority, is finally calculated.

[0041] Optionally, the assigning a transport robot from the robot queue based on the adjusted task priority and the available robot types, and generating an initial path for the transport robot according to the dynamic map, which takes into account both a short distance and a small crowd density, comprises: According to the material in the task parameter, a candidate robot type is selected from the transport robot type; Specifically, a task with the highest adjusted task priority is received from the priority management. The robot allocation program is immediately started. The first step is type matching, which selects candidate robots that meet the conditions from the list of transport robots contained in the environmental data according to the characteristics of the material to be transported in the task parameter. For example, if the task is to transport large medical equipment, only high-load robots are selected to form the candidate pool; if it is a medicine delivery, a delivery robot with a cabinet door is selected.

[0042] The expected travel time of the available robots of the candidate robot type to the task starting point is calculated, and the robot with the shortest expected travel time is selected as the allocated transport robot; Specifically, after determining the candidate robot type pool, the optimal individual selection stage is entered. The core goal of this stage is to select the robot that can respond to the task the fastest. Each robot in the candidate pool is traversed to obtain the current real-time position information, and the expected travel time of the robot from the current position to the task starting point is calculated based on the dynamic map. The travel time not only considers the physical distance, but also takes into account dynamic factors such as the density of personnel on the path. The robot with the shortest expected travel time is selected and officially marked as the allocated transport robot.

[0043] An improved path planning algorithm that sums the physical length of the path and the personnel density in the dynamic map as the travel cost is used to calculate a path with the lowest total travel cost for the allocated transport robot as the initial path.

[0044] Specifically, once the robot is allocated, an initial path is immediately generated for it. The path planning algorithm takes the dynamic map as input and uses an improved path planning algorithm to find the optimal path from the task starting point to the destination point. The improved path planning algorithm is an algorithm that finds the lowest cost path on a graph plane with multiple nodes. The cost function of the algorithm is specially designed to combine distance and environmental risk, and its calculation method is: , Wherein, represents the travel cost of any path on the map. is the physical length of the path, which is obtained from the static layer of the dynamic map. is the real-time personnel density on the path, which is obtained from the personnel density heat map layer of the dynamic map. is a nonlinear function for mapping the people density data into risk cost, the higher the crowdedness, the higher the cost. and is a preset weight coefficient for balancing the importance of the shortest distance and avoiding crowds. For high-priority tasks, the weight of can be dynamically increased to make the path planning more inclined to choose a clear path that may be slightly longer. The algorithm will find a path that minimizes the total distance. After generating the path, a final check is performed to check whether the path passes through the temporary control areas marked on the dynamic map, such as rooms being disinfected. After confirmation, the path consisting of a series of coordinate points is determined as the initial path and is output to the trajectory tracking together with the assigned transport robot information.

[0045] Optionally, the driving transport robot performs trajectory tracking along the initial path and predicts the behavior of obstacles in real time during the travel, dynamically generates and executes avoidance actions according to the prediction results, and forms an adjusted path, including: using a model predictive control algorithm to make the assigned transport robot accurately travel along the initial path and output the real-time tracking state of the robot; Specifically, model predictive control is an advanced control strategy that can predict the possible motion trajectories of the robot within a small period of time based on the current state of the robot and the kinematic model at each control period. The optimal one is selected from these trajectories by optimizing the cost function, and the goal of the cost function is to minimize the deviation between the actual position of the robot and the reference points of the initial path. The robot only executes the first control instruction of the optimal trajectory, and then repeats this process at the next control period. This rolling optimization process enables the robot to smoothly and accurately travel along the initial path, thereby generating a real-time tracking state.

[0046] When the robot's sensor detects an obstacle, a fuzzy control algorithm is used to fuzz the sensor input and perform fuzzy reasoning to obtain a behavior prediction result of the future movement intention of the obstacle; Specifically, in the process of trajectory tracking, the robot continuously perceives the surrounding environment through the sensors mounted thereon, especially pedestrians or other moving obstacles. Once an obstacle is detected to enter a preset safety area, a fuzzy control algorithm is activated for behavior prediction. Fuzzy control is an intelligent control method based on fuzzy set theory, which is particularly suitable for handling uncertain and fuzzy information. The inputs of the sensors are fuzzified, i.e., converted into fuzzy linguistic variables such as "far", "medium", "near", "fast", "slow", etc. Then, through a set of preset "IF-THEN" fuzzy rules, such as "IF the obstacle distance is near AND the relative speed is fast THEN the behavior prediction result is high risk", the fuzzy variables are reasoned, and a clear behavior prediction result is finally obtained. This result not only judges the level of collision risk, but also predicts the possible moving intention of the pedestrian, such as crossing, slow walking in the same direction, or standing still.

[0047] The behavior prediction result is fed back as a dynamic constraint to the model predictive control algorithm for trajectory re-planning, generating and executing an avoidance action, and the trajectory set of the avoidance action constitutes the adjusted path.

[0048] Specifically, when the behavior prediction result indicates a high collision risk or a clear crossing intention of the pedestrian, the behavior prediction result is fed back as a new constraint condition to the optimization process of the model predictive control algorithm. The model predictive control algorithm re-plans the trajectory within the prediction horizon, generating an avoidance path that can effectively avoid the predicted obstacle trajectory and return to the original initial path as soon as possible. In some highly complex interaction scenarios, such as human-robot encounter in a narrow passage, human-robot interaction is also triggered. This may be through the robot's voice broadcast to issue prompts such as "please be careful, the robot is passing, please go first", or through flashing lights of different colors to clearly express its intention to pass. This process is repeated until the robot safely bypasses the obstacle. The actual driving path generated by the combined action of trajectory tracking, behavior prediction, and dynamic adjustment, which contains all the local avoidance actions, is the adjusted path, which is recorded and output. As shown in Figure 3 The predictive control strategy of the present application is intuitively displayed, enabling the robot to smoothly decelerate, follow and overtake during obstacle avoidance, avoiding traditional emergency braking and path interruption, and ensuring smooth execution of the task.

[0049] Optionally, the transportation task is divided into a task chain containing several links, and the power of the transportation robot is monitored in real time during the execution of the task chain. When the power is lower than the power threshold, whether to insert a charging task is decided in combination with the adjusted task priority of the current task, and an updated task chain is obtained, including: The complete transportation task is divided into ordered links of picking up goods, taking the elevator, transporting, and delivering goods to form a task chain; Specifically, at the start of a task, the central scheduling unit will automatically decompose the transportation task into an ordered series of subtasks according to the complete transportation process logic, forming a structured task chain. A typical task chain contains four main links: picking up, taking the elevator, transportation, and delivering. Each link defines clear starting conditions, execution actions, and completion flags. For example, the "picking up" link requires the robot to navigate to the starting point, interact with the material handover equipment or personnel, and confirm that the material is loaded; the "taking the elevator" link involves communication with the elevator control, waiting, entering, and leaving the elevator; the "transportation" link involves traveling according to the planned path within the floor; and the "delivering" link involves completing the material unloading and delivery confirmation at the destination point.

[0050] continuously monitoring the real-time power value of the transportation robot and comparing it with the power threshold; Specifically, during the entire execution of the task chain, the key state of the transportation robot is continuously monitored in real time, and the most core monitoring indicator is the power value. The robot's battery management continuously reports the current accurate power percentage to the central scheduling unit. There is a configurable power preset threshold, such as 20%. In each control cycle, the real-time power value is compared with this preset threshold.

[0051] When the real-time power value is lower than the power threshold, it is determined whether the adjusted task priority of the current task is high priority. If yes, a charging task is inserted after the completion of the current task chain; if no, the current task chain is paused and a charging task is immediately inserted, forming an updated task chain.

[0052] Specifically, when it is detected that the power value of the transportation robot is first lower than the preset threshold, an emergency handling mechanism is triggered. Instead of simply instructing the robot to immediately charge, this mechanism executes an intelligent decision-making process with priority judgment. It evaluates the adjusted task priority of the task currently being executed. If the task is a high-priority task of P0 or P1 level, it is determined that it must be completed first. In this case, the robot will continue to execute the current task chain until the delivery link is completed. After the task is completed, a charging task with the highest priority is automatically generated and inserted at the top of the robot's task queue, guiding it to the nearest available charging station. Conversely, if the current execution is a low-priority task of P2 or P3 level, an interruption and switching strategy is adopted. The current task chain is paused, and the unfinished task state is saved, and a charging task with the highest priority is also generated. After the robot completes charging, it will decide whether to resume executing the interrupted task or to reinsert it into the task pool for re-scheduling according to the current situation. This decision process finally outputs an updated task chain that is dynamically modified and optimized, which may contain a newly inserted charging link.

[0053] Optionally, after the transport robot completes the updated task chain, prompting task completion and collecting transport process data to form feedback data, and transmitting the feedback data to the central scheduling unit comprises: Giving a task completion prompt to the destination person through voice and light; Specifically, when the transport robot navigates to the task destination according to the adjusted path and completes the delivery link, the task is the task completion confirmation and feedback. The robot will issue a task completion prompt to the receiving person at the destination or the surrounding environment through its integrated various human-computer interaction devices. This includes playing a preset voice prompt through the built-in speaker, such as "Hello, your ordered medicine has arrived, please take it", and emitting a specific light signal through the LED light strip on the robot body, such as changing from a blue constant light to a green flashing light, intuitively indicating that the current state has changed to "standby for handover".

[0054] Integrating path deviations, time consumption of each link, and abnormal events recorded during the execution of the transport task to form structured transport process data; Specifically, during the entire transport task execution, from receiving multi-modal instructions to finally delivering, all key process data has been recorded in real time in the robot's local log and the central scheduling unit's database. Now, these scattered data fragments are structured and integrated. The collected data mainly includes: path and navigation data, which includes the path deviation between the planned initial path and the adjusted path actually traveled by the robot, recording all obstacle avoidance and dynamic adjustment details; timestamp data, which details the creation time, allocation time, pickup time, delivery time, and other key nodes of the task, so that total time consumption, waiting time, and other performance indicators can be calculated; state information, which covers the robot's power consumption during task execution, various abnormal events encountered, and interaction records with the environment.

[0055] Packaging the transport process data as feedback data and sending it to the central scheduling unit for subsequent optimization of scheduling rules.

[0056] Specifically, after data collection and arrangement are completed, these structured data are packaged and fed back to the central scheduling unit through a wireless communication network. After receiving these feedback data, the central scheduling unit permanently stores them in the historical task database. These data not only serve as evidence of this task completion, but more importantly, they will become the core resources for self-optimization and iteration. For example, data analysis can regularly analyze path deviation data to identify "hot areas" in the hospital where congestion or human-machine interaction difficulties often occur, so as to assign higher travel costs to these areas in future path planning. Similarly, by analyzing time consumption data, the actual effect of different priority scheduling rules can be evaluated, and the weight coefficients can be adaptively adjusted to continuously optimize the overall scheduling efficiency.

[0057] Optionally, the method further comprises: Periodically refreshing the dynamic map in the environment data with the latest sensor-collected information, to maintain the real-time accuracy of the dynamic map; Specifically, regarding real-time updating of the environment data, the dynamic map is not built once and for all, and then never changed. On the contrary, a continuous data flow pipeline is established. The sensors deployed throughout the hospital, such as cameras and Wi-Fi probes, constantly send the raw data collected at a high frequency to the central server. The data processing engine on the server analyzes these new data in real time, recalculates the personnel density of each area, and updates the device location on the device status layer. This process ensures that the dynamic information in the environment data is always highly synchronized with the real situation of the hospital, thereby ensuring the real-time accuracy of the dynamic map.

[0058] According to the feedback data received by the central scheduling unit, the scheduling rules and the generation logic of the initial task priority are adaptively adjusted using machine learning algorithms; Specifically, regarding adaptive adjustment, the input is the feedback data collected and fed back to the central scheduling unit. The accumulated historical feedback data is analyzed offline periodically. By analyzing a large number of path deviations, task time consumptions, and abnormal event records, potential patterns and bottlenecks are discovered. For example, if the data shows that the time consumption of transporting high-priority tasks always exceeds the expectation during a certain period, the reason is analyzed. If it is found that the priority weight setting is improper, causing it to be disturbed by low-priority tasks, the weight coefficient in the priority calculation formula is automatically adjusted. Similarly, if it is found that a certain corridor has frequent obstacle avoidance events despite the low personnel density data, it is learned that there is a non-person flow causing difficulty in passing through the area, and the fixed passing cost of the road segment in the path planning algorithm is accordingly increased.

[0059] The environment perception, task analysis, scheduling planning, behavior prediction, and emergency handling are integrated into the closed-loop control logic of perception, decision, execution, and learning.

[0060] Specifically, based on the above two mechanisms, an end-to-end closed-loop control from environment perception to task completion is realized. The entire process forms a complete "perception-decision-execution-learning" cycle. Environment perception provides a real-time world model; task analysis sets the action target; scheduling planning, behavior prediction, and emergency handling are responsible for formulating and executing the optimal action plan; and the feedback after task completion plays the role of "evaluator" and "teacher", converting the execution result into experience for learning. This experience in turn optimizes the decision model, and the updated decision model will make more accurate and efficient decisions based on the latest environment perception data in the next task. This closed loop, which iterates continuously, is the core of the entire intelligence.

[0061] Based on the same inventive concept, as Figure 4 The application also provides a dynamic priority-based hospital multi-scene transportation robot hybrid scheduling system, as shown in the drawings, which comprises: An environment perception module is configured to acquire hospital multi-scene environment information and transportation robot types, and generate environment data containing a dynamic map and available robot types; A task analysis module is configured to receive and analyze multi-modal instructions, and extract task parameters containing to-be-transported materials, quantity, starting point and destination, wherein the multi-modal instructions are different modes of instructions that robots can receive and understand from voice, manual input and application; A priority management module is configured to preliminarily classify according to the task parameters and preset scheduling rules, obtain an initial task priority, and dynamically adjust the initial task priority in combination with the dynamic map and the current state of the transportation robot, and generate an adjusted task priority; A scheduling planning module is configured to allocate transportation robots from a robot queue based on the adjusted task priority and the available robot types, and generate an initial path that takes into account short distance and small flow density according to the dynamic map; A trajectory tracking and intervention module is configured to drive the transportation robot to track the initial path, and predict the behavior of obstacles in the process of travel in real time, dynamically generate and execute avoidance actions according to the prediction results, and form an adjusted path; An emergency handling module is configured to decompose a transportation task into a task chain containing several links, and monitor the power of the transportation robot in real time during the execution of the task chain, and when the power is lower than a power threshold, decide whether to insert a charging task in combination with the adjusted task priority of the current task, and obtain an updated task chain; An information feedback module is configured to prompt task completion after the transportation robot completes the updated task chain, collect transportation process data to form feedback data, and transmit the feedback data to a central scheduling unit.

[0062] In order to verify the feasibility of the application in implementation, the application is applied to a certain large general hospital. The hospital building structure is complex, there are many departments, the daily material transportation demand is large and the timeliness requirement is high, covering multiple scenes such as operating rooms, laboratories, pharmacies, inpatient departments, etc. The traditional manual distribution or simple robot scheduling mode has been difficult to meet the efficient, accurate and safe logistics demand, and problems such as delay of emergency material transportation, corridor congestion during peak period, uneven allocation of robot resources often occur. The hospital hopes to use the method of the application to deploy a set of heterogeneous robot queue composed of high-load robots and cabinet door distribution robots, and realize intelligent hybrid scheduling of hospital materials.

[0063] In this embodiment, the hospital deploys lidar, camera network, Wi-Fi probe equipment and Bluetooth Low Energy tags on each floor to build and update a multi-layer dynamic map containing personnel density heat map layer and equipment status layer in real time. Medical staff create transportation tasks through the touch screen of the nurse station, the special App on the personal mobile device or directly giving voice instructions to the robot. After receiving the instructions, the task parameters are generated after natural language processing and real-time feasibility verification, and enter the dynamic priority judgment and robot allocation process. In the task execution process, precise trajectory tracking and intelligent obstacle avoidance are achieved by using model predictive control and fuzzy control algorithms, while task chain management and emergency handling mechanisms ensure the continuity and reliability of the operation. After the completion of the task, the whole process data is fed back to the central scheduling unit for adaptive optimization of the scheduling rules.

[0064] To verify the beneficial effects of the present application, a typical concurrent scenario during the visiting peak period at 15:00 on October 16, 2024 is selected for detailed analysis. During this period, two tasks are received simultaneously: Task one high priority: initiated by the fifth floor operating room nurse station, requesting "emergency delivery of a sterile surgical kit from the central supply room to the seventh operating room".

[0065] Task two low priority: initiated by the third floor inpatient department nurse station, requesting "delivery of ten clean bedclothes to room 302".

[0066] First, the environment data is obtained through multi-sensor fusion technology, and the dynamic map shows that the main corridor leading to the fifth floor operating room has a high personnel density. Then, the multi-modal instructions of the two tasks are analyzed, and the accessibility of the starting point and destination is verified, and standardized task parameters are generated.

[0067] In the dynamic priority generation stage, according to the preset rule base, task one is assigned an initial priority P1 level, and task two is assigned an initial priority P3 level. Subsequently, for task one, due to the high average personnel density on its planned path, its dynamic priority score is significantly increased. For task two, since it has been waiting for 5 minutes, its priority is slightly improved due to the urgency decay factor. Finally, task one obtains a much higher adjusted task priority than task two.

[0068] In the robot allocation and path planning stage, for task one, a candidate robot of type "cabinet door delivery robot" is matched, and the nearest robot #R05 to the central supply room is selected to execute the task. The initial path generated by the improved path planning algorithm actively avoids the most densely populated elevator waiting area and selects a slightly longer but more unobstructed staff passage. For task two, the idle "high load robot" #R02 is assigned, and the standard shortest path is planned.

[0069] In the trajectory tracking and dynamic adjustment phase, when the robot #R05 is approaching the corner of the staff passage, its sensors detect a doctor pushing a treatment cart. The fuzzy control algorithm predicts that the doctor will move straight ahead at a slow speed. The model predictive control algorithm then plans a smooth deceleration following trajectory and overtake at the right moment, instead of emergency braking, without interrupting the task.

[0070] In terms of emergency handling, at 15:25 in the afternoon, the battery level of robot #R08, which is executing another P2-level regular medicine delivery task, drops to 18%, which is lower than the preset threshold of 20%. Since the task is not the highest priority, the emergency handling mechanism is triggered, the current task chain of #R08 is suspended, and a highest-priority charging task is generated to guide it to the nearest charging pile.

[0071] After the task is completed, all process data is collected. Through the analysis of one week of data, it is found that the corridors on the third floor of the inpatient department have frequent obstacle avoidance events from 14:00 to 16:00 in the afternoon due to family visits. Based on this feedback data, the adaptive optimization module automatically adjusts the path planning traffic cost weight of this area during this period to make the subsequent task planning more inclined to avoid this area, thereby improving the overall transportation efficiency.

[0072] It should be noted that the electrical connection between the above-mentioned units does not necessarily represent the direct connection of the line, and the indirect connection mode can also be applied to the embodiments of the present application as long as the purpose of the present application is achieved. The above-described is only an exemplary embodiment of the present application, and cannot limit the scope of the present application.

[0073] That is, any equivalent changes and modifications made in accordance with the teachings of the present application are still within the scope of the present application. Other embodiments of the present application will be readily apparent to those skilled in the art upon considering the specification and practice of the disclosure. The present application is intended to cover any variations, uses, or adaptive changes to the present application that follow the general principles of the present application and include common knowledge or conventional technical means in the art not disclosed by the present application.

Claims

1. A hybrid scheduling method for multi-scenario transport robots in hospitals based on dynamic priority, characterized in that, The method includes: Acquire multi-scenario environmental information and transport robot types in the hospital, and generate environmental data including dynamic maps and available robot types; The robot receives and parses multimodal instructions, extracts and forms task parameters including the goods to be transported, quantity, starting point and destination. The multimodal instructions are instructions from different modes that the robot can receive and understand, such as voice, manual input and application. The task priorities are initially classified according to the task parameters and preset scheduling rules to obtain the initial task priorities. The initial task priorities are then dynamically adjusted based on the dynamic map and the current state of the transport robot to generate the adjusted task priorities. Based on the adjusted task priority and the available robot types, transport robots are allocated from the robot queue, and an initial path that balances short distance and low crowd density is generated according to the dynamic map. The robot is driven to track the initial path and predict the behavior of obstacles in real time. Based on the prediction results, it dynamically generates and executes avoidance actions to form an adjusted path. The transportation task is broken down into a task chain containing several links, and the battery level of the transportation robot is monitored in real time during the execution of the task chain. When the battery level is lower than the battery threshold, a decision is made on whether to insert a charging task based on the adjusted task priority of the current task, and an updated task chain is obtained. After the transport robot completes the updated task chain, it will indicate that the task is complete and collect data from the entire transportation process to form feedback data, which will then be transmitted to the central scheduling unit.

2. The method for hybrid scheduling of multi-scenario transport robots in hospitals based on dynamic priority, as described in claim 1, is characterized in that... The process of acquiring multi-scenario environmental information and transport robot types in the hospital, and generating environmental data including dynamic maps and available robot types, includes: Static and dynamic information of the hospital is collected through multi-sensor fusion technology; A multi-layered dynamic map is constructed based on the static area information and dynamic information. The dynamic map includes a static map base layer and a personnel density heat map layer and an equipment status layer superimposed on it. The type of transport robot is identified, and the dynamic map is combined with the identified transport robot type to generate environmental data.

3. The method for hybrid scheduling of hospital multi-scenario transport robots based on dynamic priority as described in claim 1, characterized in that, The process of receiving and parsing multimodal instructions to extract and form task parameters including the goods to be transported, quantity, starting point, and destination point includes: Receive multimodal commands from voice, manual input, or applications; Natural language processing technology is used to parse the multimodal instructions and extract keywords related to the goods to be transported, quantity, starting point, and destination. The starting point and destination point in the keywords are verified in real time using the dynamic map. After the verification is successful, the keywords are encapsulated to form task parameters.

4. The method for hybrid scheduling of hospital multi-scenario transport robots based on dynamic priority as described in claim 1, characterized in that, Based on the task parameters and preset scheduling rules, an initial task priority is obtained through preliminary classification. This initial priority is then dynamically adjusted using the dynamic map and the current state of the transport robot, resulting in an adjusted task priority that includes: Based on the scheduling rules that quantify the importance of resources and destinations, the task parameters are matched to obtain the initial task priority; Obtain the current status of each transport robot from the robot queue, wherein the current status includes battery level and location information; A comprehensive weighted model is constructed, which takes the initial task priority, urgency decay factor, dynamic map and the current state of the transport robot as input, performs weighted calculation, and outputs the adjusted task priority. The urgency decay factor is a parameter used to quantify and adjust the task priority. Its function is to prevent a high-priority task from occupying a high priority for a long time and ignoring other tasks that may also become urgent.

5. A method for hybrid scheduling of hospital multi-scenario transport robots based on dynamic priority, as described in claim 1, is characterized in that... The process of allocating transport robots from the robot queue based on the adjusted task priority and the available robot types, and generating an initial path that balances short distance and low pedestrian density according to the dynamic map, includes: Based on the materials in the task parameters, candidate robot types are selected from the transport robot types; Calculate the estimated travel time for available robots of the candidate robot type to reach the task start point, and select the robot with the shortest estimated travel time as the assigned transport robot; A path planning algorithm that uses a weighted sum of the physical length of the path and the population density in the dynamic map as the travel cost is used to calculate the path with the lowest total travel cost for the assigned transport robot, which is then used as the initial path.

6. A method for hybrid scheduling of hospital multi-scenario transport robots based on dynamic priority, as described in claim 5, is characterized in that, The driving transport robot tracks the initial path and predicts obstacle behavior in real time. Based on the prediction results, it dynamically generates and executes avoidance actions to form an adjusted path, including: By using model predictive control algorithms, the assigned transport robots can travel precisely along the initial path, and the real-time tracking status of the robots can be output. When the robot's sensors detect an obstacle, the fuzzy control algorithm is used to fuzzify the sensor input and perform fuzzy inference to obtain a behavioral prediction result of the obstacle's future movement intention. The behavior prediction results are used as dynamic constraints and fed back to the model prediction and control algorithm for trajectory replanning, generating and executing avoidance actions. The set of trajectories of the avoidance actions constitutes the adjusted path.

7. The method for hybrid scheduling of hospital multi-scenario transport robots based on dynamic priority as described in claim 1, characterized in that, The process involves decomposing the transportation task into a task chain containing several stages, and monitoring the battery level of the transportation robot in real time during the execution of the task chain. When the battery level falls below a threshold, a decision is made, based on the adjusted task priority of the current task, to determine whether to insert a charging task, resulting in an updated task chain including: The complete transportation task is broken down into orderly steps of picking up goods, taking the elevator, transporting goods, and delivering goods, forming a task chain; The real-time battery level of the transport robot is continuously monitored and compared with a battery threshold. When the real-time battery level is lower than the battery threshold, determine whether the adjusted task priority of the current task is high priority. If yes, insert the charging task after the current task chain is completed. If no, pause the current task chain and immediately insert the charging task to form an updated task chain.

8. A method for hybrid scheduling of hospital multi-scenario transport robots based on dynamic priority, as described in claim 1, is characterized in that, The process of indicating task completion after the transport robot completes the updated task chain, collecting data from the entire transport process to form feedback data, and transmitting the feedback data to the central scheduling unit includes: Send a task completion notification to personnel at the destination via voice and light; Integrate the route deviations, time consumption at each stage, and abnormal events recorded during the execution of transportation tasks to form structured data for the entire transportation process; The entire transportation process data is packaged into feedback data and sent to the central scheduling unit for subsequent optimization of scheduling rules.

9. A method for hybrid scheduling of hospital multi-scenario transport robots based on dynamic priority, as described in claim 1, is characterized in that, The method further includes: The dynamic map in the environmental data is periodically refreshed using the latest sensor data to maintain the real-time accuracy of the dynamic map; Based on the feedback data received by the central scheduling unit, the scheduling rules and the logic for generating initial task priorities are adaptively adjusted using machine learning algorithms. Integrate environmental perception, task analysis, scheduling and planning, behavior prediction and emergency response into a closed-loop control logic of perception, decision-making, execution and learning.

10. A hospital multi-scenario transport robot hybrid scheduling system based on dynamic priority, characterized in that, The system is used in a dynamic priority-based multi-scenario hybrid scheduling method for hospital transport robots as described in any one of claims 1-9, the system comprising: The environmental perception module is used to acquire environmental information and transport robot types in multiple hospital scenarios, and generate environmental data that includes dynamic maps and available robot types. The task parsing module is used to receive and parse multimodal instructions, extract and form task parameters including the materials to be transported, quantity, starting point and destination point. The multimodal instructions are instructions from different modes that the robot can receive and understand, such as voice, manual input and application. The priority management module is used to perform preliminary classification based on the task parameters and preset scheduling rules to obtain the initial task priority, and dynamically adjust the initial task priority in combination with the dynamic map and the current state of the transport robot to generate the adjusted task priority. The scheduling and planning module is used to allocate transport robots from the robot queue based on the adjusted task priority and the available robot types, and to generate an initial path that balances short distance and low crowd density according to the dynamic map. The trajectory tracking and intervention module is used to drive the transport robot to track the trajectory along the initial path, predict the behavior of obstacles during the journey in real time, dynamically generate and execute avoidance actions based on the prediction results, and form an adjusted path. The emergency handling module is used to decompose the transportation task into a task chain containing several links, and monitor the battery level of the transportation robot in real time during the execution of the task chain. When the battery level is lower than the battery threshold, it decides whether to insert a charging task based on the adjusted task priority of the current task, and obtains an updated task chain. The information feedback module is used to indicate that the task is completed after the transport robot completes the updated task chain, collect data from the entire transportation process to form feedback data, and transmit the feedback data to the central scheduling unit.

Images