A vehicle loading rate and route calculation method

Abstract

The application discloses a vehicle loading rate and route joint calculation method, and particularly relates to the technical field of logistics distribution linkage scheduling, and is used for solving the problems of insufficient coordination of existing vehicle loading and distribution path, leading to unloading conflicts, frequent unloading and poor scheme executability during dynamic changes. By collecting the current unloading state of the vehicle and constructing a residual load execution graph, combining the current position of the vehicle, the remaining to-be-arrived stations, the unloading time period and the distribution change event to generate a path change fingerprint, further calculating the load path mismatch degree, screening the restricted adjustment set and determining the target execution scheme, the vehicle can still be jointly calculated around the residual load state and the residual distribution state when the vehicle has completed part of the distribution and the insertion order, address change, time period change or road detour occurs on the way, unloading conflicts, frequent unloading and route mismatch are reduced, and the scheme executability and operation continuity are improved.

Description

Technical Field

[0001] This invention relates to the field of logistics distribution coordination and scheduling technology, specifically a method for calculating vehicle loading rate and route. Background Technology

[0002] With the development of community retail, chain convenience stores, and on-demand replenishment and delivery services, urban delivery vehicles typically need to complete unloading operations at multiple stations consecutively during a single trip. In existing technologies, vehicle loading optimization and delivery route planning are mostly conducted separately: one approach focuses on improving loading utilization based on cargo volume, weight, and vehicle capacity; the other focuses on determining delivery routes or adjusting the order based on station locations, delivery distances, and pickup times. This type of approach is applicable to generating initial loading plans and initial delivery schedules before vehicles leave the warehouse, but it primarily targets plan generation under static order conditions, typically assuming that loading status and delivery order remain stable during execution.

[0003] However, in actual urban distribution scenarios, after a vehicle completes deliveries to some stations, situations frequently arise such as temporary order insertions, station relocations, adjustments to scheduled pickup times, detours on certain roads, or temporary station refusal to accept deliveries. At this point, the loading location, obstruction relationships, and accessibility of the remaining goods in the vehicle have changed due to previous unloading, and a mismatch can easily occur between the original route sequence and the current remaining load status. If the original loading strategy is still used or the route is only readjusted based on the remaining stations, problems often arise such as goods at subsequent stations blocking goods at previous stations, frequent on-site unloading requiring reordering, forced rearrangement of unloading batches, and difficulty in directly implementing the results of local adjustments. In other words, although existing technologies can handle loading utilization and route arrangement separately, when the vehicle has entered the delivery execution phase and the remaining load status and remaining delivery status change simultaneously, there is a lack of a mechanism for coordinated processing based on the actual remaining load status in the vehicle, the remaining station sequence, and the on-site unloading accessibility.

[0004] Therefore, how to adjust subsequent delivery arrangements in a targeted manner, taking into account dynamic events such as order insertion, address change, time period change, and road detour, when the vehicle has completed partial delivery and the status of the remaining goods in the vehicle has changed, in combination with the loading location, obstruction relationship, unloading accessibility, and remaining station sequence of the remaining goods, and form a remaining station sequence, unloading batch arrangement, and partial reloading plan that can be directly used for on-site execution, has become an urgent technical problem to be solved in the existing technology. Summary of the Invention

[0005] To address the shortcomings of existing technologies, this invention provides a method for calculating vehicle loading rate and route, comprising:

[0006] S1. Collect the station affiliation, loading location, size, weight, and obstruction relationship of the vehicle's currently unloaded cargo, and construct the remaining load execution map;

[0007] S2. Collect the vehicle's current location, remaining destinations, pickup times at each destination, and delivery change events, and generate a route change fingerprint.

[0008] S3. Based on the remaining load execution diagram and path change fingerprint, calculate the unloading conflict amount, cargo transfer amount and time period deviation corresponding to each path adjustment action to form the load path mismatch degree.

[0009] S4. Select path adjustment actions based on load path mismatch. The path adjustment actions must satisfy the segment reachability constraint and not add any occlusion relationships. The segment reachability constraint is: under the adjusted station sequence, there must be at least one continuous passage segment between the target cargo belonging to the previous station and the door passage from the current main location along the unique unloading direction; and generate a restricted adjustment set.

[0010] S5. Perform joint calculations on the restricted adjustment set to determine the target execution plan, which includes the adjusted remaining station sequence, unloading batch sequence, and partial cargo relocation instructions.

[0011] S6. Based on the actual execution records of the target execution plan, revise the screening threshold and segment reachability constraint determination rules corresponding to the load path mismatch.

[0012] Furthermore, S1 includes:

[0013] The cargo part number is used as the primary key to collect information on station affiliation, loading location, size, weight, and obstruction from the front.

[0014] The order of participation is determined by the primary location, while the occupancy of the passageway is determined by the secondary location.

[0015] Determine the obstruction relationship based on the unique unloading direction, and form a directed edge for each pair of obstructing and obstructed components;

[0016] The remaining load execution graph file is generated using vehicle number, wave number, mapping time, rule version number, and parameter version number as header information.

[0017] Furthermore, S2 includes:

[0018] The current location of the vehicle, the remaining destinations to be reached, the pickup time of each destination, and the delivery change events are aligned in a unified time.

[0019] The first station in the current sequence of remaining stations to be reached is taken as the predetermined target station, and the deviation of the vehicle's driving direction is determined by the direction of the line connecting the vehicle's current position to the predetermined target station.

[0020] Using vehicle number, train number, observation window start and end times, rule version number, and parameter version number as header information, a path change fingerprint file is formed in the order of changes in station set, station order, delivery time period, and traffic conditions.

[0021] Furthermore, S3 includes:

[0022] Based on the remaining load execution diagram and path change fingerprint, calculate the unloading conflict amount, cargo reversal amount and time period deviation for each path adjustment action;

[0023] The load path mismatch is determined by prioritizing the unloading conflict amount, then the reloading amount, and finally the time period deviation amount. The load path mismatch is represented by a record that includes path adjustment actions, unloading conflict amount, reloading amount, time period deviation amount, and sorting position.

[0024] Furthermore, the calculation of unloading conflict volume and reverse volume includes:

[0025] Check the path of the target cargo piece leaving the container along the only unloading direction at each station in the adjusted station order;

[0026] Target cargo items that have frontal obstruction, directional obstruction, or passage blockage on the exit passage are included in the unloading conflict quantity.

[0027] Cargo items located on the obstruction chain, whose station sequence is later than the target departure station, and which need to be moved away from the original path before the target cargo item can be released, will be included in the reload volume.

[0028] Furthermore, S4 includes:

[0029] Arrange path adjustment actions in ascending order by load path mismatch degree;

[0030] First, determine the reachability constraints of the segment in the sorting order, then determine whether no new occlusion relationship will be added. If the previous determination is not true, stop the subsequent checks of the corresponding path adjustment action.

[0031] Only path adjustment actions that simultaneously satisfy the segment reachability constraint and do not introduce new occlusion relationships are written into the restricted adjustment set.

[0032] Furthermore, S5 includes:

[0033] The adjustment actions of each path in the restricted adjustment set are calculated together to obtain the adjusted remaining station sequence, unloading batch sequence, and local reloading instructions;

[0034] The target execution plan is determined in the following order: total number of partial unloading instructions, total number of unloading batches, sorting order of original load path mismatch, and path adjustment action number.

[0035] Furthermore, generate unloading batch sequence and partial unloading instructions, including:

[0036] Generate unloading batch order station by station according to the adjusted order of the remaining stations;

[0037] For obstructing cargo items that do not belong to the current batch and cannot be released unless moved, a partial reversal instruction is generated.

[0038] Update the location status of the goods based on the results of the partial repositioning instruction.

[0039] Furthermore, S6 includes:

[0040] Establish an execution checklist chain based on the target execution plan number;

[0041] Based on the execution deviation of the actual execution records in the continuous valid samples relative to the target execution plan, the upper limit of the proportion of low-confidence pending confirmation items, the warning threshold of the number of partial unloading instructions, the warning threshold of the total number of unloading batches, the preset lower limit of passage, the restricted boundary of temporary storage area, the reserved boundary of the first batch of unloading passage, and the acceptable boundary of the new location after continuing to follow the vehicle are corrected.

[0042] The revised results are registered with the rule version number and parameter version number, and a rule revision file is generated.

[0043] Compared with the prior art, the present invention has the following beneficial effects:

[0044] 1. By collecting information on the station affiliation, loading location, size, weight, and obstruction relationships of the goods currently unloaded by the vehicle, a remaining load execution map is constructed. Combined with the vehicle's current location, remaining destination stations, receiving time periods at each station, and delivery change events, a route change fingerprint is generated. Furthermore, the load-route mismatch is calculated, a restricted adjustment set is filtered, and the target execution plan is determined. This enables collaborative calculations based on the remaining load status and remaining delivery status even when the vehicle has completed part of the delivery and there are interruptions such as order insertion, address changes, time period changes, or road detours. This avoids problems such as high on-paper loading rates but on-site unloading conflicts, frequent cargo transfers, and route mismatches, improving the executability of the adjustment results and the continuity of on-site operations.

[0045] 2. By using the actual execution records of the target execution plan, the screening threshold and section reachability rules corresponding to the load path mismatch are corrected in a closed loop. This enables subsequent similar train trips to gradually reduce section reachability misjudgments, local reloading instruction deviations, and unloading batch arrangement deviations under similar vehicle types, route templates, delivery areas, and cargo categories, thereby improving the consistency between plan generation and on-site execution. Attached Figure Description

[0046] Figure 1 This is an overall flowchart of a method for calculating vehicle loading rate and route;

[0047] Figure 2 Construct a schematic diagram for the remaining load execution graph;

[0048] Figure 3 This is a schematic diagram illustrating the generation of path change fingerprints.

[0049] Figure 4 This is a schematic diagram illustrating the formation of load path mismatch.

[0050] Figure 5 A schematic diagram for generating a restricted adjustment set;

[0051] Figure 6 A closed-loop diagram illustrating the determination of the target execution plan and the modification of rules. Detailed Implementation

[0052] 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.

[0053] Example: Combined with Appendix Figure 1-6 This embodiment provides a method for calculating vehicle loading rate and route, including:

[0054] S1. Collect the station affiliation, loading location, dimensions, weight, and obstruction relationships of the vehicle's currently unloaded cargo, and construct the remaining load execution map. The specific implementation is as follows:

[0055] After the vehicle enters the delivery execution phase, the on-board terminal triggers this step at four specific times: each time the unloading at the station is confirmed, the vehicle restarts, a delivery change event is received, and the door is closed again. It obtains the station affiliation, loading location, size, weight, and obstruction relationship of each item from the currently unloaded goods and forms a remaining load execution map. The on-board terminal can be set up in the cab of the delivery vehicle and consists of an industrial control host, a storage unit, a wireless communication unit, and a timing unit. The main execution entity is the loading and calculation service process.

[0056] Ununloaded goods are limited to goods that have completed outbound processing, loading confirmation, but have not yet completed piece-level receipt confirmation and have not been cancelled. Once the goods leave the carriage and are confirmed for handover at the station, they are no longer considered ununloaded goods. If the network does not transmit the data back, the status will be changed first based on the local receipt record on the vehicle terminal. The waybill-level status must not cover the piece-level status.

[0057] The station affiliation is derived from the waybill master file and current delivery wave details issued by the transportation management system. The fields to be read include at least the waybill number, station number, scheduled pickup time, and cargo number. The station affiliation is the only unloading station that the unloaded goods should be delivered to in the current trip. It is written when the vehicle is locked upon leaving the warehouse. When the address is changed, a new station affiliation is generated while retaining the original station affiliation record. When a change of address instruction is received while the door is open, the current record is not immediately rewritten. Instead, it takes effect when the next observation window is opened after the door is closed. The reading rhythm can be set to event trigger plus 60-second polling compensation.

[0058] Loading locations are determined from loading scan records and wagon compartment location markings. The wagon compartment is divided into head, middle, and tail sections along its length; left, middle, and right sides along its width; and bottom, middle, and top sections along its height. The boundary lines are determined by dividing the wagon compartment into equal parts based on its effective internal length, width, and height. Each piece of cargo is assigned a location code during loading scans. This location code indicates the stable placement of the cargo within the wagon compartment, determined by the cargo's outer envelope. The primary location is the area where the geometric center is located. If the geometric center falls on the boundary line, the location is then determined sequentially: closer to the front of the wagon, closer to the left side wall, and closer to the bottom. For cargo located in the direction of the loading platform, cargo exceeding a single loading area is recorded as having a main loading area and an auxiliary loading area. The main loading area is involved in the sequence relationship, while the auxiliary loading area is involved in the channel occupancy relationship. The sequence relationship refers to the determination of the cargo departure inspection order and the order of unloading batches within the same station based on the relative position of the main loading area to the vehicle door, under the condition that the station order is determined after adjustment and the unloading direction is fixed. The channel occupancy relationship refers to the determination of whether the cargo has entered the target cargo departure channel, whether it has compressed the net width of the continuous passage section, and whether it has created a channel blockage, based on the extension of the auxiliary loading area's width to the adjacent loading area. Corresponding auxiliary loading areas... Figure 2 For example goods placed across different locations, only the primary location is used to determine the order of priority, while the secondary location is used to determine the channel occupancy.

[0059] The dimensions and weights are taken from the master data of the goods and the outbound verification records of the warehouse management system. The unit of dimensions is mm and the unit of weight is kg. The actual measured value after packaging is used first. When there are multiple records, they are sorted first by the validity of the status, then by the time close to the time of locking the vehicle for outbound, and then by the calibration level of the weighing equipment. When the difference between the measured value and the archived value exceeds the preset tolerance, the measured value shall prevail and it shall be marked as a specification deviation part.

[0060] In this step, the obstruction relationship is determined based on the unique unloading direction corresponding to the current station. The default is to take the rear door facing outwards. If the vehicle has multiple unloading directions, only the unique unloading direction locked by the dispatch command is used in this round of determination. Its physical meaning is the relationship between the target cargo and the door passageway, where the cargo's outline projection overlaps with the target cargo's outline projection in the passage width direction, and the cargo relationship does not meet the first-unloading condition. Figure 2 In the illustrated relationship, if a right-side cargo item is located in front of the exit passage of the target cargo item in the middle-side position along the only unloading direction to the tailgate, and its outer contour projection enters the passage width range of the target cargo item, then the right-side cargo item can form a frontal obstruction relationship with the target cargo item in the middle-side position. The establishment of this obstruction relationship is based on the spatial relationship between the only unloading direction, the tailgate position, and the exit passage of the target cargo item, rather than solely on the location name of the cargo item.

[0061] In specific judgment, within 5 seconds after the door closes, for each unloaded item, first determine a unique unloading direction, then search for items located outside the door along the direction from the door to the front of the vehicle; if the outer item and the target item have a projection overlap in the width of the passage and the outer item has not been signed for and left the compartment, it is recorded as one forward obstruction relationship; if there is no projection overlap but the reserved width of the passage exceeds the lower limit of the passage judgment, it is recorded as a passage blocking relationship and is included in the forward obstruction relationship for unified recording. The lower limit of the passage judgment follows the preset passage lower limit diameter in the section reachability rule; each pair of obstructing items and the obstructed item forms a directed edge. The same obstructing item can correspond to multiple directed edges, and self-looping edges are prohibited; when a bidirectional conflict occurs, retain the directed edge of the item closer to the door pointing to another item, delete the reverse edge and leave a trace; Appendix Figure 2 The directed occlusion edge in the diagram is used to represent the unidirectional occlusion relationship under the current observation window, the current unique unloading direction, and the current exit passage conditions. It is not represented by a double-headed arrow to indicate simultaneous bidirectional occlusion. When the same pair of cargo items may have opposite directions of obstruction at different deduction stages, the unidirectional directed edge is recorded according to the corresponding stage. If there is a conflict in the relative direction, the unidirectional directed edge is retained according to the direction of the occlusion item closer to the door pointing to the other cargo item, and the reverse edge is deleted.

[0062] The starting point of the observation window is fixed at the moment when the last piece of goods leaves the vehicle and the door closes again at this station, and the ending point is the moment when the vehicle wheel speed is first continuously greater than 0 and remains so for 3 seconds. If the ending point does not occur within 300 seconds, the window will be forcibly closed with 300 seconds as the ending point. After the window is closed, a new observation window will only be reopened when a delivery change event or a door reopening event is received.

[0063] To ensure consistency, the primary key for site affiliation, loading location, size, weight, and obstruction relationship is the cargo part number. If the part number is duplicated, the candidate records are aggregated by idempotent key consisting of waybill number, commodity code, and loading time. Then, within the same idempotent key, they are sorted by a unified timestamp, and only the first valid record is retained. Out-of-order records that are not completed within 2 seconds are placed in the waiting queue and do not participate in this round of mapping.

[0064] During the initial processing, field integrity checks are performed first, followed by time alignment, noise removal, and boundary completion. When a field is missing, the priority order for completion sources is: the most recent loading lock snapshot, manual review records, video review records, and cargo master data. If a field is still missing, it is marked as an undecidable item. Undecidable items are retained in the remaining load execution graph and participate in site affiliation and location occupancy records. Obstruction edges are not automatically generated for these items as start or end points; they are called as high-risk pending confirmation items in the next step. Time alignment is based on the vehicle terminal's time synchronization clock, with the conversion error controlled within 50ms.

[0065] Noise removal is used when two consecutive reports of different locations are made, but the following four conditions are met simultaneously: the door is closed, the parking status is maintained, the working sensor in the carriage is not triggered, and the loading handheld terminal has no login record. In such cases, the report is judged as a false alarm and the previous valid location is used.

[0066] After processing, a remaining load execution graph file is generated and written to the vehicle terminal's local persistent storage, using vehicle number, wave number, mapping time, rule version number, and parameter version number as header information. The rule version number is the version identifier of the currently effective set of judgment rules, and the parameter version number is the version identifier of the currently effective threshold, step size, boundary value, and default value set. The evidence chain log is an overwriteable trace record set formed by appending original records, correction records, anomaly records, version change records, and manual review records in chronological order. The evidence chain index is used to establish the correspondence between related files, log records, and rule versions. The idempotent key is a unique identifier combination used to constrain the same business object to retain only one valid processing result at the same business time point. It is also pushed to the dispatch center message channel for the next step to call. The file content must include at least the cargo node list, station affiliation, location code, size and weight value, occlusion edge list, default mark, specification offset mark and credibility mark. The credibility mark is judged item by item according to four items: data source integrity, time consistency, location continuity and manual review status. All items are satisfied and it is recorded as high, one item is missing and it is recorded as medium, two or more items are missing and it is recorded as low. Low credibility cargo is included in the storage but is not included in the automatic occlusion edge generation.

[0067] The caller reads only by vehicle number and the latest mapping time, and does not allow cross-wave mixing; only one copy of the same summary for the same vehicle at the same time is retained.

[0068] The maximum delay from trigger execution to local write completion can be set to 800ms. Concurrency is executed on a per-vehicle, per-instance basis. When the central receipt times out, the three most recent versions are retained and resent after 2s, 5s, and 10s.

[0069] When network interruption, missing loading snapshot, site ownership conflict, out-of-bounds location code, single item weight exceeding the vehicle's approved single item limit, or cumulative weight exceeding the vehicle's approved load limit occurs, the record will not be deleted, but written to the evidence chain log. The log will at least include the vehicle number, wave number, cargo item number, trigger time, original value, corrected value, execution rule version, operation source, and handling result.

[0070] The security and compliance boundaries are limited to processing only fields necessary for delivery operations, and not collecting private communication content or irrelevant location history.

[0071] The on-site inspection criteria can be set as follows: composition integrity rate, location consistency rate, occlusion judgment consistency rate, and success rate of calling the next step. The sample size can be set as 30 consecutive trips with no less than 50 items per trip.

[0072] Preferably, in the community retail distribution scenario, the vehicle is a 4.2m van, carrying 120 to 180 items per vehicle. The van is divided into 27 zones. The mapping trigger time is within 30 seconds after each station is signed for, the station affiliation polling cycle is 60 seconds, the dimensional tolerance is ±5mm, the weight tolerance is ±0.2kg, and the maximum waiting time for late records is 2 seconds. In the most recent 30 trips with a total of 4380 items, the mapping completeness rate reached 99.3%, the zone consistency rate reached 98.8%, the occlusion judgment consistency rate reached 97.9%, and the average mapping time per trip was 420ms. For example, if a vehicle has 96 items remaining after unloading at the 3rd station, the system generates a remaining load execution graph containing 96 item nodes and 143 occlusion edges, which is directly called in the next step.

[0073] Alternatively, in situations where loading scanning is not feasible, the loading location can be deduced from the video recordings of the time the doors are open and the loading list by the camera on the top of the carriage. Then, the location code can be corrected by a manual verification terminal after unloading at each station. As long as the remaining load execution diagram is generated according to the same station affiliation criteria, the same location division rules, and the same obstruction judgment order, it is an equivalent implementation form that can be implemented in this step.

[0074] S2. Collect the vehicle's current location, remaining destinations, pickup times at each destination, and delivery change events to generate a route change fingerprint. The specific implementation is as follows:

[0075] After the vehicle enters the delivery execution phase, this step is triggered by the on-board terminal at four specific times: each time a station is signed for, the vehicle restarts, the dispatch center issues a delivery change instruction, and the vehicle doors close again. The vehicle continuously obtains its current location, remaining destinations, delivery time slots at each station, and delivery change events from the on-board positioning terminal, transportation management system, station reservation ledger, and dispatch instruction queue, and generates a route change fingerprint based on this. The on-board terminal is located in the driver's cab and consists of an industrial control host, satellite positioning module, inertial measurement module, wheel speed acquisition module, cellular communication module, timing unit, and local storage unit. The execution entity is the route calculation service process, which calls the most recently generated remaining load execution map upstream and provides the route change fingerprint generated in this step to the load path mismatch calculation stage downstream.

[0076] The vehicle's current position is determined jointly using the latitude, longitude, speed, heading angle, and positioning timestamp from the onboard positioning terminal. Latitude and longitude are in degrees, speed is in km / h, heading angle is in degrees, and the timestamp accuracy can be set to 1ms. The vehicle's current position is not directly obtained from a single positioning point, but rather as a valid position result obtained by eliminating drift points within a continuous positioning sequence of the most recent 10 seconds according to a unified time synchronization order. The continuous positioning sequence must contain at least 5 records; if fewer than 5, the previous valid position is used and marked as low confidence. Drift point determination can be set by setting the calculated speed between two adjacent positioning points to exceed [a certain threshold]. Records showing the vehicle's maximum speed and for which the inertial measurement module did not detect a corresponding acceleration change; if multiple records remain after removal, the last valid record is taken as the current position for unified use in this step; when satellite signals are continuously lost, the current position is estimated at a 2-second rhythm, starting from the most recent valid positioning point, along the last valid heading angle and combined with the wheel speed pulse to accumulate the travel distance. If a new valid positioning point is received during the estimation period, the estimation result is overwritten by the new valid positioning point. If the estimation duration exceeds 120 seconds, the current position update is frozen, the confidence level is reduced to low, and a record is left.

[0077] The remaining pending stations are taken from the list of stations in the transportation management system that are currently locked for the current train and have not yet completed piece-level signing. The station list includes at least the station number, station latitude and longitude, station sequence number, station status, scheduled pickup time, number of completed pieces, number of incomplete pieces, and the corresponding cargo piece number set. The station status is limited to five values: pending, arriving, in operation, completed, and canceled. The remaining pending stations are limited to stations with a status of pending, arriving, or in operation and still have corresponding unloaded cargo. When a station has multiple batches of reservations, the station number plus the reservation batch number is used as the unique distinguishing unit. When the address is changed, a new station number record is generated and the original station number is retained. The original station number is no longer counted in the remaining pending stations from the time the address change takes effect.

[0078] The receiving time slots for each station are taken from the reservation fields in the station reservation ledger and waybill details. The start time and end time are recorded in pairs, with the unit being minutes, and are uniformly converted to the time zone of the vehicle terminal's time synchronization clock. Cross-day time slots are recorded continuously from the start time of the day to the end time of the next day. If the start time is later than the end time and there is no cross-day mark, it is judged as an illegal record and the default receiving time slot is called. If there are multiple receiving time slots, they are all retained in chronological order and sorted by the station number plus the time slot number.

[0079] Delivery change events are generated by the dispatch instruction queue, driver operation terminal reports, station receipt messages, and road detour messages. They are limited to eight categories: adding a station, canceling a station, changing a station address, moving the scheduled pickup time forward, moving the scheduled pickup time backward, temporarily refusing pickup at a station, road detour, and vehicle breakdown. Other reported events are not included in the generation of route change fingerprints for this round. Each type of event includes an event number, vehicle number, train number, associated station number, event type, effective time, expiration time, event source, event status, and values ​​before and after the change. The event status is limited to four values: pending, in effect, completed, and withdrawn. The status coverage order is fixed as follows: completed covers in effect, in effect covers pending, and withdrawn only covers pending and in effect, and does not cover completed.

[0080] To ensure consistent reproduction, this step first aligns the vehicle's current location, remaining destination stations, pickup times at each station, and delivery change events according to the vehicle terminal's time synchronization clock. Then, it performs noise reduction, completion, and sequence sorting. The upper limit for time alignment error can be set to 50ms. For records submitted repeatedly within 5 seconds with identical fields, only the first valid record is retained. When the content of the same event number is inconsistent, the records are sorted according to event status priority, time relevance, and source credibility level, and the record with the highest priority is retained. The source credibility level is fixed as dispatch center higher than station receipt, station receipt higher than driver manual report, and is locked according to the rule version number. When the latitude and longitude of a station are missing, the station master file is called first, followed by the station location in the most recent valid delivery record. If the location still cannot be completed, it is marked as a location pending verification station. The location pending verification station is retained in the remaining destination stations and participates in the records of station set changes, station order changes, and pickup time change, but does not participate in the records of traffic condition changes.

[0081] To define the judgment boundaries, this step uses the moment when the most recent item-level receipt was completed and the vehicle door was closed again as the starting point of the observation window, and the moment when the result of this step is first completed and successfully published as readable as the ending point of the observation window. A single observation window is limited to 300 seconds by default. Within the observation window, if a new delivery change event occurs once, or the vehicle's current position deviates from the direction of travel relative to the predetermined target station for 30 seconds, or the absolute change in either the start or end time of the receiving period for any remaining waiting station reaches 10 minutes, then the route change fingerprint reconstruction is immediately triggered. The predetermined target station is fixed as the first station in the current sequence of remaining waiting stations. The predetermined target station is determined after alignment is completed at a unified time and the remaining waiting stations in the current observation window are confirmed. It is not directly generated by the original input fields. The directional deviation is referenced by the direction of the line connecting the vehicle's current position to the predetermined target station. The condition for the deviation is that the angle between the current heading angle and the reference direction is continuously greater than 45° and the interruption recovery time is no more than 5 seconds. A single deviation caused by temporary avoidance with a duration of less than 10 seconds is not counted.

[0082] The route change fingerprint is generated in a fixed order, using a structured field set instead of free text. It first records changes in the set of stations, then changes in the order of stations, then changes in the delivery time, and finally changes in the access conditions. Each layer retains a placeholder field, and no change is recorded as 0. Within the same layer, the stations are arranged in ascending order by station number and the time when the event takes effect. When the same station changes multiple times within the same observation window, the first change value and the final state at the end of the observation window are retained. For additions that are added first and then withdrawn, only the empty placeholder result after withdrawal is retained. For changes that are modified first and then canceled, the cancellation result is retained and the change trajectory index is stored. For moves that are moved first and then moved later, the final state is retained and the cumulative change amount is recorded.

[0083] After recording is completed, a route change fingerprint file is formed using vehicle number, train number, observation window start and end time, rule version number, and parameter version number as header information and written to the local persistent storage of the vehicle terminal. At the same time, it is pushed to the dispatch center's joint computing topic through the message queue for the next step to call. The file must contain at least the current location record, the list of remaining stations to be reached, the list of receiving time periods for each station, the list of delivery change events, the route change fingerprint content, the credibility marker, and the evidence chain index. The credibility marker is judged item by item according to four items: location integrity, station integrity, event consistency, and time period validity. All items are satisfied and it is recorded as high, one item is missing and it is recorded as medium, and two or more items are missing and it is recorded as low.

[0084] To ensure consistency in idempotency, order, and deduplication, the path change fingerprint is composed of an idempotency key consisting of vehicle number, train number, observation window start point, and event summary. Only the first successfully stored version is retained for the same idempotency key. The calling side only reads based on the vehicle number and the latest observation window version, and cross-train number mixing is not allowed.

[0085] The time constraint can be set to an upper limit of 800ms for the delay from the trigger execution to the local write completion. Concurrency is executed on a single vehicle and single instance basis. When network jitter causes duplicate pushes, the dispatch center rejects duplicates by pressing the idempotent key. When the center receipt times out, the three most recent versions are retained and resent at intervals of 2s, 5s, and 10s. When situations such as location signal interruption, missing site master file, conflicting scheduled delivery time slots, simultaneous cancellation and relocation of the same site, and inconsistent road detour message sources occur, the records are not deleted, but are written to the evidence chain log. When the sources are inconsistent, the result issued by the dispatch center is always taken as the standard. When the dispatch center result is missing, the site receipt is taken as the standard.

[0086] The minimum feasible set of interfaces includes a location upload interface, a station list read interface, a delivery time period read interface, a delivery change event upload interface, and a route change fingerprint read interface. The location upload interface must receive at least the vehicle number, latitude and longitude, speed, heading angle, and timestamp. If the vehicle number is missing, it returns code 101; if the timestamp is reversed, it returns code 102; if the location is out of range, it returns code 103. The delivery change event upload interface must receive at least the event number, vehicle number, train number, associated station number, event type, effective time, and source. If the event type is not within the limited set, it returns code 201; if the event number is duplicated but the content is inconsistent, it returns code 202; if the associated station number is missing, it returns code 203. The route change fingerprint read interface must query by vehicle number and train number. If no valid version is found, it returns code 301.

[0087] The safety and compliance boundaries are limited to collecting only the location, station, and dispatch event fields necessary for the delivery business, and not collecting drivers' private communication content or historical trajectories unrelated to the current trip.

[0088] The on-site inspection criteria can be set as the current location effectiveness rate, the consistency rate of remaining stations to be reached, the event merging accuracy rate, the timeliness rate of route change fingerprint generation, and the success rate of calling the next step. Among them, the event merging accuracy rate is calculated based on the proportion of events whose manual review results are consistent with the event status retained by the system. The sampling sample size can be set to 30 consecutive trips with no less than 20 observation windows per trip.

[0089] Preferably, in the community retail urban distribution scenario, the vehicles are 4.2m vans, with each vehicle corresponding to 8 to 15 stations. The location delivery cycle is set to 2 seconds, the threshold for triggering changes in the scheduled pickup time is set to 10 minutes, the determination time for continuous directional deviation is set to 30 seconds, and the upper limit for forced closing of the observation window is set to 300 seconds. The most recent 30 trips have generated a total of 426 observation windows, with the current location effectiveness rate reaching 99.5%, the station consistency rate reaching 99.2%, the event merging accuracy rate reaching 98.7%, and the average time to generate the route change fingerprint being 380ms. For example, after a vehicle completes the signing for the 3rd station, there are still 9 stations to be reached. Within 120 seconds, the dispatch center issues one instruction to add a station and one instruction to move the pickup time of the 6th station back by 20 minutes. After comparing the station list and pickup time in the same observation window, the system generates a route change fingerprint containing the two changes and is directly called by the next step.

[0090] Alternatively, in locations without a continuously online dispatch center, the vehicle terminal can locally cache the station list, pickup time, and delivery change events confirmed by the driver. After communication is restored, the data can be reconciled with the center records. As long as the route change fingerprint is generated according to the same observation window boundary, the same event limit set, and the same ordering rules, it is an equivalent implementation that can be implemented in this step.

[0091] S3. Based on the remaining load execution diagram and path change fingerprint, calculate the unloading conflict amount, cargo transfer amount, and time period deviation corresponding to each path adjustment action to form the load path mismatch degree. The specific implementation is as follows:

[0092] After the route change fingerprint is formed, the route calculation service process immediately calls the latest valid version of the remaining load execution map for the same train number. Combined with the route change fingerprint that remains valid in the current observation window, it calculates the unloading conflict amount, cargo transfer amount and time period deviation for each route adjustment action in the current remaining waiting station sequence, thereby forming the load route mismatch degree. The latest valid version is limited to records that have been completed and have not been covered by subsequent reconstruction, and whose vehicle number, train number and rule version number are consistent.

[0093] The execution location is either the local calculation process of the vehicle terminal or the calculation node of the dispatch center. Both use the same rule version number and parameter version number. It is preferred that the vehicle terminal completes the initial calculation first and the dispatch center completes the verification calculation. If the two results are inconsistent, the one with the newer rule version and the complete evidence chain shall prevail and be recorded.

[0094] The path adjustment action is limited to single-site position adjustment and does not include multi-site combination adjustment; specifically, it includes the insertion position of the new site within the allowed insertion range, the exchange position between the original site and the adjacent site, and the position of the original site being moved backward. The range of the two positions before and after the original position includes the original position itself. The first and last positions are allowed to be used as insertion boundaries. The backward movement is allowed to extend to after the last site but must not cross the current remaining sequence of sites to be reached.

[0095] The set of stations affected by this route change fingerprint is limited to newly added stations, stations with time-period changes, new stations after relocation, stations directly adjacent to canceled stations, and stations directly covered by road detour sections. New stations are only inserted within the continuous station segment corresponding to their service area. The service area is determined by the electronic fence of the delivery area preset in the route template. The continuous station segment is the set of stations in the current remaining to-be-delivered station sequence that are located in the same delivery area and are in consecutive order. When a new station falls on the boundary of two areas, the area to which its latitude and longitude belong shall be used.

[0096] The remaining load execution graph should at least read the cargo part number, station affiliation, primary location, secondary location, forward occlusion relationship, directed occlusion edge, confidence mark, and graphing time. The path change fingerprint should at least read the current location, the list of remaining stations to be reached, the receiving time of each station, the list of delivery change events, and the start and end times of the observation window. All the above records should be aligned with the unified time synchronization of the vehicle terminal, and the time alignment error should follow the 50ms standard of the previous steps. If the train number, vehicle number, or rule version number of the two types of records are inconsistent, the calculation of this round should be stopped and written to the evidence chain log.

[0097] To ensure consistency in project reproduction, this step first associates and organizes the remaining load execution map and route change fingerprints according to the cargo part number and station number, and then establishes a list to be compared according to the order of route adjustment actions. During the organization, cargo parts that have been signed for at the piece level but are still remaining in the remaining load execution map are removed, delivery change events that have been withdrawn and have not been reactivated in the route change fingerprint are removed, and records with multiple batches of reservations at the same station are split into independent arrival units by station number plus reservation batch number.

[0098] In this step, the unloading conflict is fixed on the number of target cargo pieces whose exit is blocked as the sole counting object. Within the current observation window, based on the adjusted station order, all cargo pieces at each station are checked from front to back according to the station order. Then, according to the main location of the cargo piece from the nearest door to the farthest door, the exit path along the unique unloading direction is checked for each piece. If there is a forward obstruction relationship, directed obstruction edge, or passage blocking relationship formed by cargo pieces from subsequent stations on the exit path corresponding to the cargo piece, then the target cargo piece is counted as 1 unloading conflict. Even if the same target cargo piece is blocked by multiple obstructions under the same path adjustment action, it is still only counted as 1 conflict. Passage blocking relationship and forward obstruction relationship are not counted repeatedly. Low confidence cargo pieces are not removed, but a low confidence mark is added when counting unloading conflicts.

[0099] In this step, the number of repositioned goods is fixed as the sole counting object, not the number of repositioning operations. After the unloading conflict is determined, the goods at the station that needs to be dispatched first are located according to the adjusted station order, and the blocking chain is deployed along its departure path. Any goods located on the chain and whose station order is later than the target departure station, and which must be moved out of the original path before the target goods can be released, is counted as 1 necessary repositioned goods. Even if the same goods block multiple target goods under the same path adjustment operation, it is only counted once. After being moved and returned to the original position, it is not counted again. If it only rotates in the original position and does not leave the path, it is not included in the repositioning quantity. After each necessary repositioned goods is identified, it is temporarily removed from the blocking chain in this round of simulation, and the subsequent target goods are checked again until all target goods under the current path adjustment operation have been checked.

[0100] The time-period deviation in this step is the total deviation of the estimated arrival time of each station from the corresponding receiving time period under the target path adjustment action, with the unit fixed in minutes. The internal extrapolation accuracy is in seconds and is converted to minutes when writing records. It is obtained based on the current location signal, the adjusted station order, the receiving time period of each station, the estimated travel time of adjacent stations, and the average operation time of stations. The estimated travel time of adjacent stations takes priority to the real-time passage results in the current observation window. If it is missing, the historical valid records of the same route are used. If it is still missing, the default travel time of the route is used. The average operation time of stations takes priority to the average operation time of the stations completed by the current train. If it is missing, the historical valid records of the same type of station on the same route are used. If it is still missing, the default operation time of the route is used.

[0101] During the simulation, the estimated arrival time from the current position to the adjusted first station is calculated first. Then, the estimated travel time between adjacent stations and the average operation time of each station are added in the adjusted order. If the estimated arrival time of a station is earlier than the start time of the receiving period, the waiting time is included in the time period deviation, and this waiting time is carried over to the estimated arrival time simulation of subsequent stations. If the estimated arrival time is later than the end time of the receiving period, the late arrival time is included in the time period deviation. The deviation of each segment is not counted if it falls within the receiving period. If there are multiple receiving periods at the same station, the segment with the smallest deviation is included in the accumulation. If the smallest deviations are tied, the segment with the earlier start time is selected.

[0102] In this step, the load path mismatch is fixed as a unified evaluation record including path adjustment actions, unloading conflict amount, reloading amount, time period deviation amount, and ranking position. No arbitrary weighting is used, and no unpublished external scoring model is introduced. When comparing, the unloading conflict amount is judged item by item with a fixed priority: first unloading conflict amount, then reloading amount, and then time period deviation amount. When the unloading conflict amount of two path adjustment actions is different, the one with the smaller unloading conflict amount is better. When the unloading conflict amount is the same but the reloading amount is different, the one with the smaller reloading amount is better. When the unloading conflict amount and the reloading amount are the same, the time period deviation amount is compared. When all three items are the same, the path adjustment action with the smaller distance from the original ranking position is retained first. When the distance is the same but they are one before the other, the one with the smaller forward movement is retained first. When a new station has no original ranking position, the position closer to the middle of the continuous station segment in the same service area is retained first.

[0103] After the calculation is completed, a mismatch record file is formed with the vehicle number, train number, observation window number, route adjustment action number, rule version number, and parameter version number as header information. The file content includes at least the route adjustment action, unloading conflict amount, cargo reversal amount, time period deviation amount, load route mismatch ranking, low confidence pending confirmation item list, and evidence chain index. It is written to the local persistent storage of the vehicle terminal and published to the next step calling topic through the message queue. The mismatch record in the same vehicle and the same observation window is composed of idempotent key composed of vehicle number, train number, observation window number, and route adjustment action number. Only the first successfully written version is retained. If the summary of the repeated calculation results is consistent, the duplicate writing is rejected. If the summary is inconsistent, the latest rule version is retained and the reason for overwriting is recorded.

[0104] The time constraint can be set to a maximum delay of 1200ms from the time the remaining load execution graph and path change fingerprint are read until the mismatch record of all path adjustment actions is stored in the database. Concurrency is executed on a single vehicle and single instance basis. When the number of candidate path adjustment actions exceeds 12, they are first calculated in batches according to the order of distance from the original position from near to far, with no more than 6 positions in each batch and no more than 100ms between batches. When the central receipt times out, the three most recent versions are retained and resubmitted at intervals of 2s, 5s, and 10s.

[0105] If situations occur such as missing edges in the remaining load execution graph, missing first and second stations in the path change fingerprint, illegal receiving time period, missing average operation duration at the station, current location frozen for more than 120 seconds, or duplicate station numbers for the same path adjustment action, this step will not delete the record. Instead, the corresponding path adjustment action will be marked as an undeterminable action and a mismatch record file will be generated. The sorting position will be empty and the next step will skip this position. At the same time, it will be written into the evidence chain log for manual review.

[0106] The minimum feasible set of interfaces includes the remaining load execution graph reading interface, the route change fingerprint reading interface, the mismatch record writing interface, and the mismatch record reading interface. Among them, the mismatch record writing interface should at least receive the vehicle number, train number, observation window number, route adjustment action number, unloading conflict amount, reloading amount, and time period deviation amount. If the route adjustment action number is missing, it will return code 401; if the rule version is inconsistent, it will return code 402; if the observation window number is invalid, it will return code 403. The mismatch record reading interface should at least query by vehicle number, train number, and observation window number. If no valid record is found, it will return code 404.

[0107] The safety and compliance boundaries are limited to processing only the cargo, station, time period, and location fields required for the current delivery calculation, without collecting drivers' private communication content or accessing the full text of historical trajectories unrelated to the current delivery.

[0108] The on-site inspection criteria can be set as follows: mismatch record completeness rate, unloading conflict judgment consistency rate, cargo reversal judgment consistency rate, time period deviation calculation consistency rate, and next step call success rate. The first three items are statistically based on the proportion of path adjustment actions that are consistent with the manual review results and the system results. The sampling sample size can be set to 30 consecutive trips with no less than 10 path adjustment actions per trip.

[0109] Preferably, in the community retail urban distribution scenario, a single 4.2m van has 6 to 12 remaining destinations, and 4 to 10 candidate route adjustment actions within a single observation window. A total of 286 route adjustment actions were extracted from the last 30 trips, achieving a mismatch record completeness rate of 99.4%, a 98.6% consistency rate in unloading conflict judgment, a 97.8% consistency rate in reloading volume judgment, and a 98.9% consistency rate in time period deviation calculation. The average calculation time per observation window is 640ms. For example, if a vehicle has 8 remaining destinations after completing the 3rd station, adding one new insertion station generates 5 candidate route adjustment actions. Inserting the new station before the 4th station generates 6 unloading conflicts, 3 reloading volumes, and an 18-minute time period deviation. Inserting it after the 6th station generates 2 unloading conflicts, 1 reloading volume, and a 12-minute time period deviation. Based on this, the system generates corresponding mismatch records and ranks the latter as a better option for direct use in the next step.

[0110] Alternatively, in sites lacking onboard first-round calculation capabilities, the dispatch center's calculation nodes can use the uploaded remaining load execution map and path change fingerprint to calculate the unloading conflict amount, reloading amount, time period deviation amount, and load path mismatch degree for all path adjustment actions. As long as the mismatch degree record is generated according to the same path adjustment action limit, the same conflict counting rule, the same reloading counting rule, the same waiting propagation rule, and the same priority sorting rule, it is an equivalent implementation form that can be implemented in this step.

[0111] S4. Select path adjustment actions based on load path mismatch. The path adjustment actions must satisfy the segment reachability constraint and not create new occlusion relationships. The segment reachability constraint is: under the adjusted station sequence, there must be at least one continuous passage segment between the target cargo piece belonging to the preceding station and the door passage along the unique unloading direction from the current main location; and generate a restricted adjustment set, specifically implemented as follows:

[0112] After the load path mismatch record is formed, the path calculation service process immediately reads the latest valid version of the mismatch record file, the remaining load execution diagram, and the path change fingerprint of the same train number within the same observation window. It then performs a screening of each path adjustment action, retaining the path adjustment actions that meet the segment reachability constraints and do not create new occlusion relationships, and generates a restricted adjustment set accordingly. The latest valid version is limited to records that have been completed and have not been covered by subsequent reconstructions, and whose vehicle number, train number, rule version number, and observation window number are consistent.

[0113] The execution location can be set in the local computing process of the vehicle terminal or the computing node of the dispatch center. Both use the same rule version number and parameter version number. It is preferred that the vehicle terminal first generates the first restricted adjustment set and the dispatch center completes the consistency verification. If the two are inconsistent, the one with the newer rule version and the complete evidence chain shall prevail and be recorded.

[0114] The load path mismatch criteria used in this step include at least the path adjustment action number, unloading conflict amount, cargo reversal amount, time period deviation amount, sorting position, low confidence pending confirmation list, and undeterminable action marker; the remaining load execution graph includes at least the cargo item number, station affiliation, primary location, secondary location, forward occlusion relationship, directed occlusion edge, and confidence marker; the path change fingerprint includes at least the current location, remaining pending station list, receiving time period for each station, delivery change event list, and observation window start and end time. All of the above records are aligned with the unified time synchronization of the vehicle terminal. If the vehicle number, train number, rule version number, or observation window number is inconsistent, the current round of screening will be terminated and written to the evidence chain log.

[0115] The segment reachability constraint in this step is limited to: for any path adjustment action, under the adjusted station sequence, there must be at least one continuous passage segment between the target cargo belonging to the previous station and the door passage along the unique unloading direction from the current main location; the unique unloading direction follows the unloading direction of the current station that was locked when the remaining load was used to construct the graph, and will not be changed again in this step.

[0116] The continuous passage section is divided into segments according to the adjacent location boundaries along the length of the carriage. Each segment is determined by the net width that the target cargo can pass through within that length of the location. The net width is obtained by subtracting the width of the attached location extension and the width corresponding to the passage occupancy status from the total width of the segment. The passage occupancy status is taken from the remaining load execution diagram and limited to three values: unoccupied, partially occupied, and fully occupied. The width corresponding to full occupancy is equal to the total width of the segment, and the width corresponding to partial occupancy is less than the total width of the segment but greater than 0.

[0117] During the comparison, each segment is checked from the door to the front of the vehicle. If the net width of any segment is continuously less than the preset lower limit of passage, the entire passage is judged not to meet the segment reachability constraint. The preset lower limit of passage is taken from the vehicle configuration table and operation specifications, and is determined by adding the unloading operation margin to the outer width of the target cargo, in mm.

[0118] If there are multiple exit routes for the target cargo, only the route with the shortest length that is consistent with the single unloading direction will be retained for the determination. If the shortest routes are parallel, the one with the larger average net width will be retained first. If the average net widths are still the same, the one closer to the center line of the door will be retained.

[0119] To ensure the uniqueness of the judgment result, no new occlusion relationship is added. In this step, the number of newly added directed occlusion edges after adjustment is fixed as the criterion. The comparison scope is limited to the full set of directed occlusion edges related to the cargo items of the stations associated with the current path adjustment action. During the comparison, the corresponding directed occlusion edges in the current remaining load execution diagram are first used as the reference set. Then, without changing the physical location of the cargo items, the departure order relationship of the corresponding cargo items is reconstructed only according to the adjusted station order, and the adjusted set is regenerated. Any directed occlusion edge that only appears in the adjusted set but not in the reference set is counted as a newly added occlusion edge. Only one edge is counted for the same pair of occluding and occluding items, and they are not counted repeatedly. When the original edge direction is reversed, it is treated as adding one reverse edge and deleting the original direction edge. The new judgment only counts the number of newly added reverse edges. If the channel blocking relationship changes but no new directed occlusion edge is formed, it is not counted as a newly added occlusion relationship.

[0120] To ensure consistency in project reproduction, this step first performs preliminary sorting before proceeding to screening. During preliminary sorting, path adjustment actions marked as undecidable actions in the mismatch record are first removed, followed by path adjustment actions where the proportion of low-confidence pending confirmation items exceeds a preset upper limit. The proportion is calculated by dividing the number of low-confidence pending confirmation items by the total number of cargo items associated with the path adjustment action, expressed as a percentage. Actions equal to the preset upper limit are allowed to be retained, preferably set to 20%. Subsequently, the remaining path adjustment actions are sorted in ascending order by load path mismatch ranking. Actions with the same ranking are sorted from closest to furthest from their original ranking distance, and those with the same original ranking distance are sorted in ascending order by path adjustment action number. For the same station with three types of path adjustment actions (insertion, exchange, and extension) existing simultaneously within the same observation window, only the action with the best ranking is retained for subsequent constraint determination, while the remaining actions are recorded but do not participate in the generation of the constrained adjustment set in this round.

[0121] During the screening process, path adjustment actions are checked one by one in the above sorting order. First, the reachability constraint of the segment is determined, and then it is determined whether no new occlusion relationship is added. If the previous condition cannot be determined due to missing data, it is directly converted into an action that cannot be included in the set and subsequent checks are stopped. If the previous condition is met but the subsequent condition cannot be reconstructed due to the empty reference set and the missing status of the main location, the secondary location, or the channel occupancy, it is also converted into an action that cannot be included in the set.

[0122] The path adjustment action is written into the restricted adjustment set only if both the preceding and following conditions are met.

[0123] To prevent execution conflicts caused by multiple actions existing at the same site, only one path adjustment action is allowed to be retained at the same site in the restricted adjustment set. If a later action corresponds to the same site as an action already included in the set, their load path mismatch ranking is compared, and the one with the better ranking is retained. If the rankings are the same, the one with the smaller path adjustment range is retained. The path adjustment range is determined by the absolute difference between the adjusted position and the original position. When a new site does not have an original position, it is determined by the absolute difference between the insertion position and the center position of the continuous site segment in the same service area. The center position is the integer median of the first and last positions of the continuous site segment. If the ranking and range are the same, the one with the smaller path adjustment action number is retained.

[0124] After the filtering is completed, a restricted adjustment set file is formed with the vehicle number, train number, observation window number, rule version number, and parameter version number as header information. The file content includes at least the path adjustment action number, corresponding station number, action type, adjusted sequence, marker that meets the segment reachability constraint, marker that no new occlusion relationship is added, original load path mismatch ranking, and evidence chain index. It is written to the local persistent storage of the vehicle terminal and published to the next step calling topic through the message queue. The restricted adjustment set of the same vehicle and the same observation window is composed of the vehicle number, train number, and observation window number as idempotent keys. Only the first successfully written version is retained. If the summary of the subsequent repeated filtering results is consistent, it is rejected to write repeatedly. If the summary is inconsistent, the newer rule version is retained and the reason for overwriting is recorded.

[0125] Restricted adjustment sets are allowed to be empty; if they are empty, a restricted adjustment set file is still generated, the action list is set to empty and marked as an empty set version, and when the next step reads an empty set version, it directly enters the branch with no executable actions, without treating it as having no valid version.

[0126] Actions that cannot be included in the set are not written to the restricted set action list, but are only written to the evidence chain log for manual review.

[0127] The time constraint can be set to a maximum delay of 800ms from the time the mismatch record is read until the restricted adjustment set file is written to the database, and concurrent execution is performed on a single vehicle and single instance basis; when the number of path adjustment actions to be filtered exceeds 10, they are filtered in batches according to the sorting order, with no more than 5 actions in each batch and no more than 100ms between batches; when the central receipt times out, the three most recent versions are retained and resubmitted at intervals of 2s, 5s, and 10s.

[0128] If the following situations occur: the remaining load execution map lacks the main location, the path change fingerprint lacks the first record of the remaining unreached station, the lower limit of passage is missing, the station number corresponding to the restricted adjustment action is duplicated, the benchmark set is empty and the main location, secondary location or channel occupancy status is missing, resulting in the inability to rebuild, this step will not delete the record, but will mark the corresponding path adjustment action as an action that cannot be included in the set and write it into the evidence chain log.

[0129] The minimum feasible set of interfaces includes the mismatch record reading interface, the remaining load execution graph reading interface, the route change fingerprint reading interface, the restricted adjustment set writing interface, and the restricted adjustment set reading interface. Among them, the restricted adjustment set writing interface should at least receive the vehicle number, train number, observation window number, route adjustment action number, station number, and adjusted sequence number. If the route adjustment action number is missing, it will return code 501; if the rule version is inconsistent, it will return code 502; if the observation window number is invalid, it will return code 503. The restricted adjustment set reading interface should at least query by vehicle number, train number, and observation window number. If no valid version is found, it will return code 504.

[0130] The safety and compliance boundaries are limited to processing only the cargo, station, location, time period, and action fields required for the current train's data processing, without collecting drivers' private communication content or accessing the full text of historical trajectories unrelated to the current train.

[0131] The on-site inspection criteria can be set as follows: completeness rate of restricted adjustment set generation, consistency rate of segment reachability constraint judgment, consistency rate of judgment without adding occlusion relationship, and success rate of next step call. The first two items are statistically based on the proportion of path adjustment actions that are consistent with the results of manual review and system review. The sample size can be set to 30 consecutive trips with no less than 8 path adjustment actions per trip.

[0132] Preferably, in the community retail urban distribution scenario, there are 4 to 9 path adjustment actions to be screened within a single observation window of a 4.2m van. The upper limit of the proportion of low-confidence pending confirmation items is set at 20%. A total of 214 path adjustment actions were extracted from the most recent 30 trips. The completeness rate of the restricted adjustment set generation reached 99.1%, the consistency rate of segment reachability constraint judgment reached 98.4%, and the consistency rate of judgment without adding occlusion relationships reached 98.0%. The average single observation window screening time was 430ms. For example, after a vehicle completes the 3rd station, it generates 5 path adjustment actions. Although the 2nd action has the best ranking, after the adjustment, it causes 2 items of goods in the rear area to occupy the only path of the target goods of the preceding station, which does not meet the segment reachability constraint, so it is not included in the set. The 3rd action meets the segment reachability constraint, but after re-deduction, 2 directed occlusion edges are added, so it is not included in the set. The 4th action meets both the segment reachability constraint and does not add occlusion relationships, so it is finally included in the restricted adjustment set and can be directly called in the next step.

[0133] Alternatively, in areas where on-board screening is not available, the dispatch center's interconnected nodes can screen all path adjustment actions based on the mismatch records of the same train, the remaining load execution diagram, and the path change fingerprint. As long as the restricted adjustment set is generated according to the same segment reachability constraint judgment criteria, the same newly added occlusion edge comparison range, the same single-site unique action retention rule, and the same empty set handover rule, it is an equivalent implementation form that can be implemented in this step.

[0134] S5. Perform joint calculations on the restricted adjustment set to determine the target execution plan. The target execution plan includes the adjusted remaining station sequence, unloading batch sequence, and partial reloading instructions. The specific implementation is as follows:

[0135] After the constrained adjustment set is formed, the route calculation service process immediately reads the latest valid version of the constrained adjustment set file, the remaining load execution diagram, the route change fingerprint, and the load-route mismatch record for the same train in the same observation window. It then performs the calculation on each route adjustment action in the constrained adjustment set in sequence to determine the target execution plan. The calculation is limited to performing correlation deduction and joint determination on the adjusted remaining station sequence, the corresponding unloading batch sequence, and the local cargo reversal instructions under a single route adjustment action.

[0136] The execution location is either the local computing process of the vehicle terminal or the computing node of the scheduling center. Both use the same rule version number and parameter version number. It is preferred that the vehicle terminal first generates the target execution plan and the scheduling center completes the consistency verification. If the results of the two are inconsistent, the one with the newer rule version and the complete evidence chain shall prevail and be recorded.

[0137] The restricted adjustment set should at least include the path adjustment action number, corresponding station number, action type, adjusted sequence, marker indicating that the reachability constraint of the section is met, marker indicating that no new occlusion relationship is added, and the original load path mismatch ranking; the remaining load execution graph should at least include the cargo part number, station affiliation, primary location, secondary location, forward occlusion relationship, directed occlusion edge, and confidence marker; the path change fingerprint should at least include the current location, the list of remaining stations to be reached, the receiving time period of each station, the list of delivery change events, and the start and end times of the observation window; the load path mismatch record should at least include the path adjustment action number, unloading conflict amount, reloading amount, time period deviation amount, and ranking.

[0138] All the above records are aligned with the unified time synchronization of the vehicle terminal. If the vehicle number, train number, rule version number, or observation window number is inconsistent, the current round of joint calculation will be terminated and written to the evidence chain log.

[0139] During the initial cleanup, path adjustment actions marked as invalid in the restricted adjustment set are first removed. Invalidity is limited to actions where the corresponding station has completed the item-level acceptance, the corresponding station has been canceled or relocated and the original action no longer corresponds to the current station, the observation window has been switched to the updated version, or the restricted adjustment set of the same train has been updated to a new version. Then, multiple historical execution results corresponding to the same station are deduplicated according to the observation window number, generation time and version status, and only the latest valid record in the current observation window is retained.

[0140] For stations where the proportion of low-confidence pending confirmation items in the remaining load execution diagram exceeds the preset upper limit, participation in the joint calculation is not prohibited, but a manual confirmation mark is added to the joint calculation result. The proportion is obtained by dividing the number of low-confidence pending confirmation items at the station by the total number of unloaded items at the station, and the unit is %. If it equals the preset upper limit, it is allowed to be retained. Preferably, the preset upper limit can be set to 20%.

[0141] During the joint calculation, the adjusted order of the remaining stations is generated first based on the adjusted sequence corresponding to each path adjustment action. The order is a unique sequence obtained by performing only the insertion, exchange, or delay of the stations corresponding to the path adjustment action on the current list of remaining stations to be reached.

[0142] When the restricted adjustment set is empty, the target execution plan file is still generated. The order of the remaining stations after adjustment follows the current list of remaining stations to be reached. The unloading batch order is set to empty, the local dumping instruction is set to empty, and it is marked as the default execution plan version. This version participates in downstream execution calls, but does not participate in multi-action adjudication.

[0143] The unloading batch sequence is carried out station by station according to the adjusted remaining station sequence. Within each station, the batches are generated in order of the main location of the cargo from the nearest door to the farthest door, and in the same location, they are generated in order of the first release and the last obstruction according to the obstruction relationship. The batch formation rule is fixed as follows: any cargo that can leave the container directly along the only unloading direction without the need for a new local reloading instruction in the current state is included in the same unloading batch. Once a cargo must wait for the previous batch of cargo to leave the container, wait for the local reloading to be completed, or wait for the passage to meet the section reachability constraint again before it can leave the container, it will be transferred to the next unloading batch.

[0144] After each unloading batch is completed, based on the temporary location status of the goods that have left the container and the partial unloading instructions that have been executed, it is re-determined whether the remaining goods can be directly unloaded, and then the next batch is generated. Goods that are closer to the center line of the door are given priority to be included in the previous batch when they are released at the same time. If they are still the same, they are arranged in ascending order of the goods number.

[0145] The release point is fixed at the moment when the preceding goods blocking the cargo have finished leaving the container or the corresponding partial unloading instruction has been executed, and the departure path first continuously satisfies the section reachability constraint.

[0146] The fixed partial relocation instruction is: a temporary relocation instruction that must be implemented for obstructing cargo items in order to ensure that the unloading batch sequence can be executed, without changing the remaining station sequence after the adjustment.

[0147] For each station, check each batch of obstructing cargo that is still blocking subsequent cargo to be unloaded in the order of unloading batches. If the obstructing cargo does not belong to the current batch of cargo to be unloaded and the target cargo cannot be released without being moved, generate one partial relocation instruction. Even if the same cargo blocks multiple target cargo during the operation at the same station, only one partial relocation instruction will be generated.

[0148] Each partial relocation instruction includes at least the station number, cargo item number, trigger batch number, original main location, temporary storage location, and return requirements. The availability conditions for the temporary storage location are fixed as follows: it is not occupied, it is not located on the first batch unloading passage of the current station, it can accommodate the outer dimensions of the cargo item, it meets the load-bearing boundary, and it does not occupy the reserved passage for subsequent batches of the current station. The available location is selected first on the same side as the original main location and closer to the vehicle door. When there are multiple available locations, the one with the shorter relocation distance is retained. If the relocation distance is the same, the location with the smaller number is retained.

[0149] The first unloading route at the current station is fixed as the unique unloading route corresponding to the cargo piece closest to the center line of the vehicle door in the first batch of cargo at the current station.

[0150] The rules for choosing between repositioning or continuing to accompany the vehicle are fixed as follows: if the station to which the blocked cargo belongs still needs to maintain the existing loading order after the current station, then it will be repositioned after all unloading is completed at the current station; if repositioning will block the first batch of unloading passages at the determined subsequent stations again, then it will not be repositioned and will continue to accompany the vehicle as the new main location using the temporary storage area.

[0151] After a partial cargo relocation instruction is executed, the location status of the cargo in the current observation window is updated in real time; if the cargo continues to travel with the vehicle, the temporary storage location is used as the updated main location; if the relocation is completed, the original main location and auxiliary location are restored.

[0152] When there is no available temporary storage location, the corresponding path adjustment action is marked as an unexecutable candidate and written to the evidence chain log. In cases where the remaining load execution graph lacks a primary or secondary location, the set of directed occlusion edges is incomplete, the current list of remaining pending stations is missing, or the unloading batch sequence of the same station cannot be closed, this step will also mark the corresponding path adjustment action as an unexecutable candidate and write it to the evidence chain log, without deleting the original record.

[0153] The target execution plan is fixed as a unified execution record that includes the adjusted remaining station order, unloading batch order, partial reloading instruction list, manual confirmation mark, evidence chain index, and generation time. It does not use arbitrary weighting, nor does it call unpublished external optimization models. When determining the target execution plan, first compare the total number of partial reloading instructions generated after the joint calculation of each path adjustment action. The statistical scope is all remaining stations, and the one with fewer instructions is preferred. If the total number of partial reloading instructions is the same, then compare the total number of unloading batches of all remaining stations, and the one with fewer batches is preferred. If both are the same, then compare the original load path mismatch ranking, and the one with the better ranking is preferred. If they are still the same, the one with the smaller path adjustment action number is retained.

[0154] When a manual confirmation flag is present, it serves only as a pre-execution prompt for the driver's end and the station's operation terminal, and does not prevent the automatic issuance of the target execution plan and the invocation of the next step.

[0155] Unexecutable candidates are not included in the target execution plan, but are retained in the evidence chain log and used as a reference for manual review in the next observation window.

[0156] After the joint calculation is completed, a target execution plan file is formed using vehicle number, train number, observation window number, rule version number, and parameter version number as header information. This file is written to the local persistent storage of the vehicle terminal and published to the execution scheduling topic through a message queue for the next step to call. For the same vehicle and the same observation window, the target execution plan uses the vehicle number, train number, and observation window number as idempotent keys. Only the first successfully written version is retained. If the results of subsequent joint calculations are consistent, duplicate writing is rejected. If the results are inconsistent, the newer rule version is retained and the reason for overwriting is recorded.

[0157] The time constraint can be set to a maximum delay of 1000ms from the start of reading the restricted adjustment set to the completion of the target execution plan file being stored in the database, and concurrent execution is performed on a per-vehicle, per-instance basis; when the number of restricted adjustment actions exceeds 6, they are sorted by the original load path mismatch degree from best to worst and calculated in batches, with no more than 3 actions per batch and no more than 100ms interval between batches; when the central receipt times out, the 3 most recent versions are retained and resubmitted at intervals of 2s, 5s, and 10s.

[0158] The minimum feasible set of interfaces includes the restricted adjustment set reading interface, the remaining load execution graph reading interface, the path change fingerprint reading interface, the target execution scheme writing interface, and the target execution scheme reading interface. Among them, the target execution scheme writing interface should at least receive the vehicle number, train number, observation window number, target execution scheme number, adjusted remaining station order, and partial reloading instruction list. If the target execution scheme number is missing, it returns code 601; if the rule version is inconsistent, it returns code 602; if the observation window number is invalid, it returns code 603. The target execution scheme reading interface should at least query by vehicle number, train number, and observation window number. If no valid version is found, it returns code 604.

[0159] The safety and compliance boundaries are limited to processing only the cargo, station, location, time period, and instruction fields required for the current train's data processing, without collecting drivers' private communication content or accessing the full text of historical trajectories unrelated to the current train.

[0160] The on-site inspection criteria can be set as the completeness rate of target execution plan generation, the consistency rate of unloading batch order judgment, the consistency rate of partial unloading instructions, and the success rate of next step invocation. The first three items are statistically based on the proportion of target execution plans that are consistent with the results of manual review and system review. The sample size for random inspection can be set to 30 consecutive trips with no less than 5 target execution plans per trip.

[0161] Preferably, in the community retail distribution scenario, the number of restricted adjustment actions within a single observation window of a 4.2m van is 2 to 6. A total of 168 target execution plans were extracted from the most recent 30 trips, achieving a target execution plan generation completeness rate of 99.2%, a 98.5% consistency rate in unloading batch sequence determination, and a 97.9% consistency rate in partial reversing instructions. The average single observation window calculation time is 520ms. For example, after completing the 3rd station, a vehicle generates 3 restricted adjustment actions. The 2nd action generates the remaining station sequence as 4th, 6th, 5th, and 7th stations. Under this sequence, the 4th station generates 2 unloading batches and 1 partial reversing instruction, the 6th station generates 1 unloading batch without a partial reversing instruction, and the 5th station generates 2 unloading batches and 1 partial reversing instruction. The system ultimately determines the calculation result corresponding to this action as the target execution plan and allows it to be directly called in the next step.

[0162] S6. Based on the actual execution records of the target execution plan, revise the screening threshold and segment reachability constraint determination rules corresponding to the load path mismatch, specifically as follows:

[0163] After the target execution plan is issued and undergoes actual site operations, the rule correction service process reads the target execution plan file, execution scheduling receipt, site operation terminal record, vehicle door status record, cargo receipt record, and vehicle terminal location time sequence record according to the same train number, same observation window, and same target execution plan number. It merges the actual execution records of the target execution plan and corrects the filtering threshold and segment reachability rules corresponding to the load path mismatch degree accordingly. The execution scheduling receipt includes at least the target execution plan number, issuance time, receipt confirmation time, execution start marker, execution end marker, and reason for refusal to execute.

[0164] The rule correction service process can be deployed on the dispatch center's interconnection node, or it can be deployed in a combination of the vehicle terminal's local backhaul service and the center's review service. The preferred execution entity is the rule management process on the dispatch center side, while the vehicle terminal side is only responsible for collecting the original execution records and the initial backhaul results.

[0165] The actual execution record is limited to the station arrival time, first piece departure time, last piece departure time, start time of each unloading batch, end time of each unloading batch, trigger time of each partial unloading instruction, completion time, repositioning result, continued on-vehicle result, manual confirmation record, and record of reasons for not executing according to the plan, which are directly corresponding to the target execution plan. The above records are respectively taken from the station operation terminal click record, door opening and closing signal, barcode signing time, unloading confirmation button record, and unified time clock of the vehicle terminal. The time unit is ms, the operation time is converted to s, and the position deviation is converted to mm. All are aligned with the unified time of the vehicle terminal. If any of the vehicle number, train number, observation window number, or target execution plan number is inconsistent, the current correction is stopped and written to the evidence chain log.

[0166] To ensure consistency in project reproduction, this step first establishes an execution comparison chain based on the target execution plan number, and then performs deduplication, completion, and sequential organization on various execution records. Among them, for execution records submitted repeatedly within 5 seconds with completely identical fields, only the first valid record is retained. For records with inconsistent content for the same event number, the records are sorted in the following order: source credibility level, time recentity, and status completion level, and the record with the highest priority is retained. The source credibility level is fixed as follows: site operation terminal is higher than vehicle terminal automatic record, vehicle terminal automatic record is higher than manually supplemented record. The status completion level is fixed as follows: completed is higher than in progress, in progress is higher than triggered but not completed, triggered but not completed is higher than only planned, and is locked according to the rule version number. Missing partial reshipment completion time is first filled in with the start time of the next batch. If it still cannot be filled in, it is marked as a missing execution segment, which does not participate in this round of threshold correction but is retained as a record.

[0167] The screening thresholds are limited to the upper limit of the proportion of low-confidence pending confirmation items, the warning threshold for the number of partial reversal instructions, and the warning threshold for the total number of unloading batches; the section accessibility rules are limited to the preset lower limit of passage, the restricted boundary of temporary storage area, the reserved boundary of the first batch of unloading passage, and the acceptable boundary of new location after continuing to follow the vehicle.

[0168] The execution results are fixed as the adjusted remaining site sequence, unloading batch sequence, and partial repositioning instruction list in the target execution plan; the execution deviation is fixed as the difference between the actual execution record and the planned execution results in terms of the number of batches, the number of repositioning items, the order of channel release, and the site completion time.

[0169] The above rules are not arbitrarily modified, but are corrected item by item based on the deviation between the actual execution results and the planned execution results of consecutive valid samples under the same vehicle type, same route template, same delivery area, and same cargo category combination; the cargo category combination is determined by the cargo category with the highest proportion of pieces in the same train; if the highest proportions are tied, the category with the higher volume proportion is determined; the cargo width range is divided into increments of 50mm based on the outer width of the cargo pieces.

[0170] Continuous valid samples are selected from the most recent 30 consecutive trips that meet the following criteria: field completeness rate of not less than 95%, manual data entry ratio of not more than 10%, and target execution plan has been completed and returned in a closed loop. The denominator of the field completeness rate is the total number of fields that should be collected for the sample, and the denominator of the manual data entry ratio is the total number of execution events for the sample. Continuity is determined by the time sequence of valid samples with the same vehicle type, the same route template, the same delivery area, and the same cargo category. Invalid samples are not included and the continuity is not interrupted.

[0171] When making specific corrections, first scan consecutive valid samples one by one according to the rule items; the same sample is allowed to enter the judgment of multiple rule items at the same time.

[0172] If three or more consecutive samples at the same station meet the condition that "the first batch of goods cannot leave the container directly as planned, but actually requires one or more partial unloadings before leaving the container", then the preset lower limit of passage under the corresponding condition will be increased by one preset step. The preset step is determined according to the vehicle configuration table, and can preferably be set to 20mm to 50mm. If three or more consecutive samples meet the condition that "the plan determines that partial unloading is required, but in reality, no unloading is required to complete the first batch of goods leaving the container", then the preset lower limit of passage under the corresponding condition will be decreased by one preset step. However, the decreased value shall not be lower than the minimum safe operating margin of the vehicle model, and the increased value shall not be higher than the maximum allowable passage redundancy value of the vehicle model. The maximum allowable passage redundancy value of the vehicle model is taken from the fixed upper limit jointly determined by the vehicle configuration table and the operating specifications, and the unit is mm.

[0173] Comparing the partial relocation instruction list with the actual relocation execution results, if three or more consecutive samples meet the condition that "the temporary storage location specified by the partial relocation instruction is continuously replaced by the actual operator with an adjacent location on the same side and no new obstruction relationship is added after the replacement", then the priority of the temporary storage location for this type of cargo under this vehicle type and location combination will be adjusted to the adjacent location that appears most frequently and accounts for more than 50%; if three or more consecutive samples meet the condition that "the first batch of passages for cargo that continues to accompany the vehicle are not blocked after actual relocation at the next station", then the acceptable boundary of the new location after continuing to accompany the vehicle will be relaxed by one location level; conversely, if two or more consecutive samples meet the condition that "continuing to accompany the vehicle causes the first batch of passages at the next station to be blocked and additional partial relocation is added", then the acceptable boundary of the new location after continuing to accompany the vehicle will be tightened by one location level.

[0174] The location hierarchy is fixed and defined in the order of near-gate location, middle location, and far-gate location on the same side. It can be relaxed to allow one level backward, or tightened to allow one level forward.

[0175] The screening threshold correction corresponding to the load path mismatch is performed according to the deviation ratio: In the most recent consecutive valid samples, if the proportion of low-confidence pending confirmation items exceeds the current upper limit, but after final manual confirmation, more than 90% of the samples are still executed as planned, and execution as planned is limited to consistent station order and batch difference not exceeding 1 batch and local relocation instruction number difference not exceeding 1. The batch difference and local relocation instruction number difference are obtained by comparing the corresponding fields in the target execution plan file with the actual execution record item by item. Then, the upper limit of the proportion of low-confidence pending confirmation items under the corresponding line template will be increased by 5 percentage points; If the proportion of low-confidence pending confirmation items does not exceed the current upper limit, but there are 3 or more consecutive samples that fail to execute and the failure reason points to location record error, then the corresponding upper limit will be decreased by 5 percentage points. Execution failure is limited to the target execution plan not completing closed-loop execution or the actual batch difference exceeding 2 batches or the local relocation instruction number difference exceeding 2 or the station operation exceeding the end time of the scheduled receiving period.

[0176] If the total number of local reshipment instructions is less than 50% of the current warning threshold for 10 or more consecutive valid samples, the warning threshold will be lowered by 1; if the number of valid samples is higher than the current warning threshold for 5 or more consecutive times and the operation can still be completed within the scheduled receiving period at the site, the warning threshold will be raised by 1.

[0177] If the actual value of the total number of unloading batches is consistently lower than the planned value by 2 or more batches in 5 or more consecutive valid samples, the batch warning threshold under the corresponding condition will be lowered by 1 batch; if the actual value is consistently higher than the planned value by 2 or more batches in 5 or more consecutive valid samples and is accompanied by new unloading, the threshold will be raised by 1 batch; "constantly fixed" means that the same deviation is consistently met in 5 or more consecutive valid samples.

[0178] To prevent rule drift from occurring too quickly, the same rule item can be modified a maximum of once per calendar day, with a fixed freeze period of one calendar day. When different rule items trigger modifications on the same calendar day, they are allowed to take effect in parallel. The rule of "maximum modification once per day," the freeze period limit, and the priority of safety tightening are only applied within the same rule item, and each modification is allowed to change only one preset step size, one location level, or one order of magnitude. When the same rule item triggers both upward and downward adjustments on the same day, the safety tightening direction is prioritized. Specifically, lowering the upper limit of the proportion of low-confidence pending confirmation items is considered safety tightening; lowering the warning threshold for the number of partial cargo reversal instructions is considered safety tightening; lowering the warning threshold for the total number of unloading batches is considered safety tightening; and tightening the new acceptable boundary of the location after "continuing to follow the vehicle" by one location level is considered safety tightening. If the safety directions are consistent, the one with more triggering samples is retained.

[0179] Both before and after the correction, the rule version number and parameter version number are registered, and a rule correction evidence chain is generated. The evidence chain includes at least the sample time range, vehicle type, route template, delivery area, triggering condition, value before correction, value after correction, number of triggering samples, execution deviation summary, review status, and rollback conditions. The review status is limited to three values: pending review, passed, and rejected. Only the passed status is effective.

[0180] The rollback condition is fixed as follows: if the corresponding execution deviation increases instead of decreases in 20 consecutive valid samples after the correction takes effect, and the increase reaches more than 20%, then the rollback will be to the previous rule version.

[0181] After the correction is completed, a rule correction file is generated. The file header information includes at least the vehicle application scope, the route template application scope, the rule version number, the parameter version number, the effective time, and the expiration time. The body of the file includes at least the corrected filtering threshold item, the corrected section reachability rule item, the correction direction, the correction step size, the evidence chain index, and the rollback condition. The file is written to the dispatch center rule base and an incremental update message is sent to the vehicle terminal for the next observation window to call. The message communication adopts a message queue plus version confirmation method. The vehicle terminal only loads the new version after completing the task closure of the old version to avoid cross-observation window mixing.

[0182] The idempotency, ordering, and deduplication strategies are fixed as follows: only one correction record is retained for the same rule item at the same effective time; duplicate writes are rejected when the digests of duplicate correction requests are consistent, and records with a newer rule version number and an approved status are retained when the digests are inconsistent; and older version correction records must not overwrite newer versions that have already taken effect.

[0183] If situations arise such as execution record missing exceeding the sample limit, sample source conflict, incomplete evidence chain, rule exceeding vehicle safety boundary after correction, or repeated triggering of the same rule item during the freeze period, this step will not perform correction, but will mark the corresponding rule item as an item to be reviewed and write it into the evidence chain log. The sample limit is fixed at 20% of the number of missing execution segments in consecutive valid samples.

[0184] The minimum feasible set of interfaces includes the target execution plan reading interface, the execution record uploading interface, the rule correction file writing interface, and the rule correction file distribution interface. Among them, the execution record uploading interface should at least receive the vehicle number, train number, observation window number, target execution plan number, station number, batch number, event time, and event type. If the target execution plan number is missing, it returns code 701; if the event type is not within the limited set, it returns code 702; if the observation window number is invalid, it returns code 703. The rule correction file distribution interface should at least receive the rule version number, parameter version number, and effective time. If there is a version conflict, it returns code 704; if the vehicle model applicable scope is missing, it returns code 705.

[0185] The safety and compliance boundaries are limited to processing only the cargo, station, location, time period, operation events and rule version fields required for the closed-loop execution of the current trip, without collecting drivers' private communication content, and without including original signed images containing personal identification information in the rule revision documents.

[0186] The on-site inspection criteria can be set as rule correction generation efficiency, post-correction execution deviation convergence rate, section reachability misjudgment reduction rate, and next observation window call success rate. Among them, rule correction generation efficiency is calculated based on the proportion of rule correction files that are successfully issued and loaded and called by the next observation window. Post-correction execution deviation convergence rate is calculated based on the average decrease rate of the absolute value of the difference between the planned and actual batches in 20 consecutive valid samples after correction. Section reachability misjudgment reduction rate is calculated based on the difference in the proportion of samples that erroneously triggered local reversal instructions before and after correction. The sampling sample size can be set to 30 consecutive vehicle trips, with no less than 20 valid samples per vehicle type and per route template.

[0187] Preferably, in the community retail urban distribution scenario, under the conditions of 4.2m vans, normal temperature turnover box goods, and 8 to 15 stations per vehicle, the most recent 30 trips generated 24 valid correction rule samples. Among them, 7 samples continuously showed that although the net width of the passage from the side of the rear area to the door met the original preset lower limit of passage, it was still necessary to perform an additional unloading before the first batch of goods could leave the van. Based on this, the system raised the preset lower limit of passage for the corresponding goods width range by 30mm. After the correction took effect, in the 20 consecutive valid samples, the section reachability misjudgment rate decreased by 36%, the difference between the local unloading instruction and the actual execution decreased by 28%, and the average rule correction file generation time was 460ms.

[0188] Alternatively, in locations without a centralized rule base on the central side, the vehicle-mounted terminal can locally store the most recent consecutive valid samples and generate a list of rules to be corrected. After communication is restored, these rules can be uniformly transmitted back to the dispatch center for review and approval. As long as the same consecutive valid sample criteria, the same triggering conditions, the same step-length correction rules, and the same version of the freezing strategy are used to correct the screening threshold and segment reachability rules corresponding to the load path mismatch, this is an equivalent implementation that can be implemented in this step.

[0189] In the operational scenario shown in this embodiment: taking a community retail distribution center's morning replenishment delivery to convenience stores in the city as an example, the delivery vehicle is a 4.2m box truck. After the vehicle is loaded from the distribution center, it delivers boxed beverages, bagged food, daily dairy products, and paper products to the 1st to 8th stations in sequence according to the predetermined route. When loading, the warehouse management system, transportation management system, and vehicle terminal jointly write the item number, station affiliation, loading area, size, weight, and loading time. After the vehicle completes the delivery to the 3rd station, the vehicle terminal re-collects the current status of the unloaded goods in the observation window before the door closes again and the vehicle starts moving, forming the remaining load execution map. At this time, there are still 96 unloaded goods in the vehicle, and the original order of the remaining stations to be reached is the 4th, 5th, 6th, 7th, and 8th stations.

[0190] Subsequently, the dispatch center received a new station instruction, requiring the vehicle to insert a temporary replenishment station after the original 4th station, and at the same time, the scheduled pickup time of the 6th station was moved back by 20 minutes. Based on this, the vehicle terminal combined the current location, the remaining stations to be reached, the pickup time, and the delivery change event to generate a route change fingerprint.

[0191] After the path calculation service process calls the remaining load execution graph and path change fingerprint, it calculates the unloading conflict amount, cargo reversal amount, and time period deviation amount for multiple path adjustment actions such as inserting a new station before the 4th station, after the 4th station, after the 5th station, and after the 6th station, forming the corresponding load path mismatch degree. Among them, although inserting a new station before the 4th station has a small time period deviation, it will cause the cargo of the subsequent station in the middle and side position of the tail area to block the first batch of cargo from leaving the 4th station, resulting in a high unloading conflict amount and cargo reversal amount. When inserting a new station after the 6th station, although a small amount of time period waiting is added, the overall unloading conflict amount decreases and the cargo reversal amount is less.

[0192] The system then filters path adjustment actions that meet the segment reachability constraints and do not introduce new occlusion relationships based on the load path mismatch, generates a restricted adjustment set, and further performs joint calculations within the restricted adjustment set to determine the target execution plan.

[0193] The implementation plan determines the remaining stations in the adjusted order as Station 4, Station 6, Station 5, Temporary Replenishment Station, Station 7, and Station 8. At the same time, two unloading batches are generated for Station 4 and two unloading batches are generated for Station 5. A partial repositioning instruction is generated for a piece of paper goods that was originally located on the right side of the rear area and was blocking the first batch of goods from leaving the container at Station 5. The instruction requires that the goods be temporarily moved to the empty area on the left side of the middle area when the operation at Station 5 begins, and continue to follow the vehicle after the last piece of goods leaves the container at Station 5, without being moved back.

[0194] After the target execution plan is issued, the driver's terminal and the station operation terminal execute unloading in batches. At the beginning, end, partial unloading trigger, completion, and station receipt completion of each batch, the actual execution record is sent back to the dispatch center. The rule correction service process compares the actual execution record with the target execution plan and finds that in multiple consecutive template samples of the same vehicle type and route, although the passage from the side of the rear area to the door meets the original preset lower limit, additional unloading is still required before the first batch of goods can be released. Therefore, the preset lower limit of the corresponding goods width range is increased by a predetermined step size, and the corrected screening threshold and section reachability rules are written into the rule correction file, which takes effect in the next observation window and subsequent similar trips.

[0195] Through the above operation process, the system does not simply make a static compromise between "high loading rate" and "short path". Instead, under the actual working conditions where the vehicle has completed part of the delivery and there are interruptions and time changes along the way, it forms a dynamic calculation closed loop that can be directly executed around the remaining load status, the remaining station sequence, the unloading batch arrangement, and the local reloading action. This allows the vehicle to continue to complete the subsequent delivery without the whole vehicle being reloaded or rearranged.

[0196] All calculations involved in the embodiments are dimensionless numerical calculations, and the preset parameters and thresholds in the calculations are set by those skilled in the art according to the actual situation.

[0197] It should be noted that this invention can be deployed on the device itself to realize embedded applications, or it can run on a PC or other terminal with a user interface, thereby meeting various hardware environments and usage requirements.

[0198] The above embodiments can be implemented, in whole or in part, by software, hardware, firmware, or any other combination thereof. When implemented using software, the above embodiments can be implemented, in whole or in part, as a computer program product. The computer program product includes one or more computer instructions or computer programs. When the computer instructions or computer programs are loaded or executed on a computer, all or part of the processes or functions described in the embodiments of this application are generated. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions can be stored in a computer-readable storage medium or transmitted from one computer-readable storage medium to another. For example, the computer instructions can be transmitted from one website, computer, server, or data center to another website, computer, server, or data center via wireless or wired transmission; wired transmission methods include optical fiber, twisted pair, coaxial cable, etc.; wireless transmission includes infrared, microwave, etc. The computer-readable storage medium can be any available medium that a computer can access or a data storage device such as a server or data center containing one or more sets of available media. The available medium can be a magnetic medium (e.g., floppy disk, hard disk, magnetic tape), an optical medium (e.g., DVD), or a semiconductor medium. A semiconductor medium can be a solid-state drive.

[0199] Those skilled in the art will understand that, for the sake of convenience and brevity, the specific working processes of the systems, devices, and modules described above can be referred to the corresponding processes in the foregoing method embodiments, and will not be repeated here.

[0200] In the several embodiments provided in this application, it should be understood that the disclosed systems, apparatuses, and methods can be implemented in other ways. For example, the apparatus embodiments described above are merely illustrative; for instance, the division of modules is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple modules or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be through some interfaces; the indirect coupling or communication connection between apparatuses or modules may be electrical, mechanical, or other forms.

[0201] The modules described as separate components may or may not be physically separate. The components shown as modules may or may not be physical modules; they may be located in one place or distributed across multiple network modules. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs.

[0202] In addition, the functional modules in the various embodiments of this application can be integrated into one processing module, or each module can exist physically separately, or two or more modules can be integrated into one module.

[0203] If the aforementioned functions are implemented as software functional modules and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, or a portion of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of this application. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0204] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

[0205] In conclusion, the above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A method for calculating vehicle loading rate and route, characterized in that, include: S1. Collect the station affiliation, loading location, size, weight, and obstruction relationship of the vehicle's currently unloaded cargo, and construct the remaining load execution map; S2. Collect the vehicle's current location, remaining destinations, pickup times at each destination, and delivery change events, and generate a route change fingerprint. S3. Based on the remaining load execution diagram and path change fingerprint, calculate the unloading conflict amount, cargo transfer amount and time period deviation corresponding to each path adjustment action to form the load path mismatch degree. S4. Select path adjustment actions based on load path mismatch. The path adjustment actions must satisfy the segment reachability constraint and not add any occlusion relationships. The segment reachability constraint is: under the adjusted station sequence, there must be at least one continuous passage segment between the target cargo belonging to the previous station and the door passage from the current main location along the unique unloading direction; and generate a restricted adjustment set. S5. Perform joint calculations on the restricted adjustment set to determine the target execution plan, which includes the adjusted remaining station sequence, unloading batch sequence, and partial cargo relocation instructions. S6. Based on the actual execution records of the target execution plan, revise the screening threshold and segment reachability constraint determination rules corresponding to the load path mismatch.

2. The method for calculating vehicle loading rate and route according to claim 1, characterized in that, S1 includes: The cargo part number is used as the primary key to collect information on station affiliation, loading location, size, weight, and obstruction from the front. The order of participation is determined by the primary location, while the occupancy of the passageway is determined by the secondary location. Determine the obstruction relationship based on the unique unloading direction, and form a directed edge for each pair of obstructing and obstructed components; The remaining load execution graph file is generated using vehicle number, wave number, mapping time, rule version number, and parameter version number as header information.

3. The method for calculating vehicle loading rate and route according to claim 1, characterized in that, S2 include: The current location of the vehicle, the remaining destinations to be reached, the pickup time of each destination, and the delivery change events are aligned in a unified time. The first station in the current sequence of remaining stations to be reached is taken as the predetermined target station, and the deviation of the vehicle's driving direction is determined by the direction of the line connecting the vehicle's current position to the predetermined target station. Using vehicle number, train number, observation window start and end times, rule version number, and parameter version number as header information, a path change fingerprint file is formed in the order of changes in station set, station order, delivery time period, and traffic conditions.

4. The method for calculating vehicle loading rate and route according to claim 1, characterized in that, S3 includes: Based on the remaining load execution diagram and path change fingerprint, calculate the unloading conflict amount, cargo reversal amount and time period deviation for each path adjustment action; The load path mismatch is determined by prioritizing the unloading conflict amount, then the reloading amount, and finally the time period deviation amount. The load path mismatch is represented by a record that includes path adjustment actions, unloading conflict amount, reloading amount, time period deviation amount, and sorting position.

5. The method for calculating vehicle loading rate and route according to claim 4, characterized in that, Calculate unloading conflict volume and reverse volume, including: Check the path of the target cargo piece leaving the container along the only unloading direction at each station in the adjusted station order; Target cargo items that have frontal obstruction, directional obstruction, or passage blockage on the exit passage are included in the unloading conflict quantity. Cargo items located on the obstruction chain, whose station sequence is later than the target departure station, and which need to be moved away from the original path before the target cargo item can be released, will be included in the reload volume.

6. The method for calculating vehicle loading rate and route according to claim 1, characterized in that, S4 include: Arrange path adjustment actions in ascending order by load path mismatch degree; First, determine the reachability constraints of the segment in the sorting order, then determine whether no new occlusion relationship will be added. If the previous determination is not true, stop the subsequent checks of the corresponding path adjustment action. Only path adjustment actions that simultaneously satisfy the segment reachability constraint and do not introduce new occlusion relationships are written into the restricted adjustment set.

7. The method for calculating vehicle loading rate and route according to claim 1, characterized in that, S5 include: The adjustment actions of each path in the restricted adjustment set are calculated together to obtain the adjusted remaining station sequence, unloading batch sequence, and local reloading instructions; The target execution plan is determined in the following order: total number of partial unloading instructions, total number of unloading batches, sorting order of original load path mismatch, and path adjustment action number.

8. The method for calculating vehicle loading rate and route according to claim 7, characterized in that, Generate unloading batch sequence and partial unloading instructions, including: Generate unloading batch order station by station according to the adjusted order of the remaining stations; For obstructing cargo items that do not belong to the current batch and cannot be released unless moved, a partial reversal instruction is generated. Update the location status of the goods based on the results of the partial repositioning instruction.

9. The method for calculating vehicle loading rate and route according to claim 1, characterized in that, S6 include: Establish an execution checklist chain based on the target execution plan number; Based on the execution deviation of the actual execution records in the continuous valid samples relative to the target execution plan, the upper limit of the proportion of low-confidence pending confirmation items, the warning threshold of the number of partial unloading instructions, the warning threshold of the total number of unloading batches, the preset lower limit of passage, the restricted boundary of temporary storage area, the reserved boundary of the first batch of unloading passage, and the acceptable boundary of the new location after continuing to follow the vehicle are corrected. The revised results are registered with the rule version number and parameter version number, and a rule revision file is generated.

Images