Intelligent logistics distribution path optimization method under multiple constraints

Abstract

The present invention provides an intelligent logistics distribution path optimization method under multiple constraints, which relates to the field of logistics distribution technology, including: constructing a constrained graph model based on time windows, vehicle information and real-time traffic API data; using an improved genetic algorithm for initial path planning; performing dynamic path adjustment based on the initial path set and real-time traffic API data; combining the optimized path set and customer preference information in the order information, screening the optimal solution through a plant rhizome growth optimization algorithm, and then outputting the final path optimization strategy according to a preset rule engine. The present invention uses a plant rhizome optimization algorithm to simulate the laws of biological growth to achieve multi-objective co-evolution, while avoiding premature convergence, taking into account the flexibility of customer preferences and task migration; through flexible adjustment of constraints and algorithm parameters, it can adapt to more scenarios, and can still maintain stable solution capabilities in the case of continuous congestion or a surge in orders, significantly improving the risk resistance of the logistics system.

Description

Technical Field

[0001] The present invention relates to the field of logistics distribution technology, and in particular to an intelligent logistics distribution path optimization method under multiple constraint conditions. Background Art

[0002] As a core link in modern commerce, logistics and distribution have consistently evolved in technology and industrial structure inextricably linked to the needs of economic and social development. From traditional manual dispatch to intelligent systems, logistics and distribution have undergone a remarkable transformation from decentralized operations to networked and digitalized operations. At the industrial level, the restructuring of global supply chains and the explosive growth of e-commerce are driving the evolution of delivery services towards high-frequency, small-volume, and time-sensitive operations. Processing millions of orders per day and fulfilling orders within minutes have become the norm, forcing the delivery system to upgrade to automation and intelligence. The complexity of end-to-end delivery scenarios has spurred the convergence of diverse technologies, including drones, unmanned vehicles, and smart express lockers. The global nature of cross-border logistics has driven breakthroughs in the application of blockchain technology in international freight forwarding and customs clearance. Currently, the logistics and distribution system is accelerating its transition to end-to-end digitization. Digital twin technology is building a virtual mapping system that enables real-time interaction between the physical distribution network and virtual optimization models, providing flexible solutions to address demand fluctuations, resource constraints, and unexpected risks.

[0003] Logistics and delivery route optimization is a core research area in modern supply chain management. Its essence is to achieve a comprehensive balance between cost, efficiency, and service quality through the scientific planning of transportation routes and resource scheduling. With the expansion of global trade and the rapid development of e-commerce, logistics networks have become multi-layered, multi-node, and dynamic. A single delivery task often involves multiple complex variables, such as vehicle scheduling, time window constraints, traffic response, and cargo loading restrictions.

[0004] Conventional logistics and distribution route optimization methods often employ a single model or static rules, resulting in frequent conflicts between constraints and difficulty generating feasible optimization solutions. Solutions based on single genetic algorithms or ant colony algorithms have significant shortcomings: genetic algorithms are prone to premature convergence and lack global search capabilities; ant colony algorithms experience delayed pheromone updates in dynamic traffic scenarios and struggle to adapt to sudden congestion. Traditional methods ignore customer preferences and optimize solely based on a single objective, such as time or cost, resulting in low customer satisfaction. Existing methods are prone to falling into local optimality in extreme scenarios such as surges in orders and persistent congestion, resulting in high solution failure rates and insufficient system stability.

[0005] In order to solve the defects in the above-mentioned prior art, this technical solution proposes an intelligent logistics distribution path optimization method under multiple constraints. Summary of the Invention

[0006] The present invention provides an intelligent logistics distribution path optimization method under multiple constraint conditions to address the defects in the prior art.

[0007] In one aspect, the present invention provides a method for optimizing intelligent logistics distribution paths under multiple constraints, comprising:

[0008] Determine vehicle information, time window, and starting point information based on logistics order information;

[0009] Build a constrained graph model based on time windows, vehicle information, and real-time traffic API data;

[0010] Combining the constrained graph model and the starting point information, an improved genetic algorithm is used to perform initial path planning and output the initial path set.

[0011] Based on the initial path set and real-time traffic API data, dynamic path adjustment is performed to output the optimized path set;

[0012] Combining the optimized path set and customer preference information in the order information, the optimal solution is screened through the plant root growth optimization algorithm, and the final path optimization strategy is output according to the preset rule engine.

[0013] According to the intelligent logistics distribution path optimization method under multiple constraints provided by the present invention, the steps of constructing a constraint graph model include:

[0014] According to the time window, define the hard time window violation penalty coefficient and output the time window constraint matrix;

[0015] Based on vehicle information, the vehicle capacity upper limit and path energy consumption model are constructed, and the vehicle capacity constraint table is output;

[0016] Build a road network map based on starting point information and real-time traffic API data;

[0017] Combining the time window constraint matrix, vehicle capacity constraint table and road network graph, a constraint graph model is constructed.

[0018] According to the intelligent logistics distribution path optimization method under multiple constraints provided by the present invention, the steps of using the improved genetic algorithm to perform initial path planning include:

[0019] Set up chromosome encoding structure; including vehicle ID and node sequence;

[0020] Based on the vehicle ID, a greedy algorithm is used to generate the initial path;

[0021] Predict travel time based on the historical average speed of the node sequence, combine vehicle ID and initial path, and output the initial chromosome encoding scheme;

[0022] According to the initial chromosome encoding scheme, penalty coefficient and vehicle capability constraint table, the fitness function of the improved genetic algorithm is constructed;

[0023] The population size of the improved genetic algorithm is set, and the initial path is optimized according to the initial chromosome encoding scheme and fitness function, and the initial path set is output.

[0024] According to the intelligent logistics distribution path optimization method under multiple constraints provided by the present invention, the steps of generating an initial path using a greedy algorithm include:

[0025] Based on the logistics order information, the time window urgency is calculated for each vehicle, and each vehicle is sorted according to the urgency, and a list of vehicles sorted by urgency is output;

[0026] Set the current position of each vehicle in the vehicle list as the starting point, clear the path list of each vehicle, complete the path initialization, and output the initialized vehicle status table;

[0027] For the current vehicle, traverse the initialized vehicle state table, filter vehicles that meet the preset constraints, and output the initialized vehicle path list; add the current vehicle to the initialized vehicle path list and output the initial path set.

[0028] According to the intelligent logistics distribution path optimization method under multiple constraints provided by the present invention, the fitness function formula is expressed as:

[0029]

[0030] Where Fit is the fitness function, ω1 is the time cost weight, ω2 is the penalty weight for exceeding the vehicle capability limit, and ω3 is the time window violation penalty weight; T represents the total driving time, OP is the penalty for exceeding the vehicle capability limit, and TWP represents the time window violation penalty.

[0031] According to the intelligent logistics distribution path optimization method under multiple constraints provided by the present invention, the calculation method of T, OP and TWP is expressed as follows:

[0032]

[0033]

[0034]

[0035] Where, d i,i+1 represents the actual road network distance from node i to node i+1, v i,i+1 represents the average travel speed from node i to node i+1, t wi represents the waiting time of the vehicle at node i, q total Indicates the total weight of the cargo currently loaded on the vehicle, Q indicates the maximum load of the vehicle, and C ol represents the overload penalty coefficient, e irepresents the lower limit of the time window required by customer i, l i represents the upper limit of the time window required by customer i, t i represents the actual time when the vehicle arrives at customer i, C early is the early arrival penalty coefficient, C late is the penalty coefficient for lateness.

[0036] According to the intelligent logistics distribution path optimization method under multiple constraints provided by the present invention, the steps of performing dynamic path adjustment include:

[0037] Based on real-time traffic API data, obtain the road congestion index and edge weight in the initial path set and output the latest road network weight;

[0038] Ant colony algorithm is used to re-plan the sub-paths with high weights in the latest road network and detect time window conflicts, and then a local optimized path set is output;

[0039] According to the total weight of the cargo currently loaded on the vehicle and the maximum load of the vehicle, the vehicle task migration is performed in combination with the local optimized path set, the nodes of the high-load vehicle are migrated to the low-load vehicle, the path sequence is updated, and the optimized path set is output.

[0040] According to the intelligent logistics distribution path optimization method under multiple constraints provided by the present invention, the step of replanning the sub-paths with high latest road network weights includes:

[0041] According to the constrained graph model, the pheromone matrix of the ant colony algorithm is initialized. The pheromone matrix includes the initial pheromone concentration of each road section and the heuristic value of each road section. The initial pheromone concentration is expressed as:

[0042]

[0043] Where, E ij represents the historical traffic efficiency of road section (i, j), represents the initial pheromone concentration; the heuristic value is expressed as:

[0044]

[0045] Where η ij represents the heuristic value, α and β represent the pheromone weight and heuristic weight, w ij represents the road congestion penalty weight, d ij is the road section distance, t ij is the historical average travel time;

[0046] According to the initial pheromone concentration and heuristic value, a path selection rule is formulated. The pheromone is updated after each vehicle passes through the road section, and the optimal path is output. The path selection rule is expressed as:

[0047]

[0048] Where, P ij represents the probability of transferring to road segment (i, j), c ij represents the time window urgency factor, c ik represents the time window urgency factor, t ik represents the pheromone concentration, or ik represents the heuristic value, k represents the set of allowed transfer paths of the current path, It means traversing all allowed transfer paths and calculating the comprehensive attraction of the road segment (i, k). represents the comprehensive attractiveness of road segment (i, j);

[0049] According to the optimal path, the pheromone is enhanced, and the pheromone of the continuously congested section is reset. The alternative path is re-explored, and the latest pheromone matrix and the global optimal path are output. The method of resetting the pheromone of the continuously congested section is expressed as:

[0050]

[0051] Where, τ ij2 represents the latest pheromone matrix, δ represents the reinforcement coefficient, Δτ elite is the real-time advantage of the global optimal path, θ ij Indicates vehicle load status;

[0052] If the global optimal path violates the vehicle capacity constraint table, the 2-opt algorithm is used to adjust the order of delivery points and recalculate pheromones.

[0053] According to the intelligent logistics distribution path optimization method under multiple constraints provided by the present invention, the method of updating pheromones is expressed as follows:

[0054]

[0055] Where G represents the quantitative feedback of path efficiency, ρ is the forgetting rate of path information, and τ ij0 represents the pheromone matrix before updating, τ ij1 represents the updated pheromone matrix, Dt priority Indicates the order priority gain factor.

[0056] According to the intelligent logistics distribution path optimization method under multiple constraints provided by the present invention, the step of selecting the optimal solution through the plant root growth optimization algorithm includes:

[0057] Based on the plant root growth optimization algorithm, an initial root population is generated, and each root system corresponds to a delivery path plan;

[0058] Randomly select a parent vehicle allocation path and perform taproot extension and fibrous root diffusion operations on the parent vehicle allocation path. The taproot extension operation is expressed as follows:

[0059]

[0060] Where Seed1 represents the updated delivery path of the x-th vehicle, represents the current delivery path of the x-th vehicle in the c-th iteration, α1 is a randomly generated real number in the range of (-0.5, 1.5), X best represents the current global optimal vehicle delivery path, represents the delivery path of another vehicle randomly selected from the current population; the root diffusion operation is represented as follows:

[0061]

[0062] Where Seed2 represents the path length of other vehicles in c+1 iterations; represents the average delivery path length of all vehicles in the cth iteration, α2 is a randomly generated real number in the range of (−1,1); represents the delivery path of a vehicle randomly selected from the current population;

[0063] According to the results of the taproot extension operation and the fibrous root diffusion operation, the feasible delivery path after the iteration of the selected parent vehicle allocation path is determined;

[0064] Randomly swap the order of two vehicles and calculate the feasible delivery path to guide the delivery paths of the other two selected vehicles;

[0065] Preference guided search is performed based on the guided delivery path, which is expressed as:

[0066]

[0067] Where, X lbest represents the local optimal path, Seed4 represents the global optimal path, and X r3 、X r4 represents two vehicle paths randomly selected from the current population, X d Indicates that the current vehicle allocation path guides the other two vehicles selected X r3 and X r4 Guided delivery path.

[0068] The intelligent logistics distribution path optimization method under multiple constraints provided by the present invention integrates time windows, vehicle loads, energy consumption, and real-time traffic data through a constrained graph model. Combining an improved genetic algorithm with dynamic ant colony optimization, it effectively addresses uncertainties such as order fluctuations and traffic congestion, reduces the probability of conflict between hard constraints in path planning, and improves the feasibility of the solution. By adopting a plant rhizome optimization algorithm to introduce taproot extension and fibrous root diffusion operations, it simulates biological growth patterns to achieve multi-objective co-evolution, while avoiding premature convergence while taking into account customer preferences and the flexibility of task migration. By relying on real-time traffic API data and the pheromone update mechanism of the ant colony algorithm, pheromone reset and path replanning are performed on congested sections of road. Furthermore, through a task migration strategy with vehicle load balancing, transportation efficiency is dynamically optimized, reducing idle driving rate and fuel consumption. Through quantitative indicators such as the time window urgency factor and the overload penalty coefficient, the cost of delays and the risk of violations are systematically reduced. At the same time, a preference-guided search mechanism is used to improve customer satisfaction, achieving multi-objective optimization of time, cost, and service quality. Through flexible adjustment of constraints and algorithm parameters, it can adapt to more scenarios and maintain stable solving capabilities in cases of continued congestion or a surge in orders, significantly improving the risk resistance of the logistics system. BRIEF DESCRIPTION OF THE DRAWINGS

[0069] In order to more clearly illustrate the technical solutions in the present invention or the prior art, a brief introduction is given below to the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0070] Figure 1 This is a step diagram of the intelligent logistics distribution path optimization method under multiple constraints provided by an embodiment of the present invention;

[0071] Figure 2 This is a diagram of the steps for screening the optimal solution through a plant rhizome growth optimization algorithm provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0072] To make the objectives, technical solutions, and advantages of the present invention more clear, the technical solutions of the present invention will be clearly and completely described below in conjunction with the accompanying drawings. Obviously, the embodiments described are only some of the embodiments of the present invention, not all of them. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts shall fall within the scope of protection of the present invention.

[0073] Example 1:

[0074] The following combination Figure 1-Figure 2The intelligent logistics distribution path optimization method under multiple constraints of the present invention is described.

[0075] like Figure 1-Figure 2 As shown, the intelligent logistics distribution path optimization method under multiple constraints provided by the embodiment of the present invention includes:

[0076] Determine vehicle information, time window, and starting point information based on the logistics order information. Logistics order information includes delivery demand data, namely cargo weight, volume, delivery time requirements (time window), starting point location, etc.

[0077] Based on the time window, vehicle information, and real-time traffic API data, a constrained graph model is constructed. The steps to construct a constrained graph model include:

[0078] Based on the time window, a hard time window violation penalty coefficient is defined and the time window constraint matrix is ​​output. The hard time window violation penalty coefficient is used to quantify the penalty parameter of the time window violation degree, including the penalty weights for early arrival and late arrival.

[0079] Based on vehicle information, the vehicle capacity upper limit and path energy consumption model are constructed, and the vehicle capacity constraint table is output.

[0080] Based on the origin information and real-time traffic API data, a road network diagram is constructed. The road network diagram consists of nodes (delivery points) and edges (road segments), and the weight of the edge is determined by the congestion index and the historical average travel time.

[0081] A constraint graph model is constructed by combining the time window constraint matrix, vehicle capacity constraint table, and road network diagram. The vehicle capacity limit refers to the maximum cargo weight and volume that a vehicle can carry in a single delivery. Exceeding this limit is considered overloading.

[0082] Combining the constraint graph model and the starting point information, the improved genetic algorithm is used to perform initial path planning and output the initial path set. The steps of using the improved genetic algorithm for initial path planning include:

[0083] Set the chromosome encoding structure, which is the encoding form of the routing plan, including vehicle ID, node sequence, and time allocation.

[0084] The initial path is generated using a greedy algorithm, that is, the initial path is generated by selecting the local optimal solution (such as the shortest distance or the lowest time) at each step.

[0085] The travel time is predicted based on the historical average speed of the node sequence, and the initial chromosome encoding scheme is output in combination with the vehicle ID and the initial path.

[0086] According to the initial chromosome encoding scheme, penalty coefficient and vehicle capability constraint table, the fitness function of the improved genetic algorithm is constructed. The formula of the fitness function is expressed as:

[0087]

[0088] Where Fit is the fitness function, ω1 is the time cost weight, ω2 is the penalty weight for exceeding the vehicle's capacity limit, and ω3 is the time window violation penalty weight. T represents the total travel time, i.e., the total time consumed by the route. OP is the penalty for exceeding the vehicle's capacity limit, i.e., the cost incurred due to overweight cargo. TWP represents the time window violation penalty, i.e., the cost incurred due to violating the time window requirement. T, OP, and TWP are calculated as follows:

[0089]

[0090]

[0091]

[0092] Where, d i,i+1 represents the actual road network distance from node i to node i+1, v i,i+1 represents the average travel speed from node i to node i+1, t wi represents the waiting time of the vehicle at node i, q total Indicates the total weight of the cargo currently loaded on the vehicle, Q indicates the maximum load of the vehicle, and C ol represents the overload penalty coefficient, e i represents the lower limit of the time window required by customer i, l i represents the upper limit of the time window required by customer i, t i represents the actual time when the vehicle arrives at customer i, C early is the early arrival penalty coefficient, C late is the penalty coefficient for lateness.

[0093] The population size of the improved genetic algorithm is set, and the initial path is optimized according to the initial chromosome encoding scheme and fitness function, and the initial feasible path set is output.

[0094] Based on the initial route set and real-time traffic API data, dynamic route adjustment is performed to output the optimized route set. The steps for dynamic route adjustment include:

[0095] According to the real-time traffic API data, the congestion index and edge weight of the road sections in the initial path set are obtained, and the latest road network weight is output.

[0096] The ant colony algorithm is used to re-plan the sub-paths with high weight in the latest road network and detect time window conflicts, and output a set of local optimized paths. The steps for re-planning the sub-paths with high weight in the latest road network include:

[0097] Based on the constrained graph model, the ant colony algorithm's pheromone matrix, or segment attractiveness, is initialized. This matrix includes the initial pheromone concentration for each segment and the heuristic value for each segment. The heuristic value reflects the inherent attractiveness of the segment and is inversely proportional to distance and time. The initial pheromone concentration is expressed as:

[0098]

[0099] Where, E ij represents the historical traffic efficiency of road section (i, j), represents the initial pheromone concentration, based on the historical traffic efficiency E ij Calculation. The heuristic value is expressed as:

[0100]

[0101] Where η ij represents the heuristic value, α and β represent the pheromone weight and heuristic weight, w ij represents the road congestion penalty weight, d ij is the road section distance, t ij is the historical average travel time.

[0102] Based on the initial pheromone concentration and heuristic value, a path selection rule is formulated. The pheromone is updated after each vehicle passes through the road section, and the optimal path is output. The path selection rule is expressed as:

[0103]

[0104]

[0105] Where, P ij represents the probability of transferring to road segment (i, j), c ij Indicates the time window urgency factor. The more urgent the time window (such as less remaining time), the ij The larger the value, the more the algorithm prefers this path. c ik Indicates the urgency of the time window, t ik represents the pheromone concentration, or ik represents the heuristic value, k represents the set of allowed transfer paths of the current path, It means traversing all allowed transfer paths and calculating the comprehensive attraction of the road segment (i, k). represents the comprehensive attractiveness of road segment (i, j), t now Indicates the current time. If the node has timed out ( t now >l i ), γ ij Drop to a negative value to force the path to be excluded.

[0106] The way to update pheromones is expressed as:

[0107]

[0108] Where G is the path efficiency, i.e. the increment of pheromone. ρ is the forgetting rate of path information, τ ij0 represents the pheromone matrix before updating, τ ij1 represents the updated pheromone matrix, Dt priority represents the order priority gain factor, expressed as:

[0109]

[0110] The pheromone concentration is dynamically adjusted by path efficiency and forgetting rate, and the paths of high-priority orders will obtain higher pheromone increments.

[0111] According to the optimal path, the pheromone is enhanced and the pheromone of the continuously congested section is reset. The alternative path is re-explored and the latest pheromone matrix and the global optimal path are output. The method of resetting the pheromone of the continuously congested section is expressed as:

[0112]

[0113]

[0114] Where, τ ij2 Represents the latest pheromone matrix, δ represents the reinforcement coefficient, which is used to accelerate the learning of sufficient paths. That is, when an order needs to be delivered first, increase γ to strengthen the pheromone of the order route. Δτ elite is the real-time advantage of the global optimal path (such as the road section after congestion is relieved). ij2 Below the threshold), reset its pheromone to the initial value, and then force the algorithm to re-explore alternative paths (such as detour backup routes). ij Indicates the vehicle load status. When the vehicle load is low, it is encouraged to use the route to avoid the risk of overloading. The lower the vehicle load (q total ≤0.8Q), θ ij The higher the load, the more likely it is that the path will be used to reserve space for loading. total >Q), θ ij Reduce to avoid repeated path selection.

[0115] If the global optimal path violates the vehicle capacity constraint table, the 2-opt algorithm is used to adjust the order of delivery points to ensure that the load and time windows are met, and the pheromone is recalculated.

[0116] According to the total weight of the cargo currently loaded on the vehicle and the maximum load of the vehicle, the vehicle task migration is performed in combination with the local optimized path set. The nodes of the high-load vehicle are migrated to the low-load vehicle to balance the load and avoid overloading. The path sequence is updated and the optimized path set is output.

[0117] Combining the optimized path set and customer preference information in the order information, the plant rhizome growth optimization algorithm is used to select the optimal solution, and then the final path optimization strategy is output according to the preset rule engine. The steps of selecting the optimal solution using the plant rhizome growth optimization algorithm include:

[0118] Based on a plant rhizome growth optimization algorithm, an initial root population is generated. Each root system corresponds to a delivery route solution. A greedy algorithm is used to generate high-cost, high-timeliness routes (e.g., electric vehicles are prioritized for cold chain orders). Low-cost, high-latency routes are randomly generated (e.g., fuel vehicles are routed around non-congested areas). The globally optimal route is prioritized based on the lowest total cost and no timeouts. If multiple solutions exist, the path that covers the most customer nodes is selected. Solutions of the same rank are weighted according to customer preference, prioritizing routes that meet cold chain orders (even if the total cost is slightly higher). Paths with a high probability of meeting the time window are compensated for low latency penalties.

[0119] Randomly select an incidental vehicle allocation path and perform taproot extension and fibrous root diffusion operations on the incidental vehicle allocation path. First, the taproot extension operation is expressed as follows:

[0120]

[0121] Where Seed1 represents the updated delivery path of the x-th vehicle. represents the current delivery path of the x-th vehicle in the c-th iteration. α1 is a randomly generated real number in the range of (-0.5, 1.5) and is used to control the search direction. If the value is greater than 0, it moves closer to the historical optimal path. If the value is less than 0, it explores a new area in the opposite direction. best It represents the current global optimal vehicle delivery path, which is used to guide all vehicles to learn from the optimal path and accelerate convergence. Represents the delivery path of another vehicle randomly selected from the current population, which is used to increase randomness and prevent all vehicles from falling into the same local optimal path.

[0122] The root system diffusion operation is expressed as follows:

[0123]

[0124] Where Seed2 represents the path length of other vehicles in c+1 iterations. represents the average delivery path length or time of all vehicles in the cth iteration. α2 is a randomly generated real number in the range (−1, 1) that controls the optimization direction: when the value is greater than 0, the path is shortened. When the value is less than 0, random perturbations are introduced. Represents the delivery path of a vehicle randomly selected from the current population, increasing diversity and simulating the efficiency differences of different vehicles in different areas.

[0125] Based on the results of the taproot extension operation and the fibrous root diffusion operation, the feasible delivery path after the selected vehicle allocation path iteration is determined. That is, boundary constraints are applied to limit the path search range to ensure the feasibility of the solution. It is expressed as follows:

[0126]

[0127] Where Seed3 represents the feasible delivery path after iteration c+1 of the selected vehicle allocation path. Represents the boundary of the problem space, such as the maximum distance or time limit of the delivery path, which is used to limit the search scope and ensure the feasibility of the solution. α3 is the boundary constraint strength and ranges from (0,1].

[0128] Randomly exchange the order of two vehicles (simulating the scattered growth of adventitious roots), and calculate the guided delivery paths of the other two vehicles selected under the guidance of the feasible delivery path, which can be expressed as:

[0129]

[0130] Where, X r3 、X r4 Represents two vehicle paths randomly selected from the current population to increase randomness and simulate the dispersed growth characteristics of the root system. d Indicates that the current vehicle allocation path guides the other two vehicles selected X r3 and X r4 The guided distribution path is α4, which is a randomly generated real number in the range of [0,1) and is used to control the diffusion intensity. When α4>0, limited diffusion is performed to high potential paths. When α4=0, random diffusion is performed.

[0131] Then, preference guided search is performed based on the guided delivery path, which is expressed as:

[0132]

[0133] Where, X lbest represents the local optimal path (such as the delivery plan with the highest customer satisfaction), and Seed4 represents the global optimal path.

[0134] In summary, the intelligent logistics distribution path optimization method under multiple constraints provided by the present invention integrates time windows, vehicle load, energy consumption and real-time traffic data through a constraint graph model, and combines an improved genetic algorithm with dynamic ant colony optimization to effectively deal with uncertainties such as order fluctuations and traffic congestion, reduce the probability of conflict of hard constraints in path planning, and improve the feasibility of the solution. By adopting the plant rhizome optimization algorithm to introduce taproot extension and fibrous root diffusion operations, the biological growth law is simulated to achieve multi-objective co-evolution, while avoiding premature convergence, taking into account customer preferences and the flexibility of task migration. By relying on real-time traffic API data and the pheromone update mechanism of the ant colony algorithm, pheromone reset and path replanning are performed on congested sections, and through the task migration strategy of vehicle load balancing, transportation efficiency is dynamically optimized to reduce empty driving rate and fuel consumption. Through quantitative indicators such as time window urgency factor and overload penalty coefficient, the delay cost and violation risk are systematically reduced, and the preference-guided search mechanism is used to improve customer satisfaction, achieving multi-objective optimization of time, cost and service quality. Through flexible adjustment of constraints and algorithm parameters, it can adapt to more scenarios and maintain stable solving capabilities in cases of continued congestion or a surge in orders, significantly improving the risk resistance of the logistics system.

[0135] Through the above description of the embodiments, those skilled in the art will clearly understand that each embodiment can be implemented using software plus a necessary general-purpose hardware platform, or of course, hardware. Based on this understanding, the essence of the above technical solution, or the portion that contributes to the prior art, can be embodied in the form of a software product. This computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, a magnetic disk, or an optical disk, and includes a number of instructions for causing a computer device (such as a personal computer, server, or network device) to execute the methods described in each embodiment or certain portions of the embodiments.

[0136] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the various embodiments of the present invention.

Claims

1. An intelligent logistics distribution path optimization method under multiple constraints, characterized by: include: Determine vehicle information, time window, and starting point information based on logistics order information; Constructing a constrained graph model based on the time window, the vehicle information, and the real-time traffic API data; the steps of constructing the constrained graph model include: According to the time window, a hard time window violation penalty coefficient is defined, and a time window constraint matrix is ​​output; Based on the vehicle information, construct a vehicle capacity upper limit and a path energy consumption model, and output a vehicle capacity constraint table; Constructing a road network map based on the starting point information and the real-time traffic API data; Combining the time window constraint matrix, the vehicle capability constraint table and the road network graph, constructing the constraint graph model; Combining the constrained graph model and the starting point information, an improved genetic algorithm is used to perform initial path planning, and an initial path set is output. The steps of performing initial path planning using the improved genetic algorithm include: Set up chromosome encoding structure; including vehicle ID and node sequence; Generate an initial path using a greedy algorithm based on the vehicle ID; Predicting travel time based on the historical average speed of the node sequence, combining the vehicle ID and the initial path, and outputting an initial chromosome encoding scheme; Constructing a fitness function of the improved genetic algorithm according to the initial chromosome encoding scheme, the penalty coefficient and the vehicle capability constraint table; Setting the population size of the improved genetic algorithm, optimizing the initial path according to the initial chromosome encoding scheme and the fitness function, and outputting the initial path set; Performing dynamic route adjustments based on the initial route set and the real-time traffic API data, and outputting an optimized route set; Combining the optimized path set and the customer preference information in the order information, screening the optimal solution through the plant rhizome growth optimization algorithm, and then outputting the final path optimization strategy according to the preset rule engine; The steps of screening the optimal solution through the plant root growth optimization algorithm include: Based on the plant root growth optimization algorithm, an initial root population is generated, and each root system corresponds to a delivery path plan; A parent vehicle allocation path is randomly selected, and a taproot extension operation and a fibrous root diffusion operation are performed on the parent vehicle allocation path; the taproot extension operation is represented as follows: ； Where Seed1 represents the updated delivery path of the x-th vehicle, represents the current delivery path of the x-th vehicle in the c-th iteration, α1 is a randomly generated real number in the range of (-0.5, 1.5), X best represents the current global optimal vehicle delivery path, represents the delivery path of another vehicle randomly selected from the current population; the root diffusion operation is represented as follows: ； Where Seed2 represents the path length of other vehicles in c+1 iterations; represents the average delivery path length of all vehicles in the cth iteration, α2 is a randomly generated real number in the range of (−1,1); represents the delivery path of a vehicle randomly selected from the current population; Determining a feasible delivery path after iteration of the selected parent vehicle allocation path based on the results of the taproot extension operation and the fibrous root diffusion operation; Randomly swap the order of two vehicles, and calculate the guided delivery paths of the other two selected vehicles guided by the feasible delivery path; According to the guided delivery path, a preference guided search is performed, which is expressed as: ； Where, X lbest represents the local optimal path, Seed4 represents the global optimal path, and X r3 、X r4 represents two vehicle paths randomly selected from the current population, X d Indicates that the current vehicle allocation path guides the other two vehicles selected X r3 and X r4 Guided delivery path.

2. The intelligent logistics distribution path optimization method under multiple constraints according to claim 1 is characterized in that: The steps to generate the initial path using the greedy algorithm include: Calculating the urgency of the time window for each vehicle based on the logistics order information, sorting each vehicle based on the urgency, and outputting a list of vehicles sorted by urgency; The current position of each vehicle in the vehicle list is set as the starting point, and the path list of each vehicle is cleared to complete the path initialization, and the initialized vehicle state table is output; For the current vehicle, the initialized vehicle state table is traversed, vehicles that meet preset constraints are screened, and an initialized vehicle path list is output; the current vehicle is added to the initialized vehicle path list, and the initial path set is output.

3. The intelligent logistics distribution path optimization method under multiple constraints according to claim 2 is characterized in that: The fitness function formula is expressed as: ； Where, Fit is the fitness function, ω1 is the time cost weight, ω2 is the penalty weight for exceeding the vehicle capability limit, and ω3 is the time window violation penalty weight; T represents the total driving time, OP is the penalty for exceeding the upper limit of the vehicle's capability, and TWP represents the time window violation penalty.

4. The intelligent logistics distribution path optimization method under multiple constraints according to claim 3 is characterized in that: The calculation method of T, OP and TWP is expressed as: ； ； ； Where, d i,i+1 represents the actual road network distance from node i to node i+1, v i,i+1 represents the average travel speed from node i to node i+1, t wi represents the waiting time of the vehicle at node i, q total Indicates the total weight of the cargo currently loaded on the vehicle, Q indicates the maximum load of the vehicle, and C ol represents the overload penalty coefficient, e i represents the lower limit of the time window required by customer i, l i represents the upper limit of the time window required by customer i, t i represents the actual time when the vehicle arrives at customer i, C early is the early arrival penalty coefficient, C late is the penalty coefficient for lateness.

5. The intelligent logistics distribution path optimization method under multiple constraints according to claim 2 is characterized in that: The steps for dynamic path adjustment include: According to the real-time traffic API data, obtain the road congestion index and edge weight in the initial path set, and output the latest road network weight; Using an ant colony algorithm to replan sub-paths with high values ​​of the latest road network weight and perform time window conflict detection, and output a local optimized path set; According to the total weight of the cargo currently loaded on the vehicle and the maximum load of the vehicle, the vehicle task migration is performed in combination with the local optimized path set, the nodes of the high-load vehicle are migrated to the low-load vehicle, the path sequence is updated, and the optimized path set is output.

6. The intelligent logistics distribution path optimization method under multiple constraints according to claim 5 is characterized in that: The step of replanning the sub-paths with high values ​​of the latest road network weights includes: According to the constrained graph model, the pheromone matrix of the ant colony algorithm is initialized. The pheromone matrix includes the initial pheromone concentration of each road section and the heuristic value of each road section. The initial pheromone concentration is expressed as: ； Where, E ij represents the historical traffic efficiency of road section (i, j), represents the initial pheromone concentration; the heuristic value is expressed as: ； Where η ij represents the heuristic value, α and β represent the pheromone weight and heuristic weight, w ij represents the road congestion penalty weight, d ij is the road section distance, t ij is the historical average travel time; A path selection rule is formulated based on the initial pheromone concentration and the heuristic value; and the pheromone is updated after each vehicle passes through the road section, and the optimal path is output; the path selection rule is expressed as: ； Where, P ij represents the probability of transferring to road segment (i, j), γ ij represents the time window urgency factor, γ ik represents the time window urgency factor, τ ik represents the pheromone concentration, η ik represents the heuristic value, k represents the set of allowed transfer paths of the current path, It means traversing all allowed transfer paths and calculating the comprehensive attraction of the road segment (i, k). represents the comprehensive attractiveness of road segment (i, j); According to the optimal path, the pheromone is enhanced, and the pheromone of the continuously congested road section is reset, and alternative paths are re-explored to output the latest pheromone matrix and the global optimal path; the method of resetting the pheromone of the continuously congested road section is expressed as follows: ； Where, τ ij2 represents the latest pheromone matrix, τ ij1 represents the updated pheromone matrix, δ represents the reinforcement coefficient, Δτ elite is the real-time advantage of the global optimal path, θ ij Indicates vehicle load status; If the global optimal path violates the vehicle capacity constraint table, the 2-opt algorithm is used to adjust the order of the delivery points and recalculate the pheromone.

7. The intelligent logistics distribution path optimization method under multiple constraints according to claim 6 is characterized in that: The method of updating the pheromone is expressed as: ； Where G represents the quantitative feedback of path efficiency, ρ is the forgetting rate of path information, and τ ij0 represents the pheromone matrix before updating, τ ij1 represents the updated pheromone matrix, Δτ priority Indicates the order priority gain factor.

Images