Unmanned delivery cross-regional collaborative scheduling re-optimization method for dynamic delivery demand
By constructing a multi-regional dynamic delivery network and optimizing operators, the problems of low computational efficiency and insufficient optimization accuracy in dynamic path planning of unmanned logistics are solved, and efficient and low-cost cross-regional collaborative scheduling is achieved.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2026-01-19
- Publication Date
- 2026-03-27
AI Technical Summary
Existing technologies suffer from low computational efficiency and insufficient optimization accuracy when dealing with dynamic path planning for unmanned logistics that balances cross-regional collaboration and cost re-optimization, leading to increased transportation costs and low efficiency in collaborative scheduling.
A multi-regional dynamic delivery network is constructed, a uniform time period is divided, a vertex set and an arc set are defined, and constraints on customer service, vehicle capacity and time window are set. Cost reduction and exchange reduction removal operators are adopted, and the solution path is optimized by combining the O(1) time window compatibility insertion method and the simulated annealing acceptance criterion.
It improved computational efficiency by 74.24%, matched 7 optimal solutions in 20 benchmark instances, and had an average error of only 0.5%, achieving low-cost route planning under dynamic delivery requirements.
Smart Images
Figure CN121745795A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of unmanned logistics technology, and specifically to a method for re-optimizing cross-regional collaborative scheduling of unmanned delivery to meet dynamic delivery needs. Background Technology
[0002] In unmanned logistics operations, since the actual delivery needs of customers are often dynamic, unmanned logistics operators must constantly adjust the delivery plans of unmanned vehicles to cope with the new delivery demands in real time, while minimizing the total operating cost under capacity constraints.
[0003] Despite continuous advancements in vehicle resource planning (VRP) technology, the complexities of the real world, such as dynamic delivery demands, unpredictable traffic conditions, and the need for efficient computing algorithms, have led to problems such as low computational efficiency, difficulty in handling complex large-scale regional deliveries, and failure to consider the time cost of re-optimization processes. These issues affect collaborative scheduling or transportation efficiency and increase the cost of delivery solutions. Summary of the Invention
[0004] To address the problems existing in the prior art, this invention provides a cross-regional collaborative scheduling and re-optimization method for unmanned delivery oriented towards dynamic delivery needs. This method solves the problems of low computational efficiency, insufficient optimization accuracy, and lack of cost balance in the dynamic path planning of unmanned logistics when dealing with the cost balance between cross-regional collaboration and re-optimization.
[0005] The method for cross-regional collaborative scheduling and re-optimization of unmanned delivery for dynamic delivery needs includes: when constructing a multi-regional dynamic delivery network, the operating time range [0,H] is divided into P uniform time periods, and the vertex set and arc set of each time period are defined. For a given dynamic delivery task, with the goal of minimizing the total transportation cost of unmanned delivery vehicles, customer service constraints, vehicle capacity constraints, time window constraints, and delivery area affiliation constraints are set. Then, the optimal multi-regional dynamic delivery path is obtained through the initial solution construction stage and the optimization solution stage.
[0006] The objective function is:
[0007] In the formula, For the p-th period (Note: p is the period index, lowercase), The arc set of ) For unmanned delivery vehicles to pass through the arc Transportation costs, 0-1 decision variables ( =1 indicates that the vehicle passed through the arc. , =0 indicates that it is not passed.
[0008] Furthermore, the vertex set for each cycle is defined as follows:
[0009]
[0010] In the formula, For fixed storage parking lots, For the set of customers to be served in the p-th period (note: p is the period index, lowercase), These are temporary parking lots used for waiting in the previous period. They are converted from customer locations where services were not completed and vehicles were parked, and they meet the following criteria:
[0011]
[0012] Furthermore, the vehicle capacity constraint is expressed as:
[0013]
[0014] In the formula, k represents the set of unmanned delivery vehicles. For customer i's order demand, 0-1 decision variables ( =1 indicates that vehicle k serves customer i). This represents the maximum rated load capacity of the unmanned delivery vehicle k.
[0015] Furthermore, the time window constraint is expressed as:
[0016]
[0017] In the formula, The time window for customer C (i.e., the time range within which the customer can accept delivery services is dynamically adjusted over the period). The service start time is when the unmanned delivery vehicle arrives at customer c, and the following conditions are met:
[0018]
[0019] in, For the service time of node i, The transit time from node i to node j (including the service time of node i).
[0020] Furthermore, the optimization solution stage includes: performing destruction operations using 6 removal operators, including cost reduction removal operators and exchange reduction removal operators; performing repair operations using a time window compatible O(1) insertion method; dynamically adjusting the selection probability based on operator performance for adaptive updates; and simulating annealing acceptance using an acceptance criterion.
[0021] Furthermore, the judgment condition for the cost reduction removal operator is:
[0022]
[0023] In the formula, , For customer v, the nodes before and after in the original path. , The arc node of the path to be inserted. Let $x$ be the distance cost from node $x$ to node $y$.
[0024] The condition for determining the swap reduction removal operator is:
[0025]
[0026] In the formula, , For paired customers, , for Nodes before and after the original path , for The nodes before and after the original path.
[0027] Furthermore, the time window compatibility O(1) insertion method includes:
[0028] Feasibility verification is performed by verifying the following conditions:
[0029]
[0030] In the formula, For the first in path L The earliest service start time for each node. For the first in path L The latest service start time for each node. , Let c be the adjacent node in path L, and c be the customer to be inserted.
[0031] The calculation methods for e(L) and l(L) are as follows:
[0032]
[0033]
[0034]
[0035]
[0036] In the formula, Let p be the start time of the p-th period. To optimize the duration, H represents the number of customers in the path, and H represents the total operating time. , For nodes The time window parameters.
[0037] Furthermore, the acceptance criteria include: setting an initial temperature. And propose a decreasing deceleration rate To lower the temperature:
[0038] In the formula, Let S represent the current temperature at the t-th iteration. Even if a suboptimal solution S' is obtained from the current solution S, the algorithm will still use it as the current solution and continue execution.
[0039] Furthermore, it also includes establishing a model of the impact of warehouse optimization cycle and unmanned vehicle fixed costs on total transportation costs, providing a basis for decision-making in practical applications.
[0040] Specifically, benefit analysis and effect evaluation were conducted, and the results showed that when the total number of optimization cycles P satisfies At that time, the target cost showed a negative growth trend:
[0041] The relationship between vehicle fixed cost F and the number of vehicles used K satisfies:
[0042]
[0043] In the formula, C represents the total transportation cost.
[0044] The beneficial effects of this invention include:
[0045] The initial path is generated by a greedy insertion algorithm, and adaptive large neighborhood search parameters are set. The algorithm is triggered in each re-optimization cycle. A new cost-oriented removal operator is used to remove customers, including cost reduction removal and exchange reduction removal. Then, an O(1) time window compatibility verification method is used. Based on the earliest service start time e(L) and latest service start time l(L) of the path node, only the inserted customer repair path is verified. Combined with the simulated annealing criterion, non-improved solutions are accepted. The computational efficiency is improved by 74.24% compared with the traditional O(n) method. When dealing with 288 customers, the average computation time per cycle is only 9.15s. Through the new cost-oriented removal operator, 7 optimal known solutions are matched in 20 benchmark instances with an average error of only 0.5%. A quantitative relationship between re-optimization cycle and cost is established to provide enterprises with operable decision parameters. It supports dynamic logistics scenarios of different scales and types, and can be adapted to specific needs through parameter adjustment. Attached Figure Description
[0046] Figure 1 This is a flowchart illustrating a dynamic path planning method for unmanned logistics that balances cross-regional collaboration and cost re-optimization, as described in an embodiment of this application.
[0047] Figure 2 This is an example diagram of dynamic re-optimization of unmanned logistics routes involved in the embodiments of this application. Detailed Implementation
[0048] To make the objectives, technical solutions, and advantages of the embodiments of this application clearer, the technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of this application, and not all of the embodiments. Therefore, the following detailed description of the embodiments of this application provided in the accompanying drawings is not intended to limit the scope of the claimed application, but merely represents selected embodiments of this application. All other embodiments obtained by those skilled in the art based on the embodiments of this application without creative effort are within the scope of protection of this application.
[0049] Example 1
[0050] Specifically, see the attached document below. Figure 1 The specific implementation process of the present invention will be described in detail;
[0051] A dynamic path planning method for unmanned logistics that balances cross-regional collaboration and cost reoptimization includes: when constructing a multi-regional dynamic delivery network, the operating time range [0,H] is divided into P uniform time periods, and the vertex set and arc set of each time period are defined. For a given dynamic delivery task, with the goal of minimizing the total transportation cost of unmanned delivery vehicles, customer service constraints, vehicle capacity constraints, time window constraints, and delivery area affiliation constraints are set. Then, the optimal multi-regional dynamic delivery path is obtained through the initial solution construction stage and the optimization solution stage.
[0052] The vertex set for each period is defined as follows:
[0053]
[0054] In the formula, For fixed storage parking lots, For the set of customers to be served in the p-th period (note: p is the period index, lowercase), These are temporary parking lots used for waiting in the previous period. They are converted from customer locations where services were not completed and vehicles were parked, and they meet the following criteria:
[0055]
[0056] The objective function is:
[0057] In the formula, For the p-th period (Note: p is the period index, lowercase), The arc set of ) For unmanned delivery vehicles to pass through the arc Transportation costs, 0-1 decision variables ( =1 indicates that the vehicle passed through the arc. , =0 indicates that it is not passed.
[0058] The customer service constraints include: each order point to be delivered is served only once by one unmanned vehicle;
[0059] The vehicle capacity constraint is expressed as follows:
[0060]
[0061] In the formula, k represents the set of unmanned delivery vehicles. For customer i's order demand, 0-1 decision variables ( =1 indicates that vehicle k serves customer i). This represents the maximum rated load capacity of the unmanned delivery vehicle k.
[0062] The time window constraint is expressed as follows:
[0063]
[0064] In the formula, The time window for customer C (i.e., the time range within which the customer can accept delivery services is dynamically adjusted over the period). The service start time is when the unmanned delivery vehicle arrives at customer c, and the following conditions are met:
[0065]
[0066] in, For the service time of node i, The transit time from node i to node j (including the service time of node i).
[0067] The delivery area ownership constraints include: the unmanned vehicle must depart from and return from its assigned parking lot.
[0068] The optimization solution stage includes: using 6 removal operators to perform destruction operations, including cost reduction removal operators and exchange reduction removal operators; using a time window compatible O(1) insertion method to perform repair operations; dynamically adjusting the selection probability according to the operator performance and performing adaptive updates; and using an acceptance criterion to simulate annealing acceptance.
[0069] The criterion for determining the cost reduction removal operator is:
[0070]
[0071] In the formula, , For customer v, the nodes before and after in the original path. , The arc node of the path to be inserted. Let $x$ be the distance cost from node $x$ to node $y$.
[0072] The condition for determining the swap reduction removal operator is:
[0073]
[0074] In the formula, , For paired customers, , for Nodes before and after the original path , for The nodes before and after the original path.
[0075] The time window compatibility O(1) insertion method includes:
[0076] Feasibility verification is performed by verifying the following conditions:
[0077]
[0078] In the formula, For the first in path L The earliest service start time for each node. For the first in path L The latest service start time for each node. , Let c be the adjacent node in path L, and c be the customer to be inserted.
[0079] The calculation methods for e(L) and l(L) are as follows:
[0080]
[0081]
[0082]
[0083]
[0084] In the formula, Let p be the start time of the p-th period. To optimize the duration, H represents the number of customers in the path, and H represents the total operating time. , For nodes The time window parameters.
[0085] The acceptance criteria include: setting an initial temperature , and proposing a decreasing rate to reduce the temperature:
[0086] In the formula, represents the current temperature at the t-th iteration. Even if the sub-optimal solution S’ obtained from the current solution S, the algorithm will still continue to execute with it as the current solution.
[0087] Figure 2 Briefly shows the operating state of MD-DVRPTW during the time period t ∈ [Tp−1, Tp] (1 ≤ p < P). Figure 2 (a) The positions of each site, including the warehouse and customer points, are clearly marked by graphic labels. Figure 2 (b) Shows the time series data corresponding to each site. The red dashed line indicates that the planned route before the time Tp−1 has approached or reached the customer point, so this customer point is regarded as a temporary warehouse. The red mark represents the temporary warehouse transformed from the original customer. The blue solid line represents the newly planned route at the time Tp−1, and the blue mark shows the newly added customer points during the time period [Tp−1, Tp]. The routes of existing customers (black marks) and unserved customers (blue marks) will be adjusted at the decision point Tp. According to the setting, the vehicle leaving the warehouse must finally return to the original warehouse; if the vehicle arrives at vertex i earlier than vertex ai, it needs to wait.
[0088] Finally, it also involves establishing an impact model of the warehousing re-optimization cycle and the fixed cost of the unmanned vehicle on the total transportation cost, providing a decision-making basis for practical applications.
[0089] Specifically, a benefit analysis and effect evaluation are carried out. The results show that: when the total number of re-optimization cycles P satisfies , the target cost shows a negative growth trend:
[0090] The relationship between the vehicle fixed cost F and the number of vehicles used K satisfies:
[0091]
[0092] In the formula, C is the total transportation cost.
[0093] During the implementation process, first, an experimental environment (3.6GHz Intel Core i7-7700 processor, 16GB RAM, coded in C++ and compiled with Visual Studio 2019, running single-threaded) is built, and basic data such as the warehouse coordinates, customer demands, dynamics, and unmanned vehicle capacity of the unmanned logistics scenario are imported.
[0094] Twenty test instances were generated based on the benchmark instance (with customer sizes ranging from 48 to 288; the time window for each customer was narrower from pr01 to pr10, and relatively wider from pr11 to pr20). The route duration and number of vehicles for each MD-VRPTW instance were limited. The base instance was generated using a randomized method, creating customer clusters around a specific number of seed points.
[0095] The initial path is generated by a greedy insertion algorithm, and adaptive large neighborhood search parameters are set. The algorithm is triggered in each re-optimization cycle. A new cost-oriented removal operator is used to remove customers, including cost reduction removal and exchange reduction removal. Then, the O(1) time window compatibility verification method is used. Based on the earliest service start time e(L) and latest service start time l(L) of the path node, only the inserted customer repair path is verified. The non-improved solution is accepted in combination with the simulated annealing criterion.
[0096] Comparative experiments show that the computational efficiency of the method provided in this invention is improved by 74.24% compared to the traditional O(n) insertion method. When processing unmanned logistics scenarios for 288 customers, the average computation time for a single optimization cycle is 9.15 seconds, and it matches 7 optimal known solutions in 20 benchmark instances with an average error of 0.5%. The traditional O(n) complexity insertion method refers to greedy insertion and regret-3 insertion.
[0097] The core design principle of this system is: O(1) verification reduces the number of verifications through bidirectional time constraints (e(L) is recursively pushed from the starting point to the end point and l(L) is recursively pushed from the end point to the beginning), and the cost-oriented removal operator selects customers with the goal of minimizing the global cost, avoiding local optimization, and ensuring that the unmanned vehicle path has both real-time performance and economy under the time window constraint.
[0098] The data obtained by solving the problem using the method described in this embodiment is shown in Table 1:
[0099] Table 1. Comparison of the solution results obtained by this method with the known optimal solution.
[0100]
[0101] In the table, Inst represents the instance name, used to identify different path test instances; m represents the number of distribution centers, representing the total number of depots participating in delivery; n represents the number of customers, i.e., the total number of customer points that need to be served; v represents the number of vehicles, referring to the total number of vehicles used to complete the delivery task; BKS represents the Best-Known Solution, which is the currently known optimal solution value in this field of research; Best represents the optimal solution obtained by this method; ALNS - Avg. represents the average solution obtained by this method over multiple runs; and Time(s) represents the average time (in seconds) required by this method to solve this instance.
[0102] As can be seen, the method described in this embodiment matched the known optimal solution in 7 out of 20 instances (shown in bold), and the problem was solved in an average of about 121.97 seconds. This shows that the method can effectively obtain near-optimal solutions in a competitive time.
[0103] Table 2 Performance comparison between our method and the state-of-the-art algorithm SÇA
[0104]
[0105] The experimental environment for SCA was an Intel Core i7-8700 @ 3.20 GHz (Passmark CPU Mark: 2655); the experimental environment for the method described in this embodiment was an Intel Core i7-7700 @ 3.60 GHz (Passmark CPU Mark: 2463).
[0106] As shown in Tables 1 and 2, the difference between the best results obtained by this method and those obtained by BKS is 0.5%.
[0107] Table 3 below compares the performance of the three different removal operators of this method to demonstrate its superiority.
[0108] General: This method (referred to as the general version) contains only 4 traditional removal operators (random removal, random path removal, Shaw removal, and worst-case removal).
[0109] CRR: Based on General, a new cost reduction removal operator is added (selecting objects to be removed by determining whether the customer's insertion cost has decreased).
[0110] ERR: In addition to General, a new swap reduction removal operator is added (selecting removal targets by comparing the costs of swapping pairs of customers).
[0111] The computation time of different algorithms varies significantly: the CRR algorithm takes an average of 165.66 seconds, exhibiting high overhead; cost reduction removal mainly affects customer choice but has limited impact on improving the target value; the ERR algorithm has advantages in both target value and computation time. This indicates that exchange reduction removal performs better in both problem-solving efficiency and accuracy. The ERR algorithm solves the problem in the shortest time (49.89 seconds) while maintaining the advantage of finding a better solution.
[0112] Table 3. Performance comparison of different removal operators in this method.
[0113]
[0114] Table 4 below compares variations of this method with different insertion methods and removal operators, focusing only on computation time to verify the "time window compatible insertion method" with O(1) complexity:
[0115] AA: General (traditional removal operator) + conventional insertion method (O(n) complexity, requires iteratively checking the feasibility of time windows for all positions of the path, as a baseline group).
[0116] AB: General + Time Window Compatibility Insertion Method (O(1) complexity, quickly determine insertion feasibility by pre-calculating "earliest / latest service time").
[0117] AC: This method proposes a new removal operator (CRR+ERR) + conventional insertion method, without using time window compatibility optimization.
[0118] AD: This method proposes a new removal operator + time window compatible insertion method.
[0119] For each scenario, the deviation calculation formula is as follows: Δ% = (Calculation time obtained using AA - Calculation time obtained using AB, AC, or ALNS) / Calculation time obtained using AA × 100%. Δ% = (AC calculation time - AD calculation time) / AC calculation time × 100%
[0120] pr01: (19.68 - 4.91) ÷ 19.68 × 100% ≈ 14.77 ÷ 19.68 ≈ 75.05% (and so on for others)
[0121] PR01 approximately 75.05%, PR02 approximately 76.34%, PR03 approximately 70.80%, PR04 approximately 68.00%, PR05 approximately 61.45%, PR06 approximately 61.57%, PR07 approximately 75.20%, PR08 approximately 69.25%, PR09 approximately 65.16%, PR10 approximately 62.37%, PR11 approximately 91.04%, PR12 approximately 89.12%, PR13 approximately 87.47%, PR14 approximately 83.06%, PR15 approximately 77.46%, PR16 approximately 76.98%, PR17 approximately 91.00%, PR18 approximately 85.37%, PR19 approximately 82.19%, PR20 approximately 73.39%.
[0122] Table 4. Comparison of computation time and deviation for various insertion methods or combinations of removal operators under this method.
[0123]
[0124] Divide the 8-hour (H=8h) timeframe of Horizon into 20 cycles (each cycle is 24 minutes long, 8×60 / 24=20), and perform a re-optimization once per cycle.
[0125] Table 5 shows the solution results for the "Dynamic" and "Static" groups, corresponding to the solutions for dynamic and static problems, respectively. Customer demands in the dynamic group emerge in real time and require periodic re-optimization; all customer demands in the static group are known.
[0126] The columns corresponding to Δ% for each group show the deviation between the target value and the computation time. In the previous settings, the re-optimization time was set to two minutes by default. However, the data in Table 9 shows that this method can efficiently solve the problem within the specified time. It is worth noting that compared to the static version of the problem with all known customer requirements, the computational cost of this method is significantly higher (the static version saves an average of 9.43%). The computation time for the dynamic solution is approximately 183.08 seconds, while the static version only requires 38.51 seconds.
[0127] According to the data in Table 5, the average time required for each re-optimization of MD-DVRPTW is 9.15 seconds, which fully meets the requirements for fast solution. The average total computation time for dynamic problems is 183.08s (average total time for 20 instances). Total average computation time ÷ number of cycles = 183.08s ÷ 20 ≈ 9.15s, which is the average computation time for a single optimization cycle.
[0128] By comparing dynamic and static problems, it was found that in instances p11-p20, the difference gradually narrowed due to the wider time window. Summing the single-instance efficiency improvement rate for all 20 instances: adding the savings rates of all instances yielded a total savings rate of approximately 1484.8%.
[0129] 75.05%+76.34%+70.80%+68.00%+61.45%+61.57%+75.20%+69.25%+65.16%+62.37%+91.04%+89.12%+87.47%+83.06%+77.46%+76.98%+91.00%+85.37%+82.19%+73.39% ≈1484.8%.
[0130] Calculate the average: Total savings rate ÷ Number of instances = 1484.8% ÷ 20 ≈ 74.24%.
[0131] Table 5 Performance comparison of this method in dynamic and static scenes.
[0132]
[0133] The implementation steps described above merely illustrate specific embodiments of this application, and while the descriptions are detailed, they should not be construed as limiting the scope of protection of this application. It should be noted that those skilled in the art can make various modifications and improvements without departing from the technical concept of this application, and these modifications and improvements all fall within the scope of protection of this application.
Claims
1. A method for cross-regional collaborative scheduling and re-optimization of unmanned delivery for dynamic delivery needs, characterized by constructing a cross-regional collaborative unmanned delivery network, dividing the operating time range [0,H] into P uniform time periods to transform the dynamic scheduling problem into multiple static scheduling problems, defining the vertex set and arc set for each time period, and setting constraints such as customer service constraints, vehicle capacity constraints, time window constraints, and delivery area affiliation for the delivery tasks within the period with the goal of minimizing the total transportation cost of unmanned delivery vehicles, and obtaining the optimal multi-regional dynamic delivery path through the initial solution construction stage and the optimization solution stage. The objective function is: In the formula, Let p be the set of arcs in period p, where p is the period index. , For unmanned delivery vehicles to pass through the arc Transportation costs, For 0-1 decision variables, =1 indicates that the vehicle passed through the arc. , =0 indicates that it is not passed through.
2. The method for re-optimizing cross-regional collaborative scheduling of unmanned delivery oriented towards dynamic delivery needs, as described in claim 1, is characterized in that... The vertex set for each period is defined as: In the formula, For fixed storage parking lots, For the set of customers to be served in the p-th period (note: p is the period index, lowercase), These are temporary parking lots used for waiting in the previous period. They are converted from customer locations where services were not completed and vehicles were parked, and they meet the following criteria: 。 3. The method for re-optimizing cross-regional collaborative scheduling of unmanned delivery oriented towards dynamic delivery needs, as described in claim 1, is characterized in that... The vehicle capacity constraint is expressed as follows: In the formula, k represents the set of unmanned delivery vehicles. For customer i's order demand, 0-1 decision variables ( =1 indicates that vehicle k serves customer i). This represents the maximum rated load capacity of the unmanned delivery vehicle k.
4. The method for cross-regional collaborative scheduling and re-optimization of unmanned delivery oriented to dynamic delivery needs, as described in claim 1, is characterized in that... The time window constraint is expressed as follows: In the formula, For customer C's time window, The service start time is when the unmanned delivery vehicle arrives at customer c, and the following conditions are met: in, For the service time of node i, The transit time from node i to node j (including the service time of node i).
5. The method for cross-regional collaborative scheduling and re-optimization of unmanned delivery oriented to dynamic delivery needs, as described in claim 1, is characterized in that... The optimization solution stage includes: using 6 removal operators to perform destruction operations, including cost reduction removal operators and exchange reduction removal operators; using a time window compatible O(1) insertion method to perform repair operations; dynamically adjusting the selection probability according to the operator performance and performing adaptive updates; and using an acceptance criterion to simulate annealing acceptance.
6. The cross-regional collaborative scheduling and re-optimization method for unmanned delivery oriented towards dynamic delivery needs, as described in claim 5, is characterized in that... The criterion for determining the cost reduction removal operator is: In the formula, , For customer v, the nodes before and after in the original path. , The arc node of the path to be inserted. The distance cost from node x to node y; The condition for determining the swap reduction removal operator is: In the formula, , For paired customers, , for Nodes before and after the original path , for The nodes before and after the original path.
7. The cross-regional collaborative scheduling and re-optimization method for unmanned delivery oriented towards dynamic delivery needs, as described in claim 5, is characterized in that... The time window compatibility O(1) insertion method includes: Feasibility verification is performed by verifying the following conditions: In the formula, For the first in path L The earliest service start time for each node. For the first in path L The latest service start time for each node. , Let c be the adjacent node in path L, and c be the customer to be inserted. The calculation methods for e(L) and l(L) are as follows: In the formula, Let p be the start time of the p-th period. The duration of the re-optimization process for the scheduling scheme. H represents the number of customers in the path, and H represents the total operating time. , For nodes The time window parameters.
8. The method for cross-regional collaborative scheduling and re-optimization of unmanned delivery oriented to dynamic delivery needs, as described in claim 5, is characterized in that... The acceptance criteria include: setting an initial temperature. And propose a decreasing rate of deceleration To lower the temperature: In the formula, This represents the current temperature at the t-th iteration.