A distribution control system and method suitable for unmanned delivery vehicles
By constructing a dynamic pressure field map and a multi-level collaborative scheduling mechanism, the problem of rigid resource allocation in the unmanned delivery vehicle system under dynamic environments was solved, enabling accurate response and efficient scheduling of real-time orders and road conditions, and improving the robustness and overall efficiency of the system.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- TAIYUAN UNIVERSITY OF TECHNOLOGY
- Filing Date
- 2026-03-31
- Publication Date
- 2026-06-30
AI Technical Summary
Existing unmanned delivery vehicle dispatching systems lack adaptive and collaborative control capabilities when faced with real-time order surges and dynamic road conditions, resulting in rigid resource allocation and a decline in overall efficiency.
A dynamic pressure field map is constructed, and dynamic responsibility areas are assigned to each unmanned delivery vehicle based on real-time order and road condition data. Through a multi-level collaborative scheduling mechanism of local task migration and global responsibility redistribution, task allocation is dynamically adjusted, and path conflicts are identified and resolved.
It achieves accurate perception and adaptive scheduling of dynamic environments, improves the robustness and resource utilization efficiency of unmanned delivery vehicle clusters in complex environments, avoids vehicle congestion and waiting, and ensures the reliable execution of scheduling schemes.
Smart Images
Figure CN122308196A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of intelligent logistics and automatic control technology, and in particular to a distribution control system and method suitable for unmanned delivery vehicles. Background Technology
[0002] Efficient scheduling of unmanned delivery vehicles is key to improving last-mile delivery efficiency. Existing scheduling methods often perform poorly when faced with real-time changes in order distribution, road conditions, and vehicle status. Their core deficiency lies in the lack of adaptive and collaborative control capabilities under dynamic perception.
[0003] Specifically, mainstream dispatching systems mostly employ allocation strategies based on fixed rules or static models. For example, they may pre-divide fixed service areas or assign pre-set task sequences to vehicles based on historical data. These methods are effective when orders and road conditions are relatively stable, but they cannot respond sensitively or flexibly to real-time "regional order backlogs" and "sudden traffic congestion" during delivery. When a local area experiences a sudden drop in delivery efficiency due to a surge in orders or congestion, the system often only performs simple task reallocation or waits, lacking a multi-layered, collaborative control mechanism that can quantify the regional "pressure" state and trigger a shift from local task reallocation to global resource reconfiguration. This results in rigid resource allocation and overall efficiency decline when facing dynamic disturbances.
[0004] Therefore, there is an urgent need for an intelligent scheduling scheme that can integrate multi-source information in real time, dynamically assess the cluster's operating status, and automatically execute hierarchical collaborative control to improve the overall operational flexibility and efficiency of unmanned delivery vehicle clusters in complex and dynamic environments. Summary of the Invention
[0005] The purpose of this invention is to provide a distribution control system and method suitable for unmanned delivery vehicles, which solves the problems of rigid resource allocation and slow response in existing scheduling schemes when dealing with real-time order surges and dynamic road conditions.
[0006] To achieve the above objectives, the present invention provides a distribution control method suitable for unmanned delivery vehicles, comprising the following steps: Step S1: Obtain order data and road condition data in the delivery area in real time, calculate the real-time pressure value of each location in the area, and construct a dynamic pressure field map; Step S2: Based on the dynamic pressure field map and the real-time location of each available unmanned delivery vehicle, assign a dynamic responsibility area to each available unmanned delivery vehicle; Step S3: Based on the divided dynamic responsibility areas, newly generated delivery orders are assigned to unmanned delivery vehicles in the corresponding responsibility areas, and the pressure relief rate of each dynamic responsibility area is calculated based on the changes in the dynamic pressure field map. Step S4: Based on the pressure relief rate, identify the dynamic responsibility area where the pressure relief rate is continuously lower than the preset threshold, and determine that the area has entered a scheduling deadlock state. Step S5: In response to the scheduling deadlock state determined in step S4, perform the following operations in sequence: Step S51, Local Task Migration: Package and migrate some orders within the dynamic responsibility area that are in a scheduling deadlock to other cooperating vehicles that meet the path fit conditions; Step S52, Global Responsibility Redistribution: If the area is still in a scheduling deadlock state after the local task migration is performed, then based on the current dynamic pressure field map, global optimization is performed again and the dynamic responsibility areas of all available vehicles are updated. Step S6: Based on the vehicle task plan determined after the scheduling response in step S5, predict and analyze the spatiotemporal conflicts of vehicles at future path nodes. Step S7: Issue and execute the final command after conflict resolution, and return the actual status information fed back by the vehicle to step S1 to update the dynamic pressure field map.
[0007] Preferably, step S1 specifically includes: Step S11: Receive the location information of orders to be delivered pushed by the order system in real time, and obtain the real-time average vehicle speed from the traffic data platform; Step S12: Divide the delivery area into grids, and for each grid... ,exist At any given moment, calculate the real-time pressure value. : ; in, express Time Grid The number of orders awaiting delivery within the specified quantity. This indicates the preset maximum capacity for a single grid order. express Time Grid Real-time congestion coefficient for the corresponding road segment; Real-time congestion coefficient Based on the obtained real-time average vehicle speed, it is determined through a preset speed-congestion mapping table.
[0008] Preferably, step S2 specifically includes: Step S21: Mark the grid where the target vehicle is currently located as the current seed grid and add it to the dynamic responsibility region set of the vehicle; Step S22: Obtain all adjacent grids of the current seed grid; Step S23: Determine whether the real-time pressure value in the dynamic pressure field map generated by each adjacent grid is greater than the preset pressure threshold. And not included in the dynamic responsibility area of any other vehicle; adjacent grids that meet the conditions will be incorporated into the dynamic responsibility area set; Step S24: Take the newly merged mesh as the new current seed mesh, and repeat steps S22 to S23 until no new mesh meets the merging condition.
[0009] Preferably, step S3 specifically includes: Step S31, for dynamic responsibility areas ,exist Record the set of real-time pressure values of all grids within its region calculated in step S1. ; Step S32, after the monitoring time interval Afterwards, Step S1 is called again at any time to obtain the dynamic responsibility area. The set of real-time pressure values after all grid updates. ; Step S33: Calculate the dynamic responsibility area exist Pressure dissipation rate over time : ; in, This indicates traversing the dynamic responsibility region. Index of all grids within.
[0010] Preferably, step S51 specifically includes: Step S511: Obtain all vehicles within a preset range that the current route plan will pass through the scheduling deadlock area as candidate cooperative vehicles; Step S512: For each candidate cooperative vehicle and migration task package Calculate its path fit : ; in, Indicates candidate cooperative vehicles The original planned path was in accepting the migration task package. Additional driving distance required afterward Indicates acceptance of migration task package For candidate collaborative vehicles The total estimated time delay caused by all other orders in the original task sequence, Indicates the preset weighting coefficients; Step S513, Select The candidate collaborative vehicle with the highest value will be selected as the collaborative vehicle to execute the migration task package.
[0011] Preferably, step S52 specifically includes: Step S521: Using the dynamic pressure field map generated in the current step S1 as input, minimize the variance of the real-time pressure values of all grids. The objective function is defined as follows: the number of virtual responsibility points equals the total number of currently available vehicles. As constraints, an optimization algorithm is used to calculate an optimal set of virtual responsibility point locations. ,in to They represent the 1st to the 1st. The two-dimensional coordinates of each virtual responsibility point within the delivery area; Step S522: Calculate the Euclidean distance from the real-time location of each available unmanned delivery vehicle to each virtual responsibility point, and use an assignment algorithm to assign a unique virtual responsibility point to each vehicle, so as to minimize the total distance of all vehicles to their assigned points. Step S523: Using the virtual responsibility point assigned to each vehicle in step S522 as the new initial point, re-execute step S2 for each vehicle to generate a new dynamic responsibility area.
[0012] Preferably, in step S521, the optimization algorithm used is simulated annealing, and its energy function is... We can directly use the objective function, that is: .
[0013] Preferably, step S6 specifically includes: Step S61: Traverse all shared path nodes. For any two vehicles A and B whose planned time windows overlap, determine that there is a potential conflict. Step S62: For each conflict, calculate the time that delays the departure of vehicle A. Alternatively, vehicle B could slow down on the road section before the conflict point to increase travel time. The adjustment plan; Step S63: Compare the total adjustment cost The calculation formula is as follows: ; in, and These are the priority weights for the tasks carried by vehicles A and B, respectively; [The text abruptly ends here, likely due to an incomplete sentence or a formatting error.] The minimum necessary adjustments will be implemented. Step S64: Based on the selected adjustment scheme, update the departure time or segment speed in the route planning instructions of the affected vehicles.
[0014] Preferably, the shared path nodes in step S6 include road intersections, delivery station entrances and exits, and charging station entrances that are predefined by the system and are prone to congestion or limited space.
[0015] The present invention also provides a distribution control system suitable for unmanned delivery vehicles, comprising: The dynamic sensing module is used to acquire order data and road condition data in the delivery area in real time, calculate the real-time pressure value of each location in the area, and build a dynamic pressure field map. The responsibility area allocation module, connected to the dynamic perception module, is used to allocate a dynamic responsibility area to each available unmanned delivery vehicle based on the dynamic pressure field map and the real-time location of each available unmanned delivery vehicle. The order scheduling module, connected to the responsibility area allocation module, is used to allocate newly emerging delivery orders to the unmanned delivery vehicles in the corresponding responsibility areas based on the divided dynamic responsibility areas, and to calculate the pressure relief rate of each dynamic responsibility area based on the changes in the dynamic pressure field map. The deadlock determination module, connected to the order scheduling module, is used to identify dynamic responsibility areas where the pressure relief rate is continuously lower than a preset threshold based on the pressure relief rate, and determine that the area has entered a scheduling deadlock state. The multi-level collaborative scheduling module, connected to the deadlock determination module, is used to respond to scheduling deadlock states and sequentially perform the following operations: The local task migration unit is used to package and migrate a portion of orders within a dynamic responsibility area that is in a scheduling deadlock state to other cooperating vehicles that meet the path fit conditions. The Global Responsibility Redistribution Unit is used to perform global optimization and update the dynamic responsibility areas of all available vehicles based on the current dynamic pressure field map when the area is still in a scheduling deadlock state after a local task migration. The conflict resolution module, connected to the multi-level collaborative scheduling module, is used to predict and resolve spatiotemporal conflicts of vehicles at future path nodes based on the task plans of each vehicle determined after the scheduling response. The instruction execution and feedback module communicates with the conflict resolution module and each unmanned delivery vehicle. It is used to issue and execute the final instruction after conflict resolution, and to return the actual status information fed back by the vehicle to the dynamic perception module to update the dynamic pressure field map.
[0016] Therefore, the present invention employs the above-mentioned distribution control system and method suitable for unmanned delivery vehicles, and the beneficial technical effects are as follows: (1) This invention constructs a dynamic pressure field map that integrates real-time order and traffic data, and uses it as the unified environmental perception basis for all scheduling decisions. This enables the system to directly and quantitatively perceive the spatiotemporal changes in order pressure and traffic resistance within the delivery area. This overcomes the shortcomings of existing methods that rely on static models and are difficult to respond to real-time dynamic disturbances, and provides a reliable basis for achieving accurate environmental adaptive scheduling.
[0017] (2) This invention identifies scheduling deadlocks by calculating and monitoring the pressure relief rate of each dynamic responsibility area, and automatically triggers a multi-level collaborative response mechanism that migrates from local tasks to global responsibility redistribution. This design enables the system to not only temporarily adjust tasks when facing local congestion or order backlog, but also reconstruct the global resource allocation pattern when necessary. This effectively solves the problems of rigid resource scheduling and lack of elastic recovery capability in existing methods, and improves the overall robustness and resource utilization efficiency of the cluster in response to emergencies.
[0018] (3) Before the final instruction is issued, the present invention introduces a spatiotemporal conflict resolution step based on the task scheme, which can predict and coordinate potential conflicts of vehicles at future path nodes. This realizes deep collaboration of vehicles at the path resource level, avoids congestion or waiting of vehicles at key nodes, and thus further ensures the reliable execution of the scheduling scheme and the overall operating efficiency in complex operating environments. Attached Figure Description
[0019] Figure 1 This is a flowchart of a distribution control method applicable to unmanned delivery vehicles according to the present invention; Figure 2 This is an architecture diagram of a distribution control system for unmanned delivery vehicles according to the present invention. Detailed Implementation
[0020] The technical solution of the present invention will be further described below with reference to the accompanying drawings and embodiments.
[0021] Unless otherwise defined, the technical or scientific terms used in this invention shall have the ordinary meaning as understood by one of ordinary skill in the art to which this invention pertains.
[0022] Example 1 This embodiment details the implementation process of the method proposed in this invention. This method is executed on a control server with computing and communication capabilities, which maintains real-time data connections with the order management system, the traffic data platform, and all unmanned delivery vehicles.
[0023] Specific steps are as follows Figure 1 As shown.
[0024] Step S1: Obtain order data and road condition data in the delivery area in real time, calculate the real-time pressure value of each location in the area, and construct a dynamic pressure field map.
[0025] The control server continuously executes this step, which is implemented as follows: (1) Data Acquisition: The server subscribes to and receives the order stream to be delivered from the order management system in real time through the API interface. Each data entry contains at least the target geographical coordinates of the order. At the same time, the server obtains the real-time average driving speed or congestion index of the road network within the delivery area through traffic data services.
[0026] (2) Environmental modeling and stress calculation: The server maintains a digital map of a delivery area and discretizes it into uniformly sized grid cells. Each grid cell serves as the basic unit for stress calculation.
[0027] For each grid At any calculation time The system counts the number of all pending orders falling within this grid, and records them as follows: .
[0028] Based on the real-time vehicle speeds of the road segments associated with the grid, a pre-set "speed-congestion coefficient" mapping table is queried to obtain the real-time congestion coefficient reflecting the current traffic difficulty. This mapping relationship is usually set as an inverse relationship, that is, the lower the vehicle speed, the higher the congestion coefficient, and the value range is within the interval [0, 1].
[0029] Calculate its real-time pressure value : ; in, express Time Grid The number of orders awaiting delivery within the specified quantity. This represents the preset maximum order capacity for a single grid, used to normalize the number of orders. Its value can be set based on the historical maximum order load of a single grid or experience.
[0030] (3) Map construction and updating: The real-time pressure values obtained from all grid calculations together form a matrix covering the entire delivery area, namely the "dynamic pressure field map". This map is the core environmental model for all subsequent scheduling decisions and is updated frequently (e.g., once per second or every few seconds) as new orders arrive, orders are completed and road conditions change.
[0031] Step S2: Based on the dynamic pressure field map and the real-time location of each available unmanned delivery vehicle, assign a dynamic responsibility area to each available unmanned delivery vehicle, where the location within the area has a pressure value higher than a set threshold.
[0032] This step is triggered whenever a vehicle becomes available or when the system periodically adjusts its regional availability. Input preparation: Obtain the latest dynamic pressure field map and the real-time GNSS positions of all unmanned delivery vehicles in the "available" state (idle or about to complete their mission), and map them onto the corresponding grid.
[0033] Execute the region growing algorithm: Perform the following iterative process independently for each available vehicle to delineate its own dynamic responsibility region: The grid where the vehicle is currently located is used as the "growth seed".
[0034] Check all neighboring grids of the seed grid (usually defined using 4-connected or 8-connected neighborhoods). Determine if neighboring grids meet two conditions: Condition 1, their pressure value is greater than a preset threshold. ( (A positive empirical value is used to filter out "high-pressure" areas that need service); Condition 2: the grid has not yet been assigned to other vehicles.
[0035] All adjacent grids that meet the above conditions are included in the current vehicle's responsibility area set.
[0036] The newly included grid is used as a new "growth seed," and the process of neighborhood exploration and condition judgment described above is repeated.
[0037] The algorithm terminates when no new grid is included in a given iteration. At this point, the vehicle's area of responsibility consists of all the included grids, forming a connected region that covers the "high-voltage zone" near the vehicle's current location.
[0038] The result of this step is that each available vehicle acquires a real-time, clearly defined service area. This area is not fixed but dynamically changes with the distribution of the pressure field and the vehicle's own position.
[0039] Step S3: Based on the divided dynamic responsibility areas, newly generated delivery orders are assigned to unmanned delivery vehicles in the corresponding responsibility areas, and the pressure relief rate of each dynamic responsibility area is calculated based on the changes in the dynamic pressure field map.
[0040] Step S31, for dynamic responsibility areas ,exist Record the set of real-time pressure values of all grids within its region calculated in step S1. ; Step S32, after the monitoring time interval Afterwards, Step S1 is called again at any time to obtain the dynamic responsibility area. The set of real-time pressure values after all grid updates. ; Step S33: Calculate the dynamic responsibility area exist Pressure dissipation rate over time : ; in, This indicates traversing the dynamic responsibility region. Index of all grids within.
[0041] Step S4: Based on the pressure relief rate, identify dynamic responsibility areas where the pressure relief rate is continuously lower than a preset threshold, and determine that the area has entered a scheduling deadlock state.
[0042] The system maintains a preset rate threshold. This serves as a standard for judging whether a region is operating normally.
[0043] At the end of each monitoring cycle, the system will calculate the results for each dynamic responsibility area. and Compare them.
[0044] If a certain area (e.g., area) The pressure digestion rate was lower than [a certain value] for several consecutive cycles (e.g., 2-3 cycles). The system then determines that the area The system has entered a "scheduling deadlock" state. This indicates that the current processing capacity of the vehicles responsible for the area (which may be affected by routes, load, and local congestion) is insufficient to effectively handle the pressure generated in the area, requiring external intervention.
[0045] Step S5: In response to the scheduling deadlock state determined in step S4, perform the following operations in sequence: Step S51, Local Task Migration: Package and migrate some orders within the dynamic responsibility area that are in a scheduling deadlock to other cooperating vehicles that meet the path fit conditions; Step S511: Obtain all vehicles within a preset range that the current route plan will pass through the scheduling deadlock area as candidate cooperative vehicles; Step S512: For each candidate cooperative vehicle and migration task package Calculate its path fit : ; in, Indicates candidate cooperative vehicles The original planned path was in accepting the migration task package. Additional driving distance required afterward Indicates acceptance of migration task package For candidate collaborative vehicles The total estimated time delay caused by all other orders in the original task sequence, Indicates the preset weighting coefficients; Step S513, Select The candidate collaborative vehicle with the highest value will be selected as the collaborative vehicle to execute the migration task package.
[0046] Step S52, Global Responsibility Redistribution: If the area is still in a scheduling deadlock state after the local task migration is performed, then based on the current dynamic pressure field map, global optimization is performed again and the dynamic responsibility areas of all available vehicles are updated.
[0047] Step S521: Using the dynamic pressure field map generated in the current step S1 as input, minimize the variance of the real-time pressure values of all grids. The objective function is defined as follows: the number of virtual responsibility points equals the total number of currently available vehicles. As constraints, an optimization algorithm is used to calculate an optimal set of virtual responsibility point locations. ,in to They represent the 1st to the 1st. The two-dimensional coordinates of each virtual responsibility point within the delivery area; The optimization algorithm used is simulated annealing, and its energy function is... We can directly use the objective function, that is: .
[0048] Step S522: Calculate the Euclidean distance from the real-time location of each available unmanned delivery vehicle to each virtual responsibility point, and assign a unique virtual responsibility point to each vehicle using an assignment algorithm (such as the Hungarian algorithm) to minimize the total distance from all vehicles to their assigned points. Step S523: Using the virtual responsibility point assigned to each vehicle in step S522 as the new initial point, re-execute step S2 for each vehicle to generate a new dynamic responsibility area.
[0049] Step S6: Based on the vehicle task plan determined after the scheduling response in step S5, predict and analyze the spatiotemporal conflicts of vehicles at future path nodes.
[0050] Step S61: Traverse all shared path nodes. For any two vehicles A and B whose planned time windows overlap, determine that there is a potential conflict. Shared path nodes include road intersections, delivery station entrances and exits, and charging station entrances that are predefined by the system and are prone to congestion or space constraints.
[0051] Step S62: For each conflict, calculate the time that delays the departure of vehicle A. Alternatively, vehicle B could slow down on the road section before the conflict point to increase travel time. The adjustment plan; Step S63: Compare the total adjustment cost The calculation formula is as follows: ; in, and These are the priority weights for the tasks carried by vehicles A and B, respectively; [The text abruptly ends here, likely due to an incomplete sentence or a formatting error.] The minimum necessary adjustments will be implemented. Step S64: Based on the selected adjustment scheme, update the departure time or segment speed in the route planning instructions of the affected vehicles.
[0052] Step S7: Issue and execute the final command after conflict resolution, and return the actual status information fed back by the vehicle to step S1 to update the dynamic pressure field map.
[0053] Command issuance: The control server will issue a set of instructions, including final task allocation, route optimization, and time or speed adjustments, to the corresponding unmanned delivery vehicle via a wireless communication network.
[0054] Vehicle execution and feedback: The unmanned delivery vehicle executes the delivery task according to the instructions and continuously transmits information such as its real location, task status (such as "picked up" or "delivered"), and vehicle status (such as battery level) back to the server through the vehicle network.
[0055] Closed-loop feedback: This real-time feedback information is continuously fed into the data processing flow of step S1. For example, when an order is confirmed as "delivered," it is removed from the number of pending orders in its grid when the dynamic stress field map is constructed, thereby directly reducing the stress value of that grid.
[0056] The invention will be further illustrated below with specific examples.
[0057] 1. Scene and parameter settings.
[0058] Delivery area: one The square area has been gridded into a digital map. A grid, each grid representing The actual area.
[0059] Unmanned vehicles: Three unmanned delivery vehicles are deployed, numbered UDV-1, UDV-2, and UDV-3. Their initial locations are as follows: UDV-1 is in the warehouse (grid index (2,2)), UDV-2 is at location (10,15), and UDV-3 is at location (15,5).
[0060] Preset parameters: Single grid order capacity limit ; pressure threshold ; Pressure dissipation rate monitoring cycle minute; Scheduling deadlock determination rate threshold (Pressure value / minute); The weight for path fit calculation is ; Conflict prioritization and weighting: expedited orders Regular orders ; Speed-Congestion Map: Vehicle speed > 30km / h =0; 20-30km / h> =0.2; 10-20km / h> =0.5; <10km / h> =0.8.
[0061] 2. Deduction of the implementation process of the method.
[0062] At 10:00 AM on a certain day (referred to as time) The system begins to run.
[0063] Step S1 (Construction of Dynamic Pressure Field): Data Input: The system received 5 orders to be delivered, with the target grid distribution as follows: Orders {O1, O2} are sent to grid A (5, 5); Order O3 is sent to grid B (12, 18); Order O4 is sent to grid C (8, 10); and Order O5 is sent to grid D (18, 8). Meanwhile, traffic data shows that due to an accident, the speed on the road section in grid B is only 8 km / h, while the speed in other areas is normal (>30 km / h).
[0064] Pressure calculation: Grid A pressure: ; Grid B pressure: ; Mesh C pressure: ; Grid D pressure: ; The pressure for the remaining grids with no orders is 0.
[0065] Output: Generate A dynamic pressure field map at any given time, where grids A and B represent significant high-pressure points.
[0066] Step S2 (Dynamic Responsibility Area Division): Input: The pressure field map above; vehicle location.
[0067] Execution: The system executes a region growing algorithm (pressure threshold) for each vehicle. ).
[0068] UDV-1: Growth begins at (2, 2). Growth stops when the pressure of adjacent grid cells falls below 0.4. Its dynamic responsibility region... It only includes the grid (2,2) where it is located.
[0069] UDV-2: Growth begins at (10, 15). No grids with pressure > 0.4 were found in its adjacent grids (grid B has high pressure but is not adjacent). Growth stops. Its dynamic responsibility region. It includes its own location and a few surrounding low-pressure grids.
[0070] UDV-3: Growth begins at (15, 5). No meshes with pressure > 0.4 were found in adjacent meshes. Growth stops. Its dynamic responsibility region. It includes its own location and a few surrounding low-pressure grids.
[0071] At this point, neither high-pressure grids A nor B are covered by the responsibility area of any vehicle because they are far from the initial positions of all vehicles and do not meet the "adjacent high-pressure" condition for region growth.
[0072] Step S3 (Task Assignment and Status Monitoring): Task Assignment: The system assigns orders {O1, O2} (target grid A) to vehicles whose responsibility area includes grid A—but there are none at present. According to the rules, the system temporarily suspends these orders, awaiting the next periodic execution of step S2 or the area redrawing after the vehicle moves. Orders O3 (target grid B), O4 (target grid C), and O5 (target grid D) are similarly suspended.
[0073] Pressure dissipation rate calculation: Since it is the initial moment and there is no historical pressure data, the rate calculation has not yet started.
[0074] The periodic execution and evolution of step S2: The system re-executes steps S1 and S2 every 30 seconds (based on updated vehicle location and order status). Assume UDV-1 starts from the warehouse and moves towards the area center. At minute 1, UDV-1 moves to grid (4, 4).
[0075] Repeat step S2: The new position (4,4) of UDV-1 is adjacent to high-pressure grid A (5,5). Using (4,4) as the seed, it is found that the pressure value of the adjacent grid (5,5) is 0.67 > 0.4, so it is merged into it. . Updated to include (4,4) and (5,5).
[0076] Re-execute step S3 (task assignment): Since grid A now belongs to the responsibility area of UDV-1. The pending orders {O1, O2} are immediately assigned to UDV-1. UDV-1 begins planning the path to grid A to execute the task.
[0077] Triggering of steps S4 (scheduling deadlock identification) and S5 (cooperative scheduling response): Scene progression: UDV-1 proceeds to grid A to perform its mission. Meanwhile, in... At the specified time, the system received three new orders {O6, O7, O8} destined for grid B, and congestion in grid B continued. The pressure value for grid B was updated to... It became a super high-pressure point.
[0078] Vehicle movement and area change: UDV-2 moved closer to grid B during its movement. At minute 1, UDV-2 moves to (11, 17), adjacent to grid B (12, 18). After re-executing step S2, grid B is incorporated into UDV-2's responsibility region. Orders O3, O6, O7, and O8 were subsequently assigned to UDV-2.
[0079] A stalemate has formed: UDV-2, fully loaded, heads towards the heavily congested grid B. During the next 5-minute monitoring period ( to Within the area, due to congestion, UDV-2 moved slowly and failed to complete any orders. The system calculated its area of responsibility. (Core grid B) Pressure relief rate: Initial pressure 2.13, ending pressure still approximately 2.13 (orders not completed, congestion continues), rate ≈ 0 << =0.05.
[0080] Deadlock determination: The system is in Determine UDV-2's area of responsibility within minutes. The system has entered a "stalemate" state.
[0081] Step S51 (Local Task Migration): The system selects two orders (e.g., O6 and O7) from the four orders for UDV-2 and packages them into a migration task package. .
[0082] The search revealed that UDV-3 is planning to travel from (15, 5) to grid D (18, 8) to execute order O5, and its path will pass near grid B.
[0083] Calculate UDV-3 contract The fit is as follows: Detouring to grid B to pick up and deliver the package is expected to increase the driving distance by 0.4km, causing a delay of about 3 minutes to its original task O5. The system determined that the fit was acceptable.
[0084] The system will migrate the task package. Assign to UDV-3 and update the task queues for both sides.
[0085] Potential triggers for step S52 (Global Responsibility Redistribution): After another 5 minutes (until) System monitoring revealed that the pressure relief rate in grid B remained extremely low. Meanwhile, UDV-1 had completed its order for grid A and was idle, located on the other side of the area.
[0086] The system determined that the local migration failed to resolve the deadlock, triggering a global redistribution of responsibilities.
[0087] The system ignores the existing vehicle positions, takes the current pressure field (grid B pressure is still extremely high, other areas have low pressure) as input, and uses 3 vehicles as constraints to run an optimization algorithm. The algorithm outputs an optimal set of virtual responsibility point locations, one of which must be very close to the high-pressure core grid B.
[0088] The idle UDV-1 is assigned to the virtual point that is close to grid B using the assignment algorithm.
[0089] Using this virtual point as a new starting point, step S2 is re-executed for UDV-1, generating a completely new responsibility area that includes part of the original grid region B. This means that some of the pressure that was originally accumulated on UDV-2 has been systematically transferred to UDV-1, realizing a global reconfiguration of fleet resources to jointly overcome bottlenecks.
[0090] Step S6 (Spatial-Time Conflict Resolution): Following the scheduling adjustments, forecasts indicate that UDV-2 and UDV-3 will arrive at the same narrow intersection leading to Grid B at approximately the same time.
[0091] The system generates the following adjustment plan: Plan 1, UDV-2 (carrying urgent order O3), Option 1: Delay by 1 minute; Option 2: UDV-3 (carrying a standard migration order, The deceleration caused a delay of 1.2 minutes.
[0092] Calculation cost: Cost of Option 1 Cost of Option Two .
[0093] Decision: Choose option two, which has a lower total cost, and send an instruction to UDV-3, requesting it to slow down appropriately before the intersection to stagger the travel time.
[0094] Step S7 (Instruction Execution and Closed-Loop Feedback): All final instructions (UDV-1 proceeding to the new area of responsibility, the adjusted missions and routes for UDV-2 and UDV-3, and the speed fine-tuning instructions for UDV-3) were issued to each vehicle.
[0095] The vehicle performs its mission. When UDV-3 completes the delivery of the migration package, the system receives real-time status feedback that "orders O6 and O7 have been completed".
[0096] Feedback to step S1: When calculating the dynamic pressure field in the next moment, the number of orders to be delivered in grid B decreases by 2, and its pressure value... This resulted in a significant decrease. Simultaneously, vehicle locations were updated.
[0097] Example 2 like Figure 2 As shown, a distribution control system suitable for unmanned delivery vehicles includes: The dynamic sensing module is used to acquire order data and road condition data in the delivery area in real time, calculate the real-time pressure value of each location in the area, and build a dynamic pressure field map. The responsibility area allocation module, connected to the dynamic perception module, is used to allocate a dynamic responsibility area to each available unmanned delivery vehicle based on the dynamic pressure field map and the real-time location of each available unmanned delivery vehicle. The order scheduling module, connected to the responsibility area allocation module, is used to allocate newly emerging delivery orders to the unmanned delivery vehicles in the corresponding responsibility areas based on the divided dynamic responsibility areas, and to calculate the pressure relief rate of each dynamic responsibility area based on the changes in the dynamic pressure field map. The deadlock determination module, connected to the order scheduling module, is used to identify dynamic responsibility areas where the pressure relief rate is continuously lower than a preset threshold based on the pressure relief rate, and determine that the area has entered a scheduling deadlock state. The multi-level collaborative scheduling module, connected to the deadlock determination module, is used to respond to scheduling deadlock states and sequentially perform the following operations: The local task migration unit is used to package and migrate a portion of orders within a dynamic responsibility area that is in a scheduling deadlock state to other cooperating vehicles that meet the path fit conditions. The Global Responsibility Redistribution Unit is used to perform global optimization and update the dynamic responsibility areas of all available vehicles based on the current dynamic pressure field map when the area is still in a scheduling deadlock state after a local task migration. The conflict resolution module, connected to the multi-level collaborative scheduling module, is used to predict and resolve spatiotemporal conflicts of vehicles at future path nodes based on the task plans of each vehicle determined after the scheduling response. The instruction execution and feedback module communicates with the conflict resolution module and each unmanned delivery vehicle. It is used to issue and execute the final instruction after conflict resolution, and to return the actual status information fed back by the vehicle to the dynamic perception module to update the dynamic pressure field map.
[0098] It is worth noting that all contents not described in detail in this invention are existing technologies and are well known to those skilled in the art.
[0099] Therefore, the present invention adopts the above-mentioned allocation control system and method suitable for unmanned delivery vehicles, which constructs a dynamic pressure field map by integrating real-time order and road condition data to achieve quantitative perception of changes in the delivery environment; automatically identifies the scheduling deadlock state of local areas by calculating and monitoring the pressure relief rate; and based on this state, sequentially executes local task migration to quickly alleviate congestion, and reconstructs the global responsibility area when necessary to completely restore system balance, and finally resolves path node conflicts.
[0100] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and not to limit them. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can still be made to the technical solutions of the present invention, and these modifications or equivalent substitutions cannot cause the modified technical solutions to deviate from the spirit and scope of the technical solutions of the present invention.
Claims
1. A distribution control method suitable for unmanned delivery vehicles, characterized in that, Includes the following steps: Step S1: Obtain order data and road condition data in the delivery area in real time, calculate the real-time pressure value of each location in the area, and construct a dynamic pressure field map; Step S2: Based on the dynamic pressure field map and the real-time location of each available unmanned delivery vehicle, assign a dynamic responsibility area to each available unmanned delivery vehicle; Step S3: Based on the divided dynamic responsibility areas, newly generated delivery orders are assigned to unmanned delivery vehicles in the corresponding responsibility areas, and the pressure relief rate of each dynamic responsibility area is calculated based on the changes in the dynamic pressure field map. Step S4: Based on the pressure relief rate, identify the dynamic responsibility area where the pressure relief rate is continuously lower than the preset threshold, and determine that the area has entered a scheduling deadlock state. Step S5: In response to the scheduling deadlock state determined in step S4, perform the following operations in sequence: Step S51, Local Task Migration: Package and migrate some orders within the dynamic responsibility area that are in a scheduling deadlock to other cooperating vehicles that meet the path fit conditions; Step S52, Global Responsibility Redistribution: If the area is still in a scheduling deadlock state after the local task migration is performed, then based on the current dynamic pressure field map, global optimization is performed again and the dynamic responsibility areas of all available vehicles are updated. Step S6: Based on the vehicle task plan determined after the scheduling response in step S5, predict and analyze the spatiotemporal conflicts of vehicles at future path nodes. Step S7: Issue and execute the final command after conflict resolution, and return the actual status information fed back by the vehicle to step S1 to update the dynamic pressure field map.
2. The allocation control method for unmanned delivery vehicles according to claim 1, characterized in that, Step S1 specifically includes: Step S11: Receive the location information of orders to be delivered pushed by the order system in real time, and obtain the real-time average vehicle speed from the traffic data platform; Step S12: Divide the delivery area into grids, and for each grid... ,exist At any given moment, calculate the real-time pressure value. : ； in, express Time Grid The number of orders awaiting delivery within the specified quantity. This indicates the preset maximum capacity for a single grid order. express Time Grid Real-time congestion coefficient for the corresponding road segment; Real-time congestion coefficient Based on the obtained real-time average vehicle speed, it is determined through a preset speed-congestion mapping table.
3. The allocation control method for unmanned delivery vehicles according to claim 1, characterized in that, Step S2 specifically includes: Step S21: Mark the grid where the target vehicle is currently located as the current seed grid and add it to the dynamic responsibility region set of the vehicle; Step S22: Obtain all adjacent grids of the current seed grid; Step S23: Determine whether the real-time pressure value in the dynamic pressure field map generated by each adjacent grid is greater than the preset pressure threshold. And not included in the dynamic responsibility area of any other vehicle; adjacent grids that meet the conditions will be incorporated into the dynamic responsibility area set; Step S24: Take the newly merged mesh as the new current seed mesh, and repeat steps S22 to S23 until no new mesh meets the merging condition.
4. The allocation control method for unmanned delivery vehicles according to claim 1, characterized in that, Step S3 specifically includes: Step S31, for dynamic responsibility areas ,exist Record the set of real-time pressure values of all grids within its region calculated in step S1. ; Step S32, after the monitoring time interval Afterwards, Step S1 is called again at any time to obtain the dynamic responsibility area. The set of real-time pressure values after all grid updates. ; Step S33: Calculate the dynamic responsibility area exist Pressure dissipation rate over time : ； in, This indicates traversing the dynamic responsibility region. Index of all grids within.
5. The allocation control method for unmanned delivery vehicles according to claim 1, characterized in that, Step S51 specifically includes: Step S511: Obtain all vehicles within a preset range that the current route plan will pass through the scheduling deadlock area as candidate cooperative vehicles; Step S512: For each candidate cooperative vehicle and migration task package Calculate its path fit : ； in, Indicates candidate cooperative vehicles The original planned path was in accepting the migration task package. Additional driving distance required afterward Indicates acceptance of migration task package For candidate collaborative vehicles The total estimated time delay caused by all other orders in the original task sequence, Indicates the preset weighting coefficients; Step S513, Select The candidate collaborative vehicle with the highest value will be selected as the collaborative vehicle to execute the migration task package.
6. The allocation control method for unmanned delivery vehicles according to claim 1, characterized in that, Step S52 specifically includes: Step S521: Using the dynamic pressure field map generated in the current step S1 as input, minimize the variance of the real-time pressure values of all grids. The objective function is defined as follows: the number of virtual responsibility points equals the total number of currently available vehicles. As constraints, an optimization algorithm is used to calculate an optimal set of virtual responsibility point locations. ,in to They represent the 1st to the 1st. The two-dimensional coordinates of each virtual responsibility point within the delivery area; Step S522: Calculate the Euclidean distance from the real-time location of each available unmanned delivery vehicle to each virtual responsibility point, and use an assignment algorithm to assign a unique virtual responsibility point to each vehicle, so as to minimize the total distance of all vehicles to their assigned points. Step S523: Using the virtual responsibility point assigned to each vehicle in step S522 as the new initial point, re-execute step S2 for each vehicle to generate a new dynamic responsibility area.
7. The allocation control method for unmanned delivery vehicles according to claim 6, characterized in that, In step S521, the optimization algorithm used is simulated annealing, and its energy function is... We can directly use the objective function, that is: .
8. The allocation control method for unmanned delivery vehicles according to claim 1, characterized in that, Step S6 specifically includes: Step S61: Traverse all shared path nodes. For any two vehicles A and B whose planned time windows overlap, determine that there is a potential conflict. Step S62: For each conflict, calculate the time that delays the departure of vehicle A. Alternatively, vehicle B could slow down on the road section before the conflict point to increase travel time. The adjustment plan; Step S63: Compare the total adjustment cost The calculation formula is as follows: ； in, and These are the priority weights for the tasks carried by vehicles A and B, respectively; [The text abruptly ends here, likely due to an incomplete sentence or a formatting error.] Implement the minimum necessary adjustments; Step S64: Based on the selected adjustment scheme, update the departure time or segment speed in the route planning instructions of the affected vehicles.
9. A distribution control method for unmanned delivery vehicles according to claim 8, characterized in that, The shared path nodes in step S6 include road intersections, delivery station entrances and exits, and charging station entrances that are predefined by the system and are prone to congestion or limited space.
10. A distribution control system suitable for unmanned delivery vehicles, characterized in that, include: The dynamic sensing module is used to acquire order data and road condition data in the delivery area in real time, calculate the real-time pressure value of each location in the area, and build a dynamic pressure field map. The responsibility area allocation module, connected to the dynamic perception module, is used to allocate a dynamic responsibility area to each available unmanned delivery vehicle based on the dynamic pressure field map and the real-time location of each available unmanned delivery vehicle. The order scheduling module, connected to the responsibility area allocation module, is used to allocate newly emerging delivery orders to the unmanned delivery vehicles in the corresponding responsibility areas based on the divided dynamic responsibility areas, and to calculate the pressure relief rate of each dynamic responsibility area based on the changes in the dynamic pressure field map. The deadlock determination module, connected to the order scheduling module, is used to identify dynamic responsibility areas where the pressure relief rate is continuously lower than a preset threshold based on the pressure relief rate, and determine that the area has entered a scheduling deadlock state. The multi-level collaborative scheduling module, connected to the deadlock determination module, is used to respond to scheduling deadlock states and sequentially perform the following operations: The local task migration unit is used to package and migrate a portion of orders within a dynamic responsibility area that is in a scheduling deadlock state to other cooperating vehicles that meet the path fit conditions. The Global Responsibility Redistribution Unit is used to perform global optimization and update the dynamic responsibility areas of all available vehicles based on the current dynamic pressure field map when the area is still in a scheduling deadlock state after a local task migration. The conflict resolution module, connected to the multi-level collaborative scheduling module, is used to predict and resolve spatiotemporal conflicts of vehicles at future path nodes based on the task plans of each vehicle determined after the scheduling response. The instruction execution and feedback module communicates with the conflict resolution module and each unmanned delivery vehicle. It is used to issue and execute the final instruction after conflict resolution, and to return the actual status information fed back by the vehicle to the dynamic perception module to update the dynamic pressure field map.