Centralized multi-robot task allocation and scheduling method based on space-time conflict prediction

Abstract

The invention discloses a centralized multi-robot task allocation and scheduling method based on space-time conflict prediction, and relates to the technical field of logistics robot scheduling, and the method comprises the steps: initializing three-dimensional layout parameters, constructing a discrete space-time state grid, and setting an initial state; generating an initial candidate task sequence for each stacker based on a nearest neighbor heuristic rule; constructing a dynamic space-time conflict probability field according to the planned tracks of all stackers; fusing the conflict probability field to carry out fine task allocation and path evaluation, and balancing execution time and conflict risks by optimizing a comprehensive cost function; generating a global conflict-free space-time reservation table and a detailed scheduling instruction according to the final optimization trajectory; and issuing an instruction and performing real-time monitoring. According to the method, a prediction mechanism based on a space-time conflict probability field is introduced, conflict risks are quantified into dynamic and continuous cost evaluation items, global optimization of task allocation and scheduling is driven, and the adaptability to uncertain factors in a complex three-dimensional motion coupling environment is enhanced.

Description

Technical Field

[0001] This invention belongs to the field of multi-logistics robot scheduling technology, and in particular relates to a centralized robot task allocation and scheduling method based on spatiotemporal conflict prediction. Background Technology

[0002] In the logistics scheduling of automated storage and retrieval systems (AS / RS), the parallel execution of storage and retrieval operations by multiple stacker cranes within densely arranged high-bay racking aisles has become a key means of improving warehousing efficiency. In this three-dimensional operation scenario, each stacker crane needs to coordinate its horizontal (moving along the aisle) and vertical (lifting and lowering between rack levels) movements simultaneously to reach the designated storage location to grab or store goods, and then transfer and hand them over with the shuttle car at the aisle entrance.

[0003] However, due to the narrow physical space of warehouse aisles and the high-density task scheduling adopted to improve throughput, the movement trajectories of multiple stacker cranes intertwine in both horizontal and vertical dimensions. These trajectories are not fixed on a single plane, but rather form complex intersections and proximity relationships in three-dimensional space as they change over time. Traditional task allocation and path planning methods often face limitations in dealing with this kind of time and space competition caused by the high coupling of vertical lifting and horizontal movement.

[0004] This leads to potential competition for the same spatial area between devices during actual system operation. This competition not only refers to complete physical overlap but also includes the risk of interlocking due to overlapping safety buffer zones. When such potential conflicts are not effectively anticipated and avoided, they can trigger a series of operational interruptions during the execution phase, such as emergency braking of equipment, temporary replanning of routes, or passive waiting of tasks. These situations disrupt the planned work rhythm, increase the uncertainty of task completion time, and may reduce the stable operating efficiency and equipment utilization consistency of the entire warehousing system. Therefore, the following solutions are proposed to address these issues. Summary of the Invention

[0005] The purpose of this invention is to provide a centralized multi-robot task allocation and scheduling method based on spatiotemporal conflict prediction. By introducing a prediction mechanism based on spatiotemporal conflict probability field, conflict risk is quantified into a dynamic and continuous cost evaluation item, which drives the global optimization of task allocation and scheduling. This solves the problem that existing methods are lagging in prediction and passive in avoidance when dealing with dynamic spatial competition conflicts caused by the coupling of multiple device motion trajectories in three-dimensional dense warehouse scenarios, and it is difficult to achieve coordinated optimization of time efficiency and operational safety at the planning source.

[0006] To solve the above-mentioned technical problems, the present invention is achieved through the following technical solution:

[0007] This invention relates to a centralized multi-robot task allocation and scheduling method based on spatiotemporal conflict prediction, the method comprising the following steps:

[0008] Step S1, System Initialization and Environment Modeling: Read the layout parameters of the working environment and the state parameters of multiple mobile robots, construct a spatiotemporal state grid, and receive the task set;

[0009] Step S2, Candidate Task Sequence Generation: For each mobile robot, tasks are selected from the task set based on preset rules to generate an initial candidate task sequence;

[0010] Step S3: Construction of dynamic spatiotemporal conflict probability field: Based on the planned motion trajectories corresponding to the candidate task sequences of all mobile robots, calculate and generate a dynamic spatiotemporal conflict probability field covering the workspace. This probability field is used to characterize the estimated probability of spatial preemption conflicts occurring at different spatiotemporal locations in the future.

[0011] Step S4, Optimize the allocation and scheduling of conflict probabilities: Define a comprehensive cost function, which includes at least the estimated time cost based on the planned motion trajectory and the conflict risk cost based on the dynamic spatiotemporal conflict probability field; optimize and adjust the candidate task sequence of each mobile robot based on the comprehensive cost function to obtain the final task allocation and motion scheduling scheme.

[0012] Step S5: Generating anti-collision scheduling instructions: Based on the final task allocation and motion scheduling scheme, generate specific motion control instructions for each mobile robot;

[0013] Step S6, Instruction Issuance and Dynamic Monitoring: The motion control instructions are issued to the corresponding mobile robot, and periodic status monitoring and rolling optimization are performed during task execution.

[0014] Furthermore, in the system initialization and environment modeling steps, the spatiotemporal state grid constructed is a three-dimensional discrete grid, where two dimensions correspond to the horizontal and vertical spatial coordinates of the working environment, and the third dimension corresponds to the discretized time step.

[0015] Furthermore, in the candidate task sequence generation step, the preset rule is a nearest neighbor heuristic rule, which uses the conflict-free estimated time for the mobile robot to reach the task target position from the current position as the selection cost to iteratively generate the initial candidate task sequence.

[0016] Furthermore, the steps for constructing the dynamic spatiotemporal conflict probability field specifically include:

[0017] For each mobile robot, its planned motion trajectory is determined based on its candidate task sequence, and the trajectory is discretized in the time dimension to obtain a series of spatiotemporal points.

[0018] Define a spatiotemporal occupancy influence function for each spatiotemporal point, which represents the influence distribution centered on the planned position of the mobile robot at that spatiotemporal point and spreading outwards to the surrounding space;

[0019] Based on the spatiotemporal occupancy influence function of different mobile robots at the same spatiotemporal point, calculate the conflict intensity between any two mobile robots;

[0020] The conflict intensity of all mobile robot pairs is aggregated and converted into conflict probability values ​​through a preset mapping function, thereby forming a dynamic spatiotemporal conflict probability field covering the entire workspace and future time periods.

[0021] Furthermore, the spatiotemporal occupancy influence function is a Gaussian diffusion function based on the planned position of the mobile robot, and its diffusion range is related to the physical size of the mobile robot, its motion state, and the system uncertainty parameters.

[0022] Furthermore, in the step of allocating and scheduling optimization of the fusion conflict probability, the comprehensive cost function is expressed as a weighted sum of the time cost term and the conflict risk cost term; wherein, the conflict risk cost term is obtained by performing an integral operation on the dynamic spatiotemporal conflict probability field in the corresponding spatiotemporal domain along the planned motion trajectory of the mobile robot.

[0023] Furthermore, in the step of allocating and scheduling optimization of the fusion conflict probability, based on the comprehensive cost function, an iterative optimization algorithm is used to jointly adjust the candidate task sequences of all mobile robots and their corresponding planned motion trajectories, so as to minimize the sum of the comprehensive cost functions of all mobile robots; during the optimization process, the dynamic spatiotemporal conflict probability field is dynamically updated according to the adjusted planned motion trajectory.

[0024] Furthermore, the anti-conflict scheduling instruction generation step also includes generating a global spatiotemporal reservation table, which is used to exclusively allocate key spatial resource units in the workspace to specific mobile robots in different time periods, thereby avoiding spatial resource conflicts when motion instructions are executed.

[0025] Furthermore, the rolling optimization in the instruction issuance and dynamic monitoring steps refers to: using the actual state of the mobile robot obtained within the monitoring period as a new initial condition, re-executing the candidate task sequence generation step, the dynamic spatiotemporal conflict probability field construction step, and the allocation and scheduling optimization step of the fusion conflict probability, so as to reallocate and schedule tasks that have not yet started execution.

[0026] The present invention has the following beneficial effects:

[0027] 1. This invention transforms conflict prediction from traditional discrete detection to continuous probability assessment by constructing a dynamic spatiotemporal conflict probability field. In the initial task allocation phase, the system calculates the conflict probability at various points in time and space based on the stacker crane's expected trajectory, forming a probability distribution field covering the three-dimensional workspace. This field quantifies potential conflict hotspots and their likelihood of occurrence, allowing conflict avoidance measures to intervene early in the planning stage. By integrating the conflict probability integral into the cost assessment of the task sequence, the system can automatically favor paths and task sequences with lower conflict risk, thereby reducing the conditions for conflict occurrence at the scheduling source. This proactive prediction mechanism helps reduce the need for interruptions and adjustments during subsequent execution, improving the continuity of the task flow and overall operational efficiency.

[0028] 2. This invention introduces a comprehensive cost function, unifying time efficiency and conflict risk within the same optimization framework. The conflict risk cost is calculated through the cumulative effect of the trajectory in the conflict probability field, reflecting a continuous measure of path safety. When adjusting task allocation, the system considers not only the shortest time path but also the conflict probability intensity of the areas traversed by the path, thereby seeking a balance among multiple competing objectives. This optimization method avoids the implicit conflicts that may arise from single-dimensional decision-making, making the generated scheduling scheme both time-efficient and operationally safe. Through iterative optimization, the system can gradually approach the globally optimal solution, improving the overall performance and resource utilization rationality of multi-robot collaborative operations.

[0029] 3. This invention employs a combination of rolling time-domain optimization and real-time status monitoring, enabling the system to continuously perceive changes in equipment status during task execution. It periodically uses the current actual state as the initial condition to repeatedly predict the conflict probability field and optimize task allocation, allowing the scheduling scheme to respond promptly to execution deviations or external interference. This dynamic adjustment mechanism allows the system to locally replan unexecuted tasks without interrupting overall operations, maintaining conflict-free operation constraints. This enhances the system's ability to cope with uncertainties and ensures scheduling stability and task completion reliability during long-term operations.

[0030] Of course, any product implementing this invention does not necessarily need to achieve all of the advantages described above at the same time. Attached Figure Description

[0031] To more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0032] Figure 1This is a flowchart illustrating the centralized multi-robot task allocation and scheduling method based on spatiotemporal conflict prediction of the present invention. Detailed Implementation

[0033] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0034] Please see Figure 1 As shown, this invention is a centralized multi-robot task allocation and scheduling method based on spatiotemporal conflict prediction, comprising the following steps:

[0035] Step S1, System Initialization and Environment Modeling: Read the layout parameters of the working environment and the state parameters of multiple mobile robots, construct a spatiotemporal state grid, and receive the task set;

[0036] Step S2, Candidate Task Sequence Generation: For each mobile robot, tasks are selected from the task set based on preset rules to generate an initial candidate task sequence;

[0037] Step S3: Construction of dynamic spatiotemporal conflict probability field: Based on the planned motion trajectories corresponding to the candidate task sequences of all mobile robots, calculate and generate a dynamic spatiotemporal conflict probability field covering the workspace. This probability field is used to characterize the estimated probability of spatial preemption conflicts occurring at different spatiotemporal locations in the future.

[0038] Step S4, Optimize the allocation and scheduling of conflict probabilities: Define a comprehensive cost function, which includes at least the estimated time cost based on the planned motion trajectory and the conflict risk cost based on the dynamic spatiotemporal conflict probability field; optimize and adjust the candidate task sequence of each mobile robot based on the comprehensive cost function to obtain the final task allocation and motion scheduling scheme.

[0039] Step S5, anti-collision scheduling instruction generation: Based on the final task allocation and motion scheduling scheme, generate specific motion control instructions for each mobile robot;

[0040] Step S6, Command Issuance and Dynamic Monitoring: Send motion control commands to the corresponding mobile robot and perform periodic status monitoring and rolling optimization during task execution.

[0041] In the system initialization and environment modeling steps, the spatiotemporal state grid is constructed as a three-dimensional discrete grid, where two dimensions correspond to the horizontal and vertical spatial coordinates of the working environment, and the third dimension corresponds to the discretized time step.

[0042] In the candidate task sequence generation step, the preset rule is the nearest neighbor heuristic rule, which uses the conflict-free estimated time for the mobile robot to reach the task target position from the current position as the selection cost to iteratively generate the initial candidate task sequence.

[0043] The specific steps for constructing a dynamic spatiotemporal conflict probability field include:

[0044] For each mobile robot, its planned motion trajectory is determined based on its candidate task sequence, and the trajectory is discretized in the time dimension to obtain a series of spatiotemporal points.

[0045] Define a spatiotemporal occupancy influence function for each spatiotemporal point, which represents the influence distribution centered on the planned position of the mobile robot at that spatiotemporal point and spreading outwards to the surrounding space;

[0046] Based on the spatiotemporal occupancy influence function of different mobile robots at the same spatiotemporal point, calculate the conflict intensity between any two mobile robots;

[0047] The conflict intensity of all mobile robot pairs is aggregated and converted into conflict probability values ​​through a preset mapping function, thereby forming a dynamic spatiotemporal conflict probability field covering the entire workspace and future time periods.

[0048] The spatiotemporal occupancy influence function is a Gaussian diffusion function based on the planned position of the mobile robot. Its diffusion range is related to the physical size of the mobile robot, its motion state, and the system uncertainty parameters.

[0049] In the optimization steps of merging conflict probability allocation and scheduling, the comprehensive cost function is expressed as a weighted sum of time cost and conflict risk cost; whereby the conflict risk cost is obtained by integrating the dynamic spatiotemporal conflict probability field in the corresponding spatiotemporal domain along the planned motion trajectory of the mobile robot.

[0050] In the allocation and scheduling optimization steps of the fusion conflict probability, based on the comprehensive cost function, an iterative optimization algorithm is used to jointly adjust the candidate task sequences of all mobile robots and their corresponding planned motion trajectories to minimize the sum of the comprehensive cost functions of all mobile robots; during the optimization process, the dynamic spatiotemporal conflict probability field is dynamically updated according to the adjusted planned motion trajectory.

[0051] The anti-conflict scheduling instruction generation step also includes generating a global spatiotemporal reservation table. The global spatiotemporal reservation table is used to exclusively allocate key spatial resource units in the workspace to specific mobile robots in different time periods, thereby avoiding spatial resource conflicts when motion instructions are executed.

[0052] Rolling optimization in the instruction issuance and dynamic monitoring steps refers to: using the actual state of the mobile robot obtained within the monitoring period as a new initial condition, re-executing the candidate task sequence generation step, the dynamic spatiotemporal conflict probability field construction step, and the allocation and scheduling optimization step of fusion conflict probability, in order to reallocate and schedule tasks that have not yet started execution.

[0053] One specific application of this embodiment is:

[0054] Step S100: System Initialization and Environment Modeling

[0055] The central control server reads the physical parameters of the warehouse layout, including: the number of aisles. Number of shelving rows in each aisle Number of floors maximum horizontal speed of stacker crane Vertical maximum speed As well as the fixed coordinates of the shuttle car handover points at each tunnel entrance; construct a three-dimensional discrete spatiotemporal state grid. Each grid cell Indicates at discrete time step Time, spatial location state, and These are discrete coordinates in the horizontal and vertical directions, respectively; initialize all stacker cranes. The initial position and state are idle; a batch of tasks to be processed is received. Each task contains the coordinates of the target cargo location. And the task type (save or retrieve).

[0056] Step S200: Generation of candidate task sequences based on spatiotemporal grid

[0057] For each idle or soon-to-be-idle stacker crane From the current unassigned task set In the middle, an initial sequence of candidate tasks is generated based on the nearest neighbor heuristic. ,in This is the task identifier; this rule is used for calculation. Move from current location or the end point of the previous task to the task. The estimated conflict-free time of the target cargo location is used as the cost. The task with the lowest cost is selected and added to the sequence. The process is iterated until the sequence length reaches the preset upper limit or the task set is empty. The estimated conflict-free time is only a simple estimate of the Euclidean distance divided by the maximum speed. At this point, fine-grained conflict prediction has not yet been performed.

[0058] Step S300: Construct a dynamic spatiotemporal conflict probability field

[0059] For all current stacker cranes, based on their candidate task sequences The planned initial motion trajectory includes horizontal movement segments and vertical ascent / recline segments. A dynamic spatiotemporal conflict probability field covering the three-dimensional working space of all tunnels is calculated. The venue indicated that at the time... Spatial location The estimated probability of a space-grabbing conflict occurring between two or more stacker cranes at a given location;

[0060] First, check each stacker crane The trajectory is represented by spatiotemporal discretization; it is derived from the sequence. The complete motion process from the start of the first task to the end of the last task, in terms of time steps. Discretized into a series of spatiotemporal points ,in For discrete-time indexing;

[0061] Define a single stacker crane At the point of spacetime Spatiotemporal occupancy influence function This function is not a simple 0-1 occupation, but rather a stacker crane-based... At any moment actual location A Gaussian diffusion function centered at the origin is used to simulate the effects of motion uncertainties and physical safety buffer zones.

[0062]

[0063] In the formula, For stacker cranes In time spatial point The intensity value of the occupation effect generated at the location; This is a number index for the stacker crane, used to identify the first stacker crane in the system. Stacker crane; These are the horizontal coordinates of the point in space. The vertical coordinates of the point in space; For a point in time; It is a natural exponential function; For stacker cranes In time The actual horizontal coordinate at that time; For stacker cranes In time The actual vertical coordinates at that time; The horizontal influence diffusion coefficient determines the size and attenuation rate of the horizontal influence range of the stacker crane, and is related to the physical width of the equipment, horizontal positioning error and safety margin. The vertical influence diffusion coefficient determines the size and attenuation rate of the vertical influence range of the stacker crane, and is related to the physical height of the equipment, vertical positioning error and safety margin.

[0064] Calculate the results of any two different stacker cranes. and At the point of spacetime Intensity of conflict The intensity is defined as the product of their respective spatiotemporal occupancy influence functions:

[0065]

[0066] In the formula, In time spatial point stacker crane With stacker crane Estimates of the intensity of the conflict between them; For the number index of two different stacker cranes, and ; For stacker cranes The influence function value at this spatiotemporal point;

[0067] The higher this value, the greater the likelihood that the two devices will spatially overlap or become too close at that point in time;

[0068] Aggregate the conflict intensities of all stacker crane pairs and map them to conflict probabilities through a saturation function to obtain the dynamic spatiotemporal conflict probability field of the entire system. :

[0069]

[0070] In the formula, In time spatial point The estimated probability of any spatial conflict occurring in the entire multi-robot system is located at [0,1]. These are the normalization coefficients; This represents the total number of stacker cranes in the system. For stacker cranes The intensity of the conflict at that point in spacetime; The operation ensures that the probability value does not exceed 1; this probability field changes dynamically in time and space, reflecting potential conflict hotspots in the near future.

[0071] Step S400: Fine-grained task allocation and path cost evaluation based on conflict probabilities.

[0072] In this step, the candidate task sequences of stacker cranes generated in step S200 are processed. Perform refined cost assessment and adjustment; instead of simply using conflict-free estimated time, define a comprehensive cost function. To evaluate the total cost of performing this task sequence:

[0073]

[0074] In the formula, To evaluate stacker cranes Execute its candidate task sequence The total overall cost; For stacker cranes A candidate task sequence is an ordered list of tasks. For sequence The estimated total execution time is obtained by calculating the motion time of each segment along the sequence trajectory based on the stacker crane motion model and summing them up.

[0075] As a consequence of conflict risk, the stacker crane is integrated into the system. The planned trajectory is calculated using the probability field values ​​of the conflicts it experiences in the spatiotemporal domain:

[0076]

[0077] In the formula, For discrete-time indexing, starting from the start time of sequence execution. Traverse to the end time ; For sequence The estimated start time of execution; For sequence The estimated end time of the execution; The symbol for a two-dimensional surface integral indicates that the integral is performed over the region of integration. Integrate on top; These are the horizontal coordinates. and vertical coordinates The symbol for differentiation indicates integration; the region of integration. It is a local buffer zone centered on the trajectory point. As a weight, it represents the contribution of the stacker crane to the conflict probability within its own influence range; this cost function quantifies the abstract conflict probability field into an accumulative cost term. The higher the frequency and the longer the time the trajectory passes through the high conflict probability spatiotemporal region, the greater its conflict risk cost.

[0078] To balance parameters, the weight between time efficiency and conflict risk is adjusted; the task sequence of each stacker crane is adjusted by optimization algorithms (such as iterative local search or simulated annealing) with the goal of minimizing the sum of the comprehensive costs of all stacker crane sequences; when adjusting the sequence, the detailed motion trajectory of the corresponding stacker crane (including the coordination of horizontal and vertical movements) is replanned, and the conflict probability field in step S300 is dynamically updated to form a closed-loop optimization.

[0079] Step S500: Generate anti-collision scheduling instructions and a time-space reservation table;

[0080] After optimization in step S400, each stacker crane is obtained. The final task sequence and its detailed spatiotemporal trajectory Based on this, the system generates a global spatiotemporal reservation table; this table allocates reservations for each critical, potentially conflicting spatial resource unit (e.g., a section of horizontal tunnel at a specific height) in the time dimension, ensuring that each spatial resource unit can be reserved by at most one stacker crane at any given time.

[0081] The reservation rules take into account the physical dimensions and safety clearances of the stacker crane; for stacker cranes For each segment of motion in the trajectory, calculate the start and end times of its occupation of a specific spatial resource unit, and mark the time interval as "reserved"; the system checks the final global reservation table to ensure that there are no overlapping reservations; if there is an overlap (very low probability, because step S400 has been optimized), then make a small speed adjustment (within the allowable range) on the stacker cranes involved in the overlap to eliminate the overlap.

[0082] Step S600: Command Issuance and Dynamic Monitoring Execution

[0083] The central control server sends the final task sequence, detailed motion path points (including time functions of horizontal and vertical coordinates), and speed commands to the corresponding stacker crane controller; the stacker crane then begins to execute the task according to the commands.

[0084] During task execution, the central control server continuously monitors the location and status of all stacker cranes in real time. Due to execution errors or unexpected delays, the system will periodically (e.g., every fixed time interval) repeat steps S200 to S500 with the current actual state as the initial condition to perform rolling time-domain optimization. However, the scope of re-optimization is limited to tasks that have not yet started execution and affected stacker cranes, in order to quickly respond to dynamic changes and recalculate and update the spatiotemporal conflict probability field to continuously ensure the conflict-free scheduling.

[0085] In the description of this specification, references to terms such as "an embodiment," "example," "specific example," etc., indicate that a specific feature, structure, material, or characteristic described in connection with that embodiment or example is included in at least one embodiment or example of the invention. In this specification, the illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples.

[0086] The preferred embodiments of the present invention disclosed above are merely illustrative of the invention. These preferred embodiments do not exhaustively describe all details, nor do they limit the invention to the specific implementations described. Clearly, many modifications and variations can be made based on the content of this specification. This specification selects and specifically describes these embodiments to better explain the principles and practical applications of the invention, thereby enabling those skilled in the art to better understand and utilize the invention. The invention is limited only by the claims and their full scope and equivalents.

Claims

1. A centralized multi-robot task allocation and scheduling method based on spatiotemporal conflict prediction, characterized in that, The method includes the following steps: Step S1, System Initialization and Environment Modeling: Read the layout parameters of the working environment and the state parameters of multiple mobile robots, construct a spatiotemporal state grid, and receive the task set; Step S2, Candidate Task Sequence Generation: For each mobile robot, tasks are selected from the task set based on preset rules to generate an initial candidate task sequence; Step S3: Construction of dynamic spatiotemporal conflict probability field: Based on the planned motion trajectories corresponding to the candidate task sequences of all mobile robots, calculate and generate a dynamic spatiotemporal conflict probability field covering the workspace. This probability field is used to characterize the estimated probability of spatial preemption conflicts occurring at different spatiotemporal locations in the future. Step S4, Optimize the allocation and scheduling of conflict probabilities: Define a comprehensive cost function, which includes at least the estimated time cost based on the planned motion trajectory and the conflict risk cost based on the dynamic spatiotemporal conflict probability field; optimize and adjust the candidate task sequence of each mobile robot based on the comprehensive cost function to obtain the final task allocation and motion scheduling scheme. Step S5: Generating anti-collision scheduling instructions: Based on the final task allocation and motion scheduling scheme, generate specific motion control instructions for each mobile robot; Step S6, Instruction Issuance and Dynamic Monitoring: The motion control instructions are issued to the corresponding mobile robot, and periodic status monitoring and rolling optimization are performed during task execution.

2. The centralized multi-robot task allocation and scheduling method based on spatiotemporal conflict prediction according to claim 1, characterized in that, In the system initialization and environment modeling steps, the spatiotemporal state grid constructed is a three-dimensional discrete grid, where two dimensions correspond to the horizontal and vertical spatial coordinates of the working environment, and the third dimension corresponds to the discretized time step.

3. The centralized multi-robot task allocation and scheduling method based on spatiotemporal conflict prediction according to claim 1, characterized in that, In the candidate task sequence generation step, the preset rule is the nearest neighbor heuristic rule, which uses the conflict-free estimated time for the mobile robot to reach the task target position from the current position as the selection cost to iteratively generate the initial candidate task sequence.

4. The centralized multi-robot task allocation and scheduling method based on spatiotemporal conflict prediction according to claim 1, characterized in that, The specific steps for constructing the dynamic spatiotemporal conflict probability field include: For each mobile robot, its planned motion trajectory is determined based on its candidate task sequence, and the trajectory is discretized in the time dimension to obtain a series of spatiotemporal points. Define a spatiotemporal occupancy influence function for each spatiotemporal point, which represents the influence distribution centered on the planned position of the mobile robot at that spatiotemporal point and spreading outwards to the surrounding space; Based on the spatiotemporal occupancy influence function of different mobile robots at the same spatiotemporal point, calculate the conflict intensity between any two mobile robots; The conflict intensity of all mobile robot pairs is aggregated and converted into conflict probability values ​​through a preset mapping function, thereby forming a dynamic spatiotemporal conflict probability field covering the entire workspace and future time periods.

5. The centralized multi-robot task allocation and scheduling method based on spatiotemporal conflict prediction according to claim 4, characterized in that, The spatiotemporal occupancy influence function is a Gaussian diffusion function based on the planned position of the mobile robot, and its diffusion range is related to the physical size of the mobile robot, its motion state, and the system uncertainty parameters.

6. The centralized multi-robot task allocation and scheduling method based on spatiotemporal conflict prediction according to claim 1, characterized in that, In the step of allocating and scheduling optimization of the fusion conflict probability, the comprehensive cost function is expressed as a weighted sum of the time cost term and the conflict risk cost term; wherein, the conflict risk cost term is obtained by performing an integral operation on the dynamic spatiotemporal conflict probability field in the corresponding spatiotemporal domain along the planned motion trajectory of the mobile robot.

7. The centralized multi-robot task allocation and scheduling method based on spatiotemporal conflict prediction according to claim 6, characterized in that, In the step of allocating and scheduling optimization of the fusion conflict probability, based on the comprehensive cost function, an iterative optimization algorithm is used to jointly adjust the candidate task sequences of all mobile robots and their corresponding planned motion trajectories so as to minimize the sum of the comprehensive cost functions of all mobile robots; during the optimization process, the dynamic spatiotemporal conflict probability field is dynamically updated according to the adjusted planned motion trajectory.

8. The centralized multi-robot task allocation and scheduling method based on spatiotemporal conflict prediction according to claim 1, characterized in that, The anti-conflict scheduling instruction generation step also includes generating a global spatiotemporal reservation table, which is used to exclusively allocate key spatial resource units in the workspace to specific mobile robots in different time periods, thereby avoiding spatial resource conflicts when motion instructions are executed.

9. The centralized multi-robot task allocation and scheduling method based on spatiotemporal conflict prediction according to claim 1, characterized in that, The rolling optimization in the instruction issuance and dynamic monitoring steps refers to: using the actual state of the mobile robot obtained within the monitoring period as a new initial condition, re-executing the candidate task sequence generation step, the dynamic spatiotemporal conflict probability field construction step, and the allocation and scheduling optimization step of the fusion conflict probability, so as to reallocate and schedule tasks that have not yet started execution.

Images