Logistics scheduling management method based on ant colony algorithm

Abstract

The invention discloses a logistics scheduling management method based on an ant colony algorithm. The method comprises the following steps: constructing a logistics network model constrained by a capacity time window; carrying out ant colony search to generate a scheduling scheme set and carrying out local modeling to obtain a scheduling constraint manifold; converting the scheduling scheme into a vehicle event graph and introducing a time phase to obtain an event phase code; defining a manifold projection operator on the scheduling constraint manifold, mapping an event phase code into a feasible code, and setting a water wave parameter according to a difference value; a water wave model with event phase codes as individuals is constructed, candidate codes are generated through translation, torsion and event subgraph reconnection, and projection is feasible; constructing a cost field and a risk field according to the transportation cost and the time window default risk, and updating event phase codes and ant colony parameters according to a gradient; and when the model converges, outputting a logistics scheduling scheme and generating a vehicle warehouse instruction. Scheduling constraint manifold and water wave optimization are introduced on the basis of ant colony search, and cost and risk are balanced under capacity constraint.

Description

Technical Field

[0001] This invention relates to the field of intelligent logistics and operations optimization technology, and in particular to a logistics scheduling and management method based on ant colony algorithm. Background Technology

[0002] Current logistics scheduling management is generally geared towards urban delivery, cross-regional trunk transportation, and integrated warehousing, trunking, and distribution scenarios. The scheduling objects include orders, stations, vehicles, and road resources. Common technical approaches are usually based on vehicle routing problem modeling, using ant colony algorithms, genetic algorithms, tabu search, and simulated annealing for solutions. Under capacity constraints and time window constraints, vehicle travel paths and customer service sequences are generated. These methods often directly construct paths on simple logistics network graphs, calculate loading feasibility based on order demand and vehicle load limits, estimate arrival times based on the start and end times of service time windows and estimated travel times, and handle time window defaults through penalty functions. In terms of scheduling scheme expression, vehicle behavior is generally described using path sequences, task sequences, or time schedules, without introducing a unified event graph structure and time phase encoding, and without distinguishing the time structure characteristics at the event level.

[0003] In terms of search mechanisms, ant colony-based scheduling methods typically use the edges between client nodes as search units, control the path selection probability through pheromone matrices and heuristic factor matrices, and improve the weight of good paths through pheromone evaporation and incremental updates. Some studies have attempted to combine local search or quadratic optimization algorithms to improve solution quality, but most scheduling schemes remain at the level of discrete path representation, without constructing a scheduling constraint manifold in the feature space, lacking modeling of the continuous relationship between capacity constraints, time window constraints and scheduling structure, and failing to systematically interface linear local models with ant colony search output. Under high-dimensional constraints, ant colony search tends to repeatedly reinforce similar paths in local regions, making it difficult to simultaneously consider capacity utilization and time window satisfaction margins, and lacking the ability to balance cost and risk in complex scenarios.

[0004] In research on high-dimensional continuous search algorithms, water wave optimization algorithms have been used for function optimization, parameter optimization, and a small number of scheduling problems. However, common approaches often use simple real vectors as water wave individual codes, adjust wavelength and wave height through fixed rules, and control the search step size and perturbation amplitude. Wavelength and wave height updates are usually driven by changes in the objective function. Existing methods typically do not use feasible event phase codes on the scheduling constraint manifold as water wave individuals, do not construct water wave parameters based on capacity deviation, time window deviation, and transportation cost changes, do not explicitly construct cost and risk fields on the logistics network graph and calculate gradients, and do not combine the cost field gradient and risk field gradient into a propagation direction through coupling coefficients, and then apply it in reverse to update event phase codes and ant colony search model parameters. They lack refined modeling methods for the coupling relationship between order event sequences, time phase structures, and global search algorithms in logistics scheduling problems.

[0005] Therefore, how to provide a logistics scheduling and management method based on ant colony algorithm is a problem that urgently needs to be solved by those skilled in the art. Summary of the Invention

[0006] One objective of this invention is to propose a logistics scheduling and management method based on the ant colony algorithm. This invention constructs a scheduling constraint manifold, event phase encoding, and manifold projection operator, and jointly optimizes vehicle travel paths and customer service sequences under the guidance of coupled gradients in the cost field and risk field. This generates a target logistics scheduling scheme, as well as vehicle instructions and warehouse instructions, that satisfy capacity constraints and time window constraints. It has the advantages of balancing scheduling costs and default risks, adapting to complex constraint scenarios, and improving solution quality and convergence stability.

[0007] A logistics scheduling and management method based on ant colony algorithm according to an embodiment of the present invention includes the following steps: Acquire order, station, vehicle, and road data to construct a logistics network diagram and a scheduling constraint model that includes capacity and time windows; An ant colony search model is established based on the logistics network graph and scheduling constraint model. The set of scheduling schemes is generated by iterative search. The set of scheduling schemes is then modeled locally linearly to obtain the scheduling constraint manifold. The scheduling scheme is represented as an event graph consisting of vehicle events, and the time phase is registered at the event node to form an event phase code; Define a manifold projection operator on the scheduling-constrained manifold to map the event phase code to the feasible event phase code on the scheduling-constrained manifold, and determine the wavelength and wave height based on the mapping difference; A water wave optimization scheduling model is constructed, with event phase codes as water wave individuals. Based on wavelength and wave height, phase translation, phase torsion, and event subgraph reconnection are selected to generate candidate event phase codes, and the manifold projection operator is called for feasibility. Based on transportation costs and customer time window default risk, a cost field and a risk field are constructed. The gradients of the cost field and the risk field are calculated. The propagation direction is obtained by combining the coupling coefficients, which is used to update the event phase encoding and ant colony search model parameters. When the water wave optimization scheduling model converges, it outputs the target logistics scheduling scheme based on the event phase encoding, and generates vehicle instructions and warehouse instructions.

[0008] Optionally, the construction of the logistics network diagram and scheduling constraint model specifically includes: Acquire order data, site data, vehicle data, and road data to form order sets, site sets, vehicle sets, and road sets; Each order in the order set includes an order identifier, order demand quantity, order originating site, order destination site, service time window start time, and service time window end time; The site collection includes warehouse sites, transit center sites, and customer sites, and each site includes a site identifier; Each vehicle in the vehicle set includes a vehicle identifier, maximum vehicle load capacity, and vehicle availability time range; Each road in the road set includes a starting station, an ending station, road length, and estimated travel time; A logistics network graph is constructed based on the set of stations and the set of roads. The logistics network graph uses the set of stations as the node set and the set of roads as the edge set. A scheduling constraint model is established based on the order set, vehicle set, and logistics network diagram. The scheduling constraint model includes capacity constraint and time window constraint. The capacity constraint limits the total order demand of each vehicle to not exceed the vehicle's maximum load capacity. The time window constraint limits the service start time of each order to be between the start time and end time of the service time window.

[0009] Optionally, the generation of the scheduling constraint manifold specifically includes: In the logistics network diagram, pheromone intensity parameters and heuristic factor parameters are set for each road in the road set to form a pheromone matrix and a heuristic factor matrix. Ant colony size, number of iteration rounds, pheromone evaporation coefficient, and pheromone increment coefficient are set as parameters of the ant colony search model. In each iteration, the starting station and initial vehicle state are determined for each ant according to the scheduling constraint model. Based on the pheromone matrix and heuristic factor matrix, the driving road and the order to be served are selected using probability rules to generate a candidate scheduling scheme that includes the vehicle driving path and the customer service order. Perform capacity constraint checks and time window constraint checks on candidate scheduling schemes, delete candidate scheduling schemes that do not meet the scheduling constraint model, and add candidate scheduling schemes that meet the scheduling constraint model to the scheduling scheme set. When the preset iteration round is reached, feature vectors are constructed based on the total transportation cost, vehicle load utilization rate, and time window margin of each scheduling scheme in the scheduling scheme set. Local neighborhoods are divided in the feature vector space according to the distance relationship. In each local neighborhood, linear fitting is performed on the feature vector and the vehicle driving path and customer service order of the scheduling scheme to form a set of local linear models. The set of local linear models together constitutes the scheduling constraint manifold.

[0010] Optionally, the generation of the event phase code specifically includes: Select a scheduling scheme from the set of scheduling schemes, and define each vehicle departure, arrival, loading, unloading and return process in the scheduling scheme as a vehicle event, forming a set of vehicle events; Within the vehicle event set, each vehicle event is assigned a vehicle identifier, station identifier, order identifier, planned start time, and planned end time. Based on the vehicle travel routes and customer service order in the scheduling scheme, an event graph is constructed on the vehicle event set. The event graph uses vehicle events as nodes and direct edges that represent the order of events connect the vehicle events. Based on the estimated travel time, planned start time, and planned end time in the logistics network diagram, calculate the time phase for each vehicle event in the event diagram and record the time phase at the corresponding node; The vehicle events in the event graph are sorted according to vehicle identification and time phase. The sorting results are encoded as event phase codes, and a one-to-one correspondence is established between the event phase codes and the scheduling scheme to form an event phase code set.

[0011] Optionally, the definition and processing of the manifold projection operator specifically include: Define a manifold projection operator on a scheduling-constrained manifold. The manifold projection operator takes event phase codes as input and feasible event phase codes on the scheduling-constrained manifold as output. Select the target event phase code from the event phase code set, and select the local linear model that is close to the target event phase code as the target local linear model based on the position of the target event phase code in the feature vector space. The manifold projection operator is invoked to map the target event phase code to the feasible event phase code on the scheduling constraint manifold under the constraints of the target local linear model. The difference between the feature vector corresponding to the target event phase code and the feature vector corresponding to the feasible event phase code is calculated and decomposed into capacity deviation, time window deviation and transportation cost change. The wavelength is set based on the capacity deviation and time window deviation, and the wave height is set based on the transportation cost change and capacity deviation. When the capacity deviation is greater than the preset threshold, the wavelength is decreased and the wave height is increased. When the capacity deviation is not greater than the preset threshold, the wavelength is increased and the wave height is decreased. The time window deviation is used as a weighting factor in the wavelength setting, and the transportation cost change is used as a weighting factor in the wave height setting.

[0012] Optionally, the construction and processing of the water wave optimization scheduling model specifically includes: A water wave optimization scheduling model is constructed. The water wave optimization scheduling model uses the set of feasible event phase codes on the scheduling constraint manifold as the search space, the objective function composed of transportation cost, vehicle load utilization rate and time window satisfaction margin as the evaluation basis, and the wavelength, wave height and energy of the water wave individual as the control parameters. The event phase codes corresponding to the water wave individuals are updated through the propagation stage, refraction stage and disappearance stage. The feasible event phase encoding set on the scheduling constraint manifold is initialized as a set of water wave individuals. For each water wave individual, the initial wavelength, initial wave height, and initial energy are set. The number of propagation rounds, group size, reception threshold, and water wave energy attenuation coefficient are set as parameters of the water wave optimization scheduling model. In each round of propagation, the wavelength and wave height of each water wave are read. The length of the time phase segment in the event phase code participating in the update is determined based on the wavelength, and the amplitude of the time phase change is determined based on the wave height. The target non-commutative neighborhood operator is selected from the phase translation non-commutative neighborhood operator, the phase torsion non-commutative neighborhood operator, and the event subgraph reconnection non-commutative neighborhood operator to generate candidate event phase codes. The phase encoding of candidate events is called with the manifold projection operator to obtain the phase encoding of feasible events on the scheduling constraint manifold. Based on the phase encoding of feasible events, the transportation cost, vehicle load utilization rate, and time window satisfaction margin are calculated to form an updated feature vector. The updated feature vector is compared with the original feature vector of the corresponding water wave individual. When the evaluation of the updated feature vector on the objective function is better than that of the original feature vector, the phase encoding of the event of the water wave individual is updated with the phase encoding of feasible events. The energy of the water wave individual and whether to retain the water wave individual are adjusted according to the acceptance threshold and the water wave energy attenuation coefficient.

[0013] Optionally, the updating of the event phase encoding and ant colony search model parameters specifically includes: Based on the transportation cost data in the road set and the mileage data in the vehicle set, cost field values ​​are configured for the station set and road set on the logistics network diagram, forming a cost field with the station set and road set as the domain; Calculate the default risk of the service time window based on the start time and end time of the service time window in the order set and the time phase recorded in the event phase code. Configure risk field values ​​for the order originating station, the order destination station and the roads on the driving route to form a risk field corresponding to the cost field. For each event phase code in the event phase code set, the cost field gradient is calculated in the cost field and the risk field gradient is calculated in the risk field according to the corresponding vehicle travel path and customer service order. The cost field gradient represents the trend of transportation cost change along the event phase disturbance direction, and the risk field gradient represents the trend of default risk change along the event phase disturbance direction. The cost field gradient and risk field gradient are linearly combined according to the preset coupling coefficient to obtain the propagation direction. The propagation direction is input into the water wave optimization scheduling model to determine the adjustment direction and adjustment magnitude of the time phase in the event phase encoding, and input into the ant colony search model to adjust the parameters of the ant colony search model.

[0014] Optionally, the output of the target logistics scheduling scheme specifically includes: In the water wave optimization scheduling model, the update status of the water wave individual set is monitored, and the change amplitude of the corresponding feature vector of the water wave individual within adjacent propagation rounds is recorded. When the change amplitude of the feature vector of all water wave individuals within a consecutive preset propagation round is less than the convergence threshold, the water wave optimization scheduling model is determined to have converged. Select a target water wave individual from the set of water wave individuals. The feature vector corresponding to the target water wave individual takes the minimum value in the comprehensive evaluation of transportation cost, vehicle load utilization rate, and time window satisfaction margin. Read the event phase code corresponding to the target water wave individual, map the event phase code to the event graph, and sort the vehicle events according to the vehicle identifier and time phase to generate the vehicle driving path and customer service order for each vehicle. Under the constraints of the scheduling model, the total transportation cost, vehicle load utilization rate and time window satisfaction margin are calculated based on the vehicle travel route and customer service sequence. When both capacity constraints and time window constraints are satisfied, the corresponding scheduling result is determined as the target logistics scheduling scheme. Vehicle instructions are generated based on the target logistics scheduling plan. These vehicle instructions include vehicle identification, departure station, departure time, travel route, loading order list, and unloading station list. Warehouse instructions are also generated based on the target logistics scheduling plan. These warehouse instructions include the loading time schedule, unloading time schedule, and work sequence for each warehouse station.

[0015] The beneficial effects of this invention are: This invention introduces scheduling constraint manifolds and event phase encoding into the traditional vehicle routing framework, unifying capacity constraints, time window constraints, vehicle travel paths, and customer service order into a single feature space representation. Furthermore, it utilizes manifold projection operators to confine the search process to the vicinity of the scheduling constraint manifold, ensuring that event phase encoding is consistently adjusted around capacity and time window constraints. Compared to scheduling methods that directly rely on penalty terms to filter infeasible solutions on the path sequence, this invention significantly reduces the number of infeasible scheduling solutions, increases the proportion of feasible scheduling solutions in the search space, and allows for better control of capacity utilization and time window satisfaction margins from the initial optimization stage. This provides a more stable constraint environment for water wave optimization scheduling models and ant colony search models.

[0016] This invention constructs a water wave optimization scheduling model, treating feasible event phase encoding as individual water waves. Wavelength and wave height are adaptively adjusted by the combined effects of capacity deviation, time window deviation, and transportation cost changes. Furthermore, non-commutative neighborhood operators such as phase translation, phase twisting, and event subgraph reconnection are used to achieve fine perturbation of the vehicle event time structure and vehicle allocation structure. The coupling of cost field gradient and risk field gradient on the logistics network graph forms a propagation direction, guiding water wave individuals to evolve along a path that balances transportation costs and time window default risks. Simultaneously, the parameters of the ant colony search model are adjusted in reverse to enhance pheromone intensity in low-cost and low-risk path regions. Through the linkage mechanism between event phase encoding, manifold projection operators, the water wave optimization scheduling model, and the ant colony search model, this invention improves the overall performance of the target logistics scheduling scheme in terms of transportation costs, vehicle load utilization, and time window satisfaction margin, while satisfying capacity and time window constraints. It is suitable for large-scale intelligent logistics scheduling scenarios with complex constraints. Attached Figure Description

[0017] The accompanying drawings are provided to further illustrate the invention and form part of the specification. They are used in conjunction with embodiments of the invention to explain the invention and do not constitute a limitation thereof. In the drawings:

[0018] Figure 1 This is a flowchart of a logistics scheduling and management method based on ant colony algorithm proposed in this invention; Figure 2 This is a schematic diagram illustrating the generation of the scheduling constraint manifold for a logistics scheduling and management method based on the ant colony algorithm proposed in this invention. Figure 3 This is a schematic diagram of the water wave optimization scheduling model structure for a logistics scheduling and management method based on the ant colony algorithm proposed in this invention. Detailed Implementation

[0019] The present invention will now be described in further detail with reference to the accompanying drawings. These drawings are simplified schematic diagrams, illustrating only the basic structure of the invention, and therefore only show the components relevant to the invention.

[0020] refer to Figure 1-3 A logistics scheduling and management method based on ant colony algorithm includes the following steps: Acquire order, station, vehicle, and road data to construct a logistics network diagram and a scheduling constraint model that includes capacity and time windows; An ant colony search model is established based on the logistics network graph and scheduling constraint model. The set of scheduling schemes is generated by iterative search. The set of scheduling schemes is then modeled locally linearly to obtain the scheduling constraint manifold. The scheduling scheme is represented as an event graph consisting of vehicle events, and the time phase is registered at the event node to form an event phase code; Define a manifold projection operator on the scheduling-constrained manifold to map the event phase code to the feasible event phase code on the scheduling-constrained manifold, and determine the wavelength and wave height based on the mapping difference; A water wave optimization scheduling model is constructed, with event phase codes as water wave individuals. Based on wavelength and wave height, phase translation, phase torsion, and event subgraph reconnection are selected to generate candidate event phase codes, and the manifold projection operator is called for feasibility. Based on transportation costs and customer time window default risk, a cost field and a risk field are constructed. The gradients of the cost field and the risk field are calculated. The propagation direction is obtained by combining the coupling coefficients, which is used to update the event phase encoding and ant colony search model parameters. When the water wave optimization scheduling model converges, it outputs the target logistics scheduling scheme based on the event phase encoding, and generates vehicle instructions and warehouse instructions.

[0021] In this embodiment, the construction of the logistics network diagram and scheduling constraint model specifically includes: Acquire order data, site data, vehicle data, and road data to form order sets, site sets, vehicle sets, and road sets. The order set describes the transportation tasks to be performed, the site set describes the locations of warehouse sites, transfer center sites, and customer sites, the vehicle set describes the transportation resources that can participate in scheduling, and the road set describes the traffic relationships and driving parameters between sites. Each order in the order set includes an order identifier, order demand quantity, order origin station, order destination station, service time window start time, and service time window end time. The order identifier is used to distinguish different orders, the order demand quantity is used to calculate the vehicle loading capacity, the order origin station and the order destination station are used to determine the transportation start point and transportation destination, and the service time window start time and service time window end time are used to limit the range of allowed service start times. Each site in the site set includes a site identifier, which is used to create nodes in the logistics network graph and establish references between the order set and the road set; Each vehicle in the vehicle set includes a vehicle identifier, maximum vehicle load capacity, and vehicle availability time range. The maximum vehicle load capacity is used to limit the total order demand allocated to the vehicle, and the vehicle availability time range is used to limit the start and end time of the vehicle's transportation task. Each road in the road set includes a starting station, an ending station, a road length, and an estimated travel time. The road length is used to estimate the transportation distance, and the estimated travel time is used to estimate the travel time between stations. A logistics network graph is constructed based on a set of stations and a set of roads. The logistics network graph uses the set of stations as the set of nodes and the set of roads as the set of edges. The connection relationship between the set of nodes and the set of edges represents the passable path between stations. A scheduling constraint model is established based on the order set, vehicle set, and logistics network graph. The scheduling constraint model includes capacity constraints and time window constraints. The capacity constraint limits the total order demand carried by each vehicle to no more than the vehicle's maximum load capacity. The time window constraint limits the service start time of each order to be between the start time and end time of the service time window. The scheduling constraint model serves as a unified constraint condition for the ant colony search model and the water wave optimization scheduling model.

[0022] In this embodiment, the generation of the scheduling constraint manifold specifically includes: In the logistics network diagram, pheromone intensity parameters and heuristic factor parameters are set for each road in the road set to form a pheromone matrix and a heuristic factor matrix. The pheromone matrix is ​​used to record the road's usage preference in each search, and the heuristic factor matrix is ​​used to record the static evaluation of the road in terms of distance, travel time, and transportation cost. The ant colony size, number of iteration rounds, pheromone evaporation coefficient, and pheromone increment coefficient are set. The ant colony size limits the number of ants participating in the search, the number of iteration rounds limits the number of pheromone updates, the pheromone evaporation coefficient controls the retention ratio of old pheromones, and the pheromone increment coefficient controls the enhancement of pheromones by high-quality scheduling schemes. The aforementioned parameters together constitute the ant colony search model parameters. In each iteration, a starting station and initial vehicle state are selected for each ant according to the scheduling constraint model. The starting station comes from the station set, and the initial vehicle state comes from the vehicle set. The scheduling constraint model excludes starting combinations that obviously do not meet the capacity constraint and time window constraint in the initial stage. Then, path selection is performed for each ant in a loop. During the path selection process, the values ​​in the pheromone matrix and heuristic factor matrix are read. A weighted evaluation is constructed based on the pheromone intensity and heuristic factor. The selection probability is calculated for the candidate road according to the weighted evaluation. The next road is selected from the candidate road set. At the same time, the pending orders associated with the target road are selected. The selected road is added to the vehicle driving path, and the selected order is added to the customer service order. This continues until there are no allocable orders or the vehicle available time range ends, forming a candidate scheduling scheme that includes the vehicle driving path and the customer service order. Candidate scheduling schemes undergo capacity constraint checks and time window constraint checks. In the capacity constraint check, the total order demand carried by each vehicle is calculated and compared with the corresponding vehicle's maximum load capacity. If the result exceeds the vehicle's maximum load capacity, the candidate scheduling scheme is marked as a capacity-incompatible scheme. In the time window constraint check, the order service start time is calculated based on the estimated road travel time and vehicle departure time and compared with the service time window start time and service time window end time. If the service start time is outside the service time window, the candidate scheduling scheme is marked as a time window-incompatible scheme. Capacity-incompatible and time window-incompatible schemes are removed from the candidate scheduling scheme set. Candidate scheduling schemes that pass both capacity and time window constraint checks are added to the scheduling scheme set. At the end of each iteration, a comprehensive evaluation value is calculated based on the transportation cost and time window default penalty for each scheduling scheme in the set. Based on the comprehensive evaluation value, the pheromone matrix undergoes evaporation and enhancement updates to ensure that scheduling schemes with lower comprehensive evaluation values ​​receive higher increments in pheromone intensity on the road. When the preset number of iterations is reached, the scheduling scheme with the lower comprehensive evaluation value is selected from the scheduling scheme set to form a scheduling sample set. A feature vector is constructed for each scheduling scheme in the scheduling sample set. The feature vector contains numerical features such as total transportation cost, vehicle load utilization rate, and time window satisfaction margin. Combined with the corresponding vehicle driving path structure and customer service sequence structure, local neighborhoods are divided in the feature vector space according to the distance relationship. In each local neighborhood, a linear fitting method is used to establish a linear relationship between the feature vector and the scheduling structure parameters, forming a set of local linear models. The set of local linear models together constitutes the scheduling constraint manifold.

[0023] In this embodiment, the generation of the event phase code specifically includes: Select a scheduling scheme from the set of scheduling schemes, and map each vehicle departure, arrival, loading, unloading and return process in the scheduling scheme to a single vehicle event, generating a set of vehicle events. Each vehicle event in the set of vehicle events contains a vehicle identifier, starting station, destination station, associated order set identifier and sequence number in the scheduling scheme, which is used to represent the discrete actions of the vehicle within the scheduling cycle. In the vehicle event set, each vehicle event is configured with a vehicle identifier, station identifier, order identifier, planned start time, and planned end time. The planned start time is obtained by adding the planned end time of the previous vehicle event to the estimated travel time in the logistics network diagram. The planned end time is obtained by adding the planned start time to the service duration. The service duration is generated by selecting the target time field from loading time, unloading time, and stopping time, so that the position of each vehicle event on the timeline is clear. Based on the vehicle travel routes and customer service order in the scheduling scheme, an event graph is constructed on the vehicle event set. The event graph uses vehicle events as nodes and direct edges that represent the order of events to connect vehicle events. Under the same vehicle identifier, the event graph connects the departure event to the arrival event, loading event, unloading event and return event according to the sequence number. Under different vehicle identifiers, an independent event subgraph structure is formed. Based on the estimated travel time, planned start time and planned end time in the logistics network diagram, the time phase is calculated for each vehicle event in the event diagram. The time phase is obtained by mapping the planned start time to a unified time reference axis and recording the time phase value at the corresponding vehicle event node to represent the time position of the vehicle event in the entire scheduling cycle. The vehicle events in the event graph are sorted according to vehicle identification and time phase. The vehicle identification is used as the group index, and the vehicle events are arranged from smallest to largest time phase. The sorting result is encoded as an event phase code. The event phase code continuously records the vehicle identification, station identification, order identification and corresponding time phase value. A one-to-one correspondence is established between the event phase code and the scheduling scheme to form an event phase code set, which provides event phase code input for the water wave optimization scheduling model.

[0024] In this embodiment, the definition and processing of the manifold projection operator specifically include: A manifold projection operator is defined on the scheduling-constrained manifold. The manifold projection operator takes the event phase code as input and the feasible event phase code on the scheduling-constrained manifold as output. Based on the capacity constraints, time window constraints and transportation cost characteristics recorded in the local linear model set, the manifold projection operator adjusts the scheduling scheme corresponding to the input event phase code so that the scheduling scheme corresponding to the output event phase code satisfies the scheduling-constrained model and falls near the scheduling-constrained manifold. Select a target event phase code from the event phase code set. Based on the position of the target event phase code in the feature vector space, select a local linear model that is close to the target event phase code as the target local linear model. Each feature vector in the feature vector space includes a cost component representing the total transportation cost, a capacity component representing the vehicle load utilization, a time window component representing the time window satisfaction, and other scheduling feature components describing the scheduling structure. The proximity relationship is determined by calculating the distance between the feature vector corresponding to the target event phase code and the central feature vector of the local linear model and selecting the local linear model with the smaller distance. The manifold projection operator is invoked to map the target event phase code to the feasible event phase code on the scheduling constraint manifold under the constraints of the target local linear model. The corresponding feature vectors of the target event phase code and the feasible event phase code are obtained. The difference between the two feature vectors is calculated in the feature vector space. The change in transportation cost is obtained based on the difference of the cost component, the capacity deviation is obtained based on the difference of the capacity component, and the time window deviation is obtained based on the difference of the time window component. The change in transportation cost, the capacity deviation, and the time window deviation represent the degree of change in the cost index, the capacity index, and the time window index, respectively. The wavelength is set based on the capacity deviation and time window deviation, and the wave height is set based on the transportation cost change and capacity deviation. When the capacity deviation is greater than a preset threshold, the wavelength is decreased and the wave height is increased. When the capacity deviation is not greater than the preset threshold, the wavelength is increased and the wave height is decreased. The time window deviation is used as a weighting factor in the wavelength setting, and the transportation cost change is used as a weighting factor in the wave height setting. The updated wavelength and wave height are written into the corresponding water wave individual parameters and used in the propagation update step of the water wave optimization scheduling model.

[0025] In this embodiment, the construction and processing of the water wave optimization scheduling model specifically includes: A water wave optimization scheduling model is constructed, based on the water wave optimization algorithm. In the water wave optimization algorithm, each water wave corresponds to an event phase code. The wavelength of each water wave controls the length of the time phase segment involved in modification in each round of update. The wave height of each water wave controls the amplitude of the time phase change. The energy of each water wave controls the number of rounds of update involved in the propagation and refraction phases. The water wave optimization scheduling model uses the set of feasible event phase codes on the scheduling constraint manifold as the search space. An objective function is set to evaluate the event phase codes. The objective function is formed by combining transportation cost indicators, vehicle load utilization rate indicators, and time window satisfaction margin indicators according to preset weights. The smaller the objective function value, the closer it is to the expected target state in terms of transportation cost, load utilization, and time window satisfaction. The water wave optimization scheduling model iteratively updates the event phase codes corresponding to each water wave in the propagation, refraction, and disappearance phases based on the objective function. The feasible event phase encoding set on the scheduling constraint manifold is initialized as a set of water wave individuals. An initial wavelength, initial wave height, and initial energy are set for each water wave individual in the set. The initial wavelength is used to limit the length of the continuous time phase segment in the event phase encoding, the initial wave height is used to limit the magnitude of the time phase change, and the initial energy is used to limit the upper limit of the number of rounds in which the water wave individual participates in the propagation and refraction phases. At the same time, the number of propagation rounds, the group size, the reception threshold, and the water wave energy attenuation coefficient are set as parameters of the water wave optimization scheduling model. In each round of propagation, the wavelength and wave height of each water wave individual in the water wave individual set are read. Based on the wavelength, a continuous time phase segment is selected in the event phase encoding. Based on the wave height, the change in time phase is calculated. Among the phase translation non-commutative neighborhood operator, the phase torsion non-commutative neighborhood operator, and the event subgraph reconnection non-commutative neighborhood operator, a target non-commutative neighborhood operator is selected according to a preset selection probability. The phase translation non-commutative neighborhood operator uniformly increases the offset of the time phase of continuous vehicle events under the same vehicle identifier. The phase torsion non-commutative neighborhood operator adjusts the adjacent time phase interval of the vehicle event sequence containing the service time window under the same vehicle identifier. The event subgraph reconnection non-commutative neighborhood operator extracts the event subgraph in the vehicle event set and changes the vehicle identifier and predecessor and successor relationships in the event subgraph. After the target non-commutative neighborhood operator is applied, candidate event phase codes are generated. The phase encoding of candidate events is called with the manifold projection operator to obtain the phase encoding of feasible events on the scheduling constraint manifold. Based on the phase encoding of feasible events, the transportation cost, vehicle load utilization rate and time window satisfaction margin are calculated to form an updated feature vector. The updated feature vector is substituted into the objective function to calculate the updated objective function value. The original feature vector of the water wave individual is substituted into the objective function to calculate the original objective function value. When the updated objective function value is less than the original objective function value, the phase encoding of the water wave individual is updated with the phase encoding of feasible events. The update result is recorded based on the acceptance threshold. The energy of the water wave individual is reduced according to the water wave energy attenuation coefficient. When the energy of the water wave individual is lower than the preset energy threshold, the corresponding water wave individual is deleted from the water wave optimization scheduling model.

[0026] In this embodiment, the updating of the event phase encoding and ant colony search model parameters specifically includes: Based on the transportation cost data in the road set and the mileage data in the vehicle set, cost field values ​​are configured for the station set and road set on the logistics network diagram, forming a cost field with the station set and road set as the domain. During the configuration process, for each road in the road set, the road cost field value is calculated based on the road length, unit transportation price, historical average fuel consumption, and mileage. For each station in the station set, the road cost field values ​​with the same starting station identifier and the same ending station identifier are weighted and summed, and the loading and unloading costs are added to obtain the station cost field value. The default risk of a time window is calculated based on the start time and end time of the service time window in the order set and the time phase recorded in the event phase code. Risk field values ​​are configured for the order originating station, the order destination station, and the roads on the driving route, forming a risk field with the station set and the road set as the domain. When calculating the default risk of a time window, the time phase is first compared with the start time and end time of the service time window to obtain the early service duration and the delayed service duration. Then, it is converted into a time window default risk index by combining the preset default penalty parameters. When configuring the risk field values, the time window default risk index is assigned to the order originating station, the order destination station, and the roads on the corresponding driving route. The risk field values ​​are accumulated at the station locations to obtain the station risk field values ​​and at the road locations to obtain the road risk field values. For each event phase code in the event phase code set, the cost field gradient is calculated in the cost field and the risk field gradient is calculated in the risk field according to the corresponding vehicle travel path and customer service sequence. When calculating the cost field gradient, the time phase increment is selected in the event phase code, and the time phase of the vehicle event is adjusted in the direction of time phase increase and time phase decrease respectively. The station cost field value and the road cost field value are re-accumulated in the cost field. The cost field gradient direction and the cost field gradient magnitude are obtained by comparing the total cost before and after adjustment. When calculating the risk field gradient, the time phase is adjusted in the same way, and the station risk field value and the road risk field value are accumulated in the risk field. The risk field gradient direction and the risk field gradient magnitude are obtained by comparing the total risk before and after adjustment. The cost field gradient and risk field gradient are weighted and combined according to the preset coupling coefficients to obtain the propagation direction. The propagation direction is input into the water wave optimization scheduling model to determine the adjustment direction and magnitude of the time phase in the event phase encoding, and input into the ant colony search model to adjust the parameters of the ant colony search model. The preset coupling coefficients include the cost field coupling coefficient and the risk field coupling coefficient. The cost field coupling coefficient is used to represent the proportion of the cost field gradient in the propagation direction, and the risk field coupling coefficient is used to represent the proportion of the risk field gradient in the propagation direction. The values ​​of the two coupling coefficients are between zero and one, and the sum of the two is set to one. The initial values ​​are obtained based on historical scheduling data. The cost field coupling coefficient and the risk field coupling coefficient are set by statistically comparing the ratio between the unit transportation cost fluctuation and the unit default loss, and are adjusted in the system parameter configuration according to the business strategy.

[0027] In this embodiment, the output of the target logistics scheduling scheme specifically includes: In the water wave optimization scheduling model, the update status of the water wave individual set is monitored. The difference norm between the feature vector corresponding to the water wave individual in each propagation round and the feature vector of the previous propagation round is recorded. The difference norm is compared with the convergence threshold. When the difference norm of all water wave individuals in a consecutive preset propagation round is less than the convergence threshold, the state of the water wave optimization scheduling model is marked as converged. In the convergence state, the water wave individual with the smallest feature vector evaluation value is extracted from the water wave individual set as the target water wave individual. The feature vector evaluation value is obtained by combining the transportation cost, vehicle load utilization rate, and time window satisfaction margin according to the preset weight coefficient. The event phase code corresponding to the target water wave individual is read and the event phase code is input into the event graph generation process. In the event graph, the vehicle events are arranged from small to large according to the vehicle identifier and time phase, forming the vehicle driving path and customer service order for each vehicle. Under the constraints of the scheduling model, the total order demand of each vehicle, the total transportation cost, and the relationship between the service start time of each order and the service time window start time and service time window end time are calculated based on the vehicle driving route and customer service order. When the total order demand of all vehicles does not exceed the maximum load of the corresponding vehicle and the service start time of each order is between the service time window start time and service time window end time, the scheduling result composed of the vehicle driving route and customer service order is confirmed as the target logistics scheduling scheme. Vehicle instructions are generated based on the target logistics scheduling plan. Each vehicle instruction includes a vehicle identifier, departure station, departure time, travel route, loading order list, and unloading station list. Warehouse instructions are also generated based on the target logistics scheduling plan. Each warehouse station includes a loading time schedule, unloading time schedule, and work sequence, which are used to drive the execution system to issue scheduling results.

[0028] Example 1: To verify the feasibility of this invention in practice, it was applied to the daily delivery scheduling of an integrated warehousing and distribution logistics park. Within a complete scheduling cycle, the park needs to replenish stock for multiple supermarkets and e-commerce self-pickup points, handling approximately 500 orders across about 50 sites, including warehouses, transfer points, and final receiving points. The fleet consists of a combination of approximately 30 medium-heavy-duty long-haul trucks and small-to-medium-sized delivery vehicles. The original scheduling method, based on a conventional ant colony algorithm, directly searches for paths between customer nodes, handling capacity overload and time window defaults through penalty coefficients. While this seems to provide feasible routes, in actual execution, vehicle departures frequently occur. With a load factor of only around 60%, some routes having long mileage, and store arrival times concentrated at or even exceeding the time window, dispatchers had to frequently reassign vehicles and adjust departure times, resulting in numerous manual interventions. Operational statistics over a period of time showed that the proportion of orders defaulting within the time window remained in the range of 7% to 8%, and the average vehicle load utilization rate remained below 70% for a long time. Complaints from stores about delayed replenishment and records of default compensation continued to exist. The operator hoped to improve vehicle load utilization, reduce the risk of defaulting within the time window, and control total transportation costs through algorithm upgrades without increasing the number of vehicles.

[0029] In this scenario, comparative experiments were conducted using data from the same scheduling period. Order data, station data, vehicle data, and road data were organized from the business system into order sets, station sets, vehicle sets, and road sets, respectively. A logistics network graph was established with warehouses, transfer points, and stores as nodes and passable roads as edges. Maximum load and available time ranges were set for each vehicle, and service time window start and end times were set for each order, forming a scheduling constraint model composed of capacity constraints and time window constraints. Based on the same fundamental data, one set of experiments used the original ant colony scheduling method, while the other set used the method of this invention. This invention first constructs an ant colony search model based on the logistics network graph and scheduling constraint model. Iteratively, a set of scheduling schemes is generated using the pheromone matrix and heuristic factor matrix. Then, local linear fitting is performed within the feature space composed of total transportation cost, vehicle load utilization rate, and time window margin to form a scheduling constraint manifold. Finally, each scheduling scheme is decomposed into departure, arrival, loading, and unloading processes. A set of vehicle events, consisting of a set of events and a return field, is used to construct an event graph. A time phase is assigned to each vehicle event, generating an event phase code. A manifold projection operator is defined on the scheduling constraint manifold to map event phase codes that deviate from the constraints back to regions that meet capacity and time window requirements. The difference between the eigenvectors before and after projection is used to obtain capacity deviation, time window deviation, and transportation cost change, which are used to adjust the wavelength and wave height of individual water waves. The water wave optimization scheduling model uses event phase codes as individual water waves and employs three types of non-commutative neighborhood operators—phase translation, phase twisting, and event subgraph reconnection—to perturb the time structure and vehicle allocation structure of vehicle events. Combined with the manifold projection operator, the feasibility of the solution is maintained. Simultaneously, a cost field and a risk field are constructed on the logistics network graph. The gradients of the cost field and the risk field are calculated based on transportation costs and time window default risks. The propagation direction is obtained through coupling coefficient combination, guiding the update of the event phase codes. The parameters of the ant colony search model are fine-tuned, gradually focusing the search on regions with lower costs and lower default risks.

[0030] To provide a quantitative comparison, experiments were conducted on a dataset with approximately 500 orders, 30 vehicles, and 50 stations. Both methods used identical order, vehicle, and station sets, as well as the same cost accounting rules. The comparative metrics included total transportation cost, average vehicle load utilization, percentage of defaulted orders within a time window, average arrival advance margin, percentage of infeasible candidate solutions, algorithm runtime, average number of service stations per vehicle, and average mileage per vehicle. The statistical results are shown in Table 1.

[0031] Table 1 Comparison of the two scheduling methods

[0032] As can be seen from the table, when the total transportation cost is taken as a benchmark of 100.0, the method of this invention reduces the cost to 88.7, increases the average vehicle load utilization rate from 68.2% to 83.5%, reduces the proportion of defaulted orders within the time window from 7.5% to 1.3%, increases the average arrival advance margin from 6.4 minutes to 14.8 minutes, reduces the proportion of infeasible candidate solutions from 21.6% to 3.2%, and increases the algorithm running time from 1.00 to 1.15, which is slightly more than the original method. However, the average number of service stations per vehicle increases from 9.1 to 10.4, and the average mileage per vehicle decreases from 1.00 to 0.92. Under the condition of unchanged vehicle scale, it comprehensively demonstrates higher load utilization, more sufficient time window margin, and more compact route structure, indicating that the present invention can effectively improve cost control and service stability in the scenario of integrated warehousing and distribution scheduling.

[0033] The above description is only a preferred embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any equivalent substitutions or modifications made by those skilled in the art within the scope of the technology disclosed in the present invention, based on the technical solution and inventive concept of the present invention, should be covered within the scope of protection of the present invention.

Claims

1. A logistics scheduling and management method based on ant colony algorithm, characterized in that, Includes the following steps: Acquire order, station, vehicle, and road data to construct a logistics network diagram and a scheduling constraint model that includes capacity and time windows; An ant colony search model is established based on the logistics network graph and scheduling constraint model. The set of scheduling schemes is generated by iterative search. The set of scheduling schemes is then modeled locally linearly to obtain the scheduling constraint manifold. The scheduling scheme is represented as an event graph consisting of vehicle events, and the time phase is registered at the event node to form an event phase code; Define a manifold projection operator on the scheduling-constrained manifold to map the event phase code to the feasible event phase code on the scheduling-constrained manifold, and determine the wavelength and wave height based on the mapping difference; A water wave optimization scheduling model is constructed, with event phase codes as water wave individuals. Based on wavelength and wave height, phase translation, phase torsion, and event subgraph reconnection are selected to generate candidate event phase codes, and the manifold projection operator is called for feasibility. Based on transportation costs and customer time window default risk, a cost field and a risk field are constructed. The gradients of the cost field and the risk field are calculated. The propagation direction is obtained by combining the coupling coefficients, which is used to update the event phase encoding and ant colony search model parameters. When the water wave optimization scheduling model converges, it outputs the target logistics scheduling scheme based on the event phase encoding, and generates vehicle instructions and warehouse instructions.

2. The logistics scheduling and management method based on ant colony algorithm according to claim 1, characterized in that, The construction of the logistics network diagram and scheduling constraint model specifically includes: Acquire order data, site data, vehicle data, and road data to form order sets, site sets, vehicle sets, and road sets; Each order in the order set includes an order identifier, order demand quantity, order originating site, order destination site, service time window start time, and service time window end time; The site collection includes warehouse sites, transit center sites, and customer sites, and each site includes a site identifier; Each vehicle in the vehicle set includes a vehicle identifier, maximum vehicle load capacity, and vehicle availability time range; Each road in the road set includes a starting station, an ending station, road length, and estimated travel time; A logistics network graph is constructed based on the set of stations and the set of roads. The logistics network graph uses the set of stations as the node set and the set of roads as the edge set. A scheduling constraint model is established based on the order set, vehicle set, and logistics network diagram. The scheduling constraint model includes capacity constraint and time window constraint. The capacity constraint limits the total order demand of each vehicle to not exceed the vehicle's maximum load capacity. The time window constraint limits the service start time of each order to be between the start time and end time of the service time window.

3. The logistics scheduling and management method based on ant colony algorithm according to claim 1, characterized in that, The generation of the scheduling constraint manifold specifically includes: In the logistics network diagram, pheromone intensity parameters and heuristic factor parameters are set for each road in the road set to form a pheromone matrix and a heuristic factor matrix. Ant colony size, number of iteration rounds, pheromone evaporation coefficient, and pheromone increment coefficient are set as parameters of the ant colony search model. In each iteration, the starting station and initial vehicle state are determined for each ant according to the scheduling constraint model. Based on the pheromone matrix and heuristic factor matrix, the driving road and the order to be served are selected using probability rules to generate a candidate scheduling scheme that includes the vehicle driving path and the customer service order. Perform capacity constraint checks and time window constraint checks on candidate scheduling schemes, delete candidate scheduling schemes that do not meet the scheduling constraint model, and add candidate scheduling schemes that meet the scheduling constraint model to the scheduling scheme set. When the preset iteration round is reached, feature vectors are constructed based on the total transportation cost, vehicle load utilization rate, and time window margin of each scheduling scheme in the scheduling scheme set. Local neighborhoods are divided in the feature vector space according to the distance relationship. In each local neighborhood, linear fitting is performed on the feature vector and the vehicle driving path and customer service order of the scheduling scheme to form a set of local linear models. The set of local linear models together constitutes the scheduling constraint manifold.

4. The logistics scheduling and management method based on ant colony algorithm according to claim 1, characterized in that, The generation of the event phase code specifically includes: Select a scheduling scheme from the set of scheduling schemes, and define each vehicle departure, arrival, loading, unloading and return process in the scheduling scheme as a vehicle event, forming a set of vehicle events; Within the vehicle event set, each vehicle event is assigned a vehicle identifier, station identifier, order identifier, planned start time, and planned end time. Based on the vehicle travel routes and customer service order in the scheduling scheme, an event graph is constructed on the vehicle event set. The event graph uses vehicle events as nodes and direct edges that represent the order of events connect the vehicle events. Based on the estimated travel time, planned start time, and planned end time in the logistics network diagram, calculate the time phase for each vehicle event in the event diagram and record the time phase at the corresponding node; The vehicle events in the event graph are sorted according to vehicle identification and time phase. The sorting results are encoded as event phase codes, and a one-to-one correspondence is established between the event phase codes and the scheduling scheme to form an event phase code set.

5. The logistics scheduling and management method based on ant colony algorithm according to claim 1, characterized in that, The definition and processing of the manifold projection operator specifically include: Define a manifold projection operator on a scheduling-constrained manifold. The manifold projection operator takes event phase codes as input and feasible event phase codes on the scheduling-constrained manifold as output. Select the target event phase code from the event phase code set, and select the local linear model that is close to the target event phase code as the target local linear model based on the position of the target event phase code in the feature vector space. The manifold projection operator is invoked to map the target event phase code to the feasible event phase code on the scheduling constraint manifold under the constraints of the target local linear model. The difference between the feature vector corresponding to the target event phase code and the feature vector corresponding to the feasible event phase code is calculated and decomposed into capacity deviation, time window deviation and transportation cost change. The wavelength is set based on the capacity deviation and time window deviation, and the wave height is set based on the transportation cost change and capacity deviation. When the capacity deviation is greater than the preset threshold, the wavelength is decreased and the wave height is increased. When the capacity deviation is not greater than the preset threshold, the wavelength is increased and the wave height is decreased. The time window deviation is used as a weighting factor in the wavelength setting, and the transportation cost change is used as a weighting factor in the wave height setting.

6. The logistics scheduling and management method based on ant colony algorithm according to claim 1, characterized in that, The construction and processing of the water wave optimization scheduling model specifically includes: A water wave optimization scheduling model is constructed. The water wave optimization scheduling model uses the set of feasible event phase codes on the scheduling constraint manifold as the search space, the objective function composed of transportation cost, vehicle load utilization rate and time window satisfaction margin as the evaluation basis, and the wavelength, wave height and energy of the water wave individual as the control parameters. The event phase codes corresponding to the water wave individuals are updated through the propagation stage, refraction stage and disappearance stage. The feasible event phase encoding set on the scheduling constraint manifold is initialized as a set of water wave individuals. For each water wave individual, the initial wavelength, initial wave height, and initial energy are set. The number of propagation rounds, group size, reception threshold, and water wave energy attenuation coefficient are set as parameters of the water wave optimization scheduling model. In each round of propagation, the wavelength and wave height of each water wave are read. The length of the time phase segment in the event phase code participating in the update is determined based on the wavelength, and the amplitude of the time phase change is determined based on the wave height. The target non-commutative neighborhood operator is selected from the phase translation non-commutative neighborhood operator, the phase torsion non-commutative neighborhood operator, and the event subgraph reconnection non-commutative neighborhood operator to generate candidate event phase codes. The phase encoding of candidate events is called with the manifold projection operator to obtain the phase encoding of feasible events on the scheduling constraint manifold. Based on the phase encoding of feasible events, the transportation cost, vehicle load utilization rate, and time window satisfaction margin are calculated to form an updated feature vector. The updated feature vector is compared with the original feature vector of the corresponding water wave individual. When the evaluation of the updated feature vector on the objective function is better than that of the original feature vector, the phase encoding of the event of the water wave individual is updated with the phase encoding of feasible events. The energy of the water wave individual and whether to retain the water wave individual are adjusted according to the acceptance threshold and the water wave energy attenuation coefficient.

7. The logistics scheduling and management method based on ant colony algorithm according to claim 1, characterized in that, The update of the event phase encoding and ant colony search model parameters specifically includes: Based on the transportation cost data in the road set and the mileage data in the vehicle set, cost field values ​​are configured for the station set and road set on the logistics network diagram, forming a cost field with the station set and road set as the domain; Calculate the default risk of the service time window based on the start time and end time of the service time window in the order set and the time phase recorded in the event phase code. Configure risk field values ​​for the order originating station, the order destination station and the roads on the driving route to form a risk field corresponding to the cost field. For each event phase code in the event phase code set, the cost field gradient is calculated in the cost field and the risk field gradient is calculated in the risk field according to the corresponding vehicle travel path and customer service order. The cost field gradient represents the trend of transportation cost change along the event phase disturbance direction, and the risk field gradient represents the trend of default risk change along the event phase disturbance direction. The cost field gradient and risk field gradient are linearly combined according to the preset coupling coefficient to obtain the propagation direction. The propagation direction is input into the water wave optimization scheduling model to determine the adjustment direction and adjustment magnitude of the time phase in the event phase encoding, and input into the ant colony search model to adjust the parameters of the ant colony search model.

8. The logistics scheduling and management method based on ant colony algorithm according to claim 1, characterized in that, The output of the target logistics scheduling scheme specifically includes: In the water wave optimization scheduling model, the update status of the water wave individual set is monitored, and the change amplitude of the corresponding feature vector of the water wave individual within adjacent propagation rounds is recorded. When the change amplitude of the feature vector of all water wave individuals within a consecutive preset propagation round is less than the convergence threshold, the water wave optimization scheduling model is determined to have converged. Select a target water wave individual from the set of water wave individuals. The feature vector corresponding to the target water wave individual takes the minimum value in the comprehensive evaluation of transportation cost, vehicle load utilization rate, and time window satisfaction margin. Read the event phase code corresponding to the target water wave individual, map the event phase code to the event graph, and sort the vehicle events according to the vehicle identifier and time phase to generate the vehicle driving path and customer service order for each vehicle. Under the constraints of the scheduling model, the total transportation cost, vehicle load utilization rate and time window satisfaction margin are calculated based on the vehicle travel route and customer service sequence. When both capacity constraints and time window constraints are satisfied, the corresponding scheduling result is determined as the target logistics scheduling scheme. Vehicle instructions are generated based on the target logistics scheduling plan. These vehicle instructions include vehicle identification, departure station, departure time, travel route, loading order list, and unloading station list. Warehouse instructions are also generated based on the target logistics scheduling plan. These warehouse instructions include the loading time schedule, unloading time schedule, and work sequence for each warehouse station.

Images