Cold-chain logistics vehicle planning system

Abstract

The invention discloses a cold-chain logistics vehicle planning system. The system comprises the following steps of initial population creation, explosion and spark generation, evaluation and selection, genetic manipulation, pheromone updating, termination judgment and optimal solution output. According to the method, a dynamic pheromone updating and path reward and punishment mechanism is introduced through improvement of pheromone updating, so that the convergence speed of the algorithm is accelerated, and the quality of a solution is improved; the exploration depth of a solution space is increased by improving the crossover operation and considering the length and quality of an introduced path; improvement of mutation operation depends on a self-adaptive mutation strategy, the exploration capability of the algorithm is enhanced, multi-objective optimization is introduced into path planning through the improved firework algorithm to consider transportation cost, temperature control and energy consumption management at the same time, and a dynamic adjustment mechanism is combined to cope with frequently changing terrain and climate conditions in the karst landform, so that the development of the karst landform is facilitated. Therefore, the adaptability of the algorithm in a complex environment is improved.

Description

Technical Field

[0001] The present invention relates to the technical field of cold chain logistics, and in particular to a cold chain logistics vehicle planning system. Background Art

[0002] Karst landform is a special landform formed by the dissolution of carbonate rocks. It has the characteristics of complex terrain, diverse landforms, developed caves, and hidden water systems. Karst landforms are distributed in Yunnan, Guizhou, Guangxi and other regions in my country. These regions are also important production and consumption areas of fresh agricultural products in my country. Therefore, cold chain logistics has important significance and value in these regions. However, karst landform also brings certain difficulties and challenges to the transportation route planning of cold chain logistics, such as tortuous roads, steep slopes, poor road conditions, traffic congestion, etc. These factors will affect the vehicle's driving speed, fuel consumption, etc. This affects transportation costs and cargo quality. At the same time, the ecological environment in the karst area is fragile, and excessive logistics activities may aggravate environmental damage. Conventional algorithms often ignore the need for ecological protection and find it difficult to balance the contradiction between efficiency and environmental protection. Cold chain logistics also involves the complexity of multi-objective optimization, including transportation time, temperature control, and energy consumption management. Rough roads and steep slopes increase vehicle fuel consumption, and refrigeration equipment consumes more energy to maintain a stable temperature, which increases overall energy consumption and maintenance costs. Conventional algorithms lack sufficient emergency response capabilities in the event of sudden disasters and cannot be adjusted quickly, affecting transportation safety. Summary of the invention

[0003] The purpose of the present invention is to provide a cold chain logistics vehicle planning system to solve the problems raised in the above background technology.

[0004] To achieve the above object, the present invention provides the following technical solution: a cold chain logistics vehicle planning system, comprising the following steps:

[0005] S1. Create an initial population, each individual represents a potential transportation path solution. These individuals are randomly placed in the search space, and their fitness is evaluated based on the total cost and efficiency of the path solution they represent;

[0006] Due to the complex terrain in the karst region, the driving speed and fuel consumption of vehicles are affected to varying degrees. Therefore, the cold chain transportation route optimization problem is subject to vehicle capacity constraints, route distance constraints, route cost constraints, cargo temperature constraints, customer time window constraints and customer service times constraints. Multiple constraints are analyzed and calculated.

[0007] The goal of cold chain transportation route optimization is to minimize the total transportation cost, which is composed of multiple aspects, including fixed cost, transportation cost, cargo damage and deterioration cost, and time window penalty cost. Multiple aspects of cost are analyzed and calculated to establish the objective function.

[0008] In karst areas, due to the complexity of the terrain, the driving speed and fuel consumption of vehicles are affected to varying degrees. Therefore, in the calculation of transportation costs and cargo damage and deterioration costs, terrain factors are introduced to consider the impact of terrain on transportation costs and cargo quality, providing a more reasonable and efficient transportation route solution for cold chain logistics in karst areas. Specifically, the terrain factor is defined as:

[0009]

[0010] where f ij represents the terrain factor between the i-th customer and the j-th customer, v ij represents the actual driving speed between the i-th customer and the j-th customer, v0 represents the ideal driving speed on flat terrain, and e ij represents the actual fuel consumption between the i-th customer and the j-th customer, e0 represents the ideal fuel consumption on flat terrain, and r ij represents the actual cargo spoilage rate between the i-th customer and the j-th customer, r0 represents the ideal cargo spoilage rate on flat terrain, and s ij is the actual slope impact between the i-th customer and the j-th customer, s0 is the road slope impact on flat terrain, the larger the terrain factor, the greater the impact of the terrain on the cost, and vice versa;

[0011] Due to the complex terrain in karst areas, the vehicle's driving speed and fuel consumption are affected to varying degrees. In the route distance constraint, route cost constraint, cargo temperature constraint and customer time window constraint, terrain factors are introduced to reflect the impact of terrain on the constraints.

[0012] S2, based on the fireworks algorithm, simulates the phenomenon of fireworks explosion in nature to solve the optimization problem, where each firework individual generates new sparks according to its fitness, and these sparks explore the solution space around the original individual to find potential improvement opportunities;

[0013] During the operation of the algorithm, each firework represents a candidate solution, and the sparks produced by the explosion of the firework represent new candidate solutions. These sparks propagate in the solution space according to their fitness values, thereby exploring new possible solutions. The fireworks algorithm can be used to solve optimization problems of the following form:

[0014] minf(x)stx∈Ω

[0015] Where x=(x1,x2,...,xd ) is the decision variable vector, f(x) is the objective function, Ω is the feasible domain, and d is the dimension of the problem;

[0016] A certain number of fireworks are randomly generated in the feasible domain Ω, each firework represents a candidate solution, and the fitness value of each firework, that is, the objective function value, is calculated;

[0017] According to the fitness value of each firework, the number of sparks and explosion radius generated by each firework are determined, which can ensure both sufficient global search capability and good local search capability;

[0018] During the spark generation process, the quality of each potential solution is evaluated by a heuristic factor, which can be a comprehensive score based on factors such as the route's historical transportation efficiency, cost, and safety record;

[0019] Calculation of heuristic factor,The calculation of heuristic factor can be performed by the following formula:

[0020] H(i)=ω1E(i)+ω2C(i)+ω3S(i)

[0021] Where H(i) is the heuristic factor of the i-th solution, E(i) is the efficiency score, C(i) is the cost score, S(i) is the safety score, ω1, ω2, ω3 are the corresponding weight coefficients, which can be adjusted according to actual conditions;

[0022]

[0023]

[0024]

[0025] Solutions with higher heuristic factors will have a greater probability of generating more sparks and thus be explored more in subsequent searches. The relationship between the number of sparks and the heuristic factor can be expressed as:

[0026] N sparks (i) = N base +α·H(i)

[0027] Dynamically adjust the heuristic factor. In order to adapt to the changing environment and data, the heuristic factor needs to be dynamically adjusted:

[0028] H(i,t+1)=H(i,t)·(1-β·ΔR(i))

[0029] If a route becomes infeasible due to an unexpected event, the heuristic factor of the associated solution will be reduced, and vice versa;

[0030] S3. Newly generated sparks will be evaluated according to their fitness, and the next generation population will be determined based on the fitness value. The best individuals are selected to maintain the progress and effectiveness of the algorithm.

[0031] For each firework, according to its number of sparks and explosion radius, a corresponding number of sparks are randomly generated around it. Each spark also represents a candidate solution. The fitness value of each spark is calculated. The spark generation formula is as follows:

[0032]

[0033] in, is the kth spark of the ith firework, x i is the i-th firework, A i is the explosion radius of the ith firework, is a random vector, each component of which follows a uniform distribution between -1 and 1.

[0034] In order to increase the diversity of the population, for each firework, a certain number of dimensions are randomly selected to perform Gaussian mutation operations to generate Gaussian sparks. Each Gaussian spark also represents a candidate solution. The fitness value of each Gaussian spark is calculated. The generation formula of Gaussian sparks is as follows:

[0035]

[0036] in, is the Gaussian spark of the ith firework, x i is the ith firework, σ is a constant used to control the intensity of Gaussian variation, g i is a random vector whose components follow a standard normal distribution;

[0037] Among all fireworks, sparks and Gaussian sparks, select the one with the best fitness value as the current optimal solution and keep it for the next generation. Then, according to the distance from each candidate solution to other candidate solutions, select a certain number of candidate solutions as the fireworks for the next generation.

[0038] In order to maintain the diversity of the population, not only the fitness value of the individual is considered, but also the differences between individuals. Even if the fitness value of an individual is not the highest, but it can bring new genetic information to the population, it may also be selected to increase the genetic diversity of the population and avoid premature convergence. The improved formula is:

[0039]

[0040] f(i) is the fitness value of the individual, δ(i) is the genetic diversity contribution value of the individual, and N is the size of the population;

[0041] Regardless of the diversity of the population, the elite retention strategy retains a certain proportion of the best individuals to ensure that the algorithm can continue to evolve towards excellent solutions. This method can ensure that the best individuals will not be lost during the evolution process, thereby improving the overall performance of the algorithm. The specific formula is:

[0042]

[0043] The roulette selection mechanism is adopted to allocate the selection probability according to the individual fitness value, and introduce random factors to increase the randomness of the selection process. This can introduce more uncertainty in the selection process and avoid falling into the local optimal solution, thereby improving the global search ability. The selection probability of the roulette selection is calculated as follows:

[0044]

[0045] Introduce a competitive selection mechanism to compare the fitness values ​​of two randomly selected individuals and select the individual with a higher fitness value;

[0046] S4. The selected individuals will undergo crossover and mutation operations. The crossover operation generates offspring individuals by combining the characteristics of the parent individuals, while the mutation operation randomly changes certain characteristics of the individuals to introduce new genetic information. The crossover and mutation operations work together on the population to increase its diversity and adaptability.

[0047] S5. The update of pheromones is based on individual performance. Excellent path solutions will increase pheromones, which will be explored more in subsequent searches, helping the algorithm remember and utilize past successful experiences.

[0048] The update of pheromone is no longer fixed, but dynamically adjusted according to the search state of the algorithm. The dynamic adjustment of pheromone volatilization rate can be expressed as:

[0049]

[0050] ρ(t) represents the pheromone volatilization rate at time t, ρ0 is the initial volatilization rate H current and H max Represent the current solution set entropy value and maximum entropy value respectively;

[0051] Path reward and punishment mechanism: After each iteration, all generated paths are rewarded or punished according to their quality. High-quality paths will gain additional pheromones, increasing their probability of being selected in subsequent iterations; while poor-quality paths will have their pheromones reduced, reducing their probability of being selected.

[0052]

[0053] The pheromone update formula, the improved pheromone update formula is as follows:

[0054] τ ij (t+1)=(1-ρ)·τ ij (t)+Δτ ij (t)

[0055] where τ ij is the pheromone concentration on path (i, j) at time t, ρ is the pheromone volatility, Δτ ij (t) is the amount of pheromone added on the path (i, j) at time t, which is determined by the number of ants passing through the path and the path quality;

[0056] When the algorithm shows signs of convergence, the pheromone volatilization rate will increase to prevent premature convergence; when the algorithm falls into a local optimum, the pheromone volatilization rate will decrease to increase the exploration of the solution space;

[0057] S6, termination judgment, check whether the algorithm reaches the preset threshold of fitness value change. If the termination condition is met, the algorithm will stop, which ensures that the algorithm will not run indefinitely. If the termination condition is not met, return to S2 and continue iteration;

[0058] Specifically, first, define the convergence standard and set a fitness threshold;

[0059] Secondly, record the fitness value. At the end of each iteration, record the current fitness value current_fitness and save it in a list fitness_history;

[0060] Finally, check whether the fitness value has converged. Traverse fitness_history and check whether the fitness value change in the most recent n_consecutive_iterations iterations is less than fitness_threshold. If the fitness value change in consecutive n_consecutive_iterations iterations is less than fitness_threshold, the algorithm is considered to have converged.

[0061] S7, output the optimal solution. Finally, after the algorithm is finished, the individual with the highest fitness is selected from the current population, and the transportation path solution represented by the individual is output as the optimal solution, ensuring that the result of the algorithm is a practical and valuable solution that can be used by the end user;

[0062] Furthermore, in step S4, the crossover operation is improved as follows: two or more firework individuals are selected as parents, and some of their features are exchanged according to certain rules to generate new firework individuals as children:

[0063] O child=α·P1+(1-α)·P2

[0064] Among them, P1 and P2 are the characteristics of the two parent individuals, α is the crossover coefficient, which is used to control the weight of the parent characteristics. The formula generates offspring by weighted combination of the parent characteristics to maintain the diversity of solutions;

[0065] The selection of the crossover point can be random or based on some heuristic rules, and the crossover operation can be single-point crossover, multi-point crossover or uniform crossover;

[0066] The crossover probability determines the probability of the parent individual performing a crossover operation, which affects the algorithm's exploration ability. Usually, the crossover probability P crossover Set to a moderate value to ensure that the algorithm can maintain diversity and converge effectively;

[0067]

[0068] N crossover is the number of individuals to be crossovered, N total is the total population.

[0069] In order to adapt to the complex path optimization problem, the path length and quality are considered in the crossover operation. When performing the crossover operation, individuals with shorter path length and higher quality are preferentially selected as parents, in the hope that the offspring can inherit better characteristics. The parent fitness value F parent The calculation formula is:

[0070]

[0071] L path is the path length, C path is the comprehensive cost of the path, γ is the trade-off coefficient between path length and cost;

[0072] Compare the fitness of individuals before and after crossover to evaluate the effect of the crossover operation. If the crossover operation can produce offspring with higher fitness, the crossover operation is successful.

[0073] The crossover operation can guide the algorithm to an excellent solution while maintaining the diversity of the population;

[0074] Further, in step S4, the formula of the mutation operation is as follows:

[0075]

[0076] O mutated (k) represents the individual characteristics after mutation, P(k) is the characteristic of the original individual, σ is the variation amplitude, N(0, 1) is the random noise introduced by the standard normal distribution, P mutation is the mutation probability;

[0077] The mutation probability determines the probability of an individual performing a mutation operation, which affects the diversity and innovation ability of the algorithm. Usually, the mutation probability is set to a lower value to ensure the stability of the algorithm and introduce enough new genetic information;

[0078] In order to adapt to the complex path optimization problem, this study introduced an adaptive mutation strategy based on path quality in the mutation operation. When performing the mutation operation, individuals with poor path quality are preferentially selected for mutation in the hope of generating better solutions through mutation. The path quality Q path The evaluation formula for (i) is:

[0079]

[0080] Compare the fitness of individuals before and after mutation to evaluate the effect of the mutation operation. If the mutation operation can produce individuals with higher fitness, then the mutation operation is considered successful.

[0081] Improvements in mutation operations With the help of adaptive mutation strategies, mutation operations can guide the algorithm to explore new solution spaces while maintaining population diversity.

[0082] Compared with the prior art, the beneficial effects of the present invention are as follows: the improvement of the heuristic factor enables the fireworks algorithm to better utilize historical data and expert knowledge, thereby guiding the search process to move toward more promising areas; secondly, the improvement of the pheromone update introduces a dynamic pheromone update and a path reward and punishment mechanism, which not only accelerates the convergence speed of the algorithm, but also improves the quality of the solution; then, the improvement of the selection operation adopts the introduction of diversity preservation and elite retention strategies to ensure the diversity of the population and the retention of excellent individuals; in addition, the improvement of the crossover operation uses the introduction of path length and quality considerations to increase the exploration depth of the solution space; finally, the improvement of the mutation operation relies on an adaptive mutation strategy to enhance the exploration ability of the algorithm and prevent the algorithm from The improved Fireworks algorithm introduces multi-objective optimization in path planning to simultaneously consider transportation cost, temperature control and energy consumption management, and combines it with a dynamic adjustment mechanism to cope with the frequently changing terrain and climate conditions in karst landforms, thereby improving the adaptability of the algorithm in complex environments; and the improved Fireworks algorithm shows high applicability in solving the cold chain logistics path optimization problem. In the event of sudden disasters, the improvement of heuristic factors enables the Fireworks algorithm to use historical data with sufficient emergency response capabilities, and can quickly optimize vehicle scheduling and path planning, ensure the safety of cold chain logistics transportation, improve transportation efficiency, reduce operating costs, and improve cargo quality and customer satisfaction. BRIEF DESCRIPTION OF THE DRAWINGS

[0083] Figure 1 is a system flow chart of the present invention;

[0084] Figure 2 Schematic diagram of the crossover operation of the present invention. DETAILED DESCRIPTION

[0085] The technical solution of the present invention is described in detail below through specific embodiments.

[0086] See also Figure 1-2 The present invention provides a cold chain logistics vehicle planning system, comprising the following steps:

[0087] S1. Create an initial population, each individual represents a potential transportation path solution. These individuals are randomly placed in the search space, and their fitness is evaluated based on the total cost and efficiency of the path solution they represent;

[0088] Transportation route plan: First, due to the complex terrain of the karst region, the vehicle's driving speed and fuel consumption are affected to varying degrees. Therefore, the cold chain transportation route optimization problem is subject to vehicle capacity constraints, route distance constraints, route cost constraints, cargo temperature constraints, customer time window constraints, and customer service times constraints. Multiple constraints are analyzed and calculated;

[0089] Vehicle capacity constraints: the load of each vehicle cannot exceed its maximum capacity, that is:

[0090]

[0091] Among them, q i represents the demand of the ith customer, x ik Indicates whether the k-th vehicle serves the i-th customer, Qk represents the maximum capacity of the k-th vehicle, n represents the number of customers, and m represents the number of vehicles;

[0092] Cargo temperature constraints: the cargo temperature of each vehicle must be within a certain range, namely:

[0093] T min ≤T ik ≤T max ,i=1,2,...,n,k=1,2,...,m

[0094] Among them, T ik represents the cargo temperature of the kth vehicle when serving the ith customer, T min and T max Indicate the minimum and maximum temperatures of the cargo;

[0095] Customer time window constraint: Each vehicle must arrive and complete service within the customer’s service time window, that is:

[0096] ai ≤t ik ≤ b i ,i=1,2,...,n,k=1,2,...,m

[0097] where t ik represents the arrival time of the kth vehicle when serving the ith customer, a i and b i represents the start and end time of the service time window of the i-th customer;

[0098] Route distance constraint: the driving distance of each vehicle cannot exceed its maximum driving distance, that is:

[0099]

[0100] Among them, d ij represents the distance between the i-th customer and the j-th customer, D k represents the maximum driving distance of the kth vehicle;

[0101] Route cost constraint: the travel cost of each vehicle cannot exceed its maximum travel cost, that is:

[0102]

[0103] Among them, c ij represents the cost between the i-th customer and the j-th customer, C k represents the maximum travel cost of the kth vehicle;

[0104] Route connectivity constraints: the driving route of each vehicle must be connected, that is:

[0105]

[0106] Among them, x ik and x jk Indicates whether the k-th vehicle serves the i-th customer and the j-th customer;

[0107] Route integrity constraint: Each vehicle’s route must start from the distribution center and return, that is:

[0108] x 0k =x k0 =1, k=1, 2, ..., m

[0109] Among them, x 0k and x k0 Indicates whether the kth vehicle departs from and returns to the distribution center, 0 indicates the distribution center;

[0110] Customer service frequency constraint: each customer can only be served once by a car, that is:

[0111]

[0112] Among them, x ik Indicates whether the k-th vehicle serves the i-th customer;

[0113] The goal of cold chain transportation route optimization is to minimize the total transportation cost, which is composed of multiple aspects, including fixed cost, transportation cost, cargo damage and deterioration cost, and time window penalty cost. Multiple aspects of cost are analyzed and calculated to establish the objective function.

[0114] Fixed costs refer to the costs of purchasing, maintaining, and depreciating cold chain transport vehicles that do not change with the volume of transport. They are related to the number and type of vehicles and can be calculated using the following formula:

[0115]

[0116] Among them, H1 represents fixed cost, x 0k The decision variable indicating whether the kth vehicle departs from the distribution center, γ k represents the unit fixed cost of the kth vehicle;

[0117] The transportation cost is the fuel consumption, tolls, labor costs, etc. of cold chain transportation vehicles that vary with the transportation volume. The transportation cost is related to factors such as vehicle load, distance, and fees, and can be calculated using the following formula:

[0118]

[0119] Among them, H2 represents the transportation cost, y ijk The decision variable indicating whether the k-th vehicle reaches the j-th customer from the i-th customer, δ k represents the fuel consumption coefficient per unit load of the kth vehicle, q i represents the demand of the i-th customer, d ij represents the distance between the i-th customer and the j-th customer, c ij represents the cost between the i-th customer and the j-th customer;

[0120] The cost of cargo damage and deterioration is the cost of cargo quality degradation or deterioration caused by temperature changes of cargo during cold chain transportation or loss of cold air during unloading. The cost of cargo damage and deterioration is related to factors such as the initial quality of the cargo, the corruption rate, and the unloading time. It can be calculated using the following formula:

[0121]

[0122] Among them, H3 represents the cost of goods damage and deterioration, x ik The decision variable indicating whether the k-th vehicle serves the i-th customer, ∈i represents the unit price of the goods of the i-th customer, q i represents the demand of the i-th customer, λ k

[0123] represents the transportation corruption rate of the cargo of the kth vehicle, t ik represents the travel time of the k-th vehicle when serving the i-th customer, μk represents the unloading corruption rate of the cargo of the k-th vehicle, s i represents the unloading time of the cargo of the i-th customer.

[0124] The time window penalty cost is the cost of customer dissatisfaction or loss caused by the vehicle's arrival time not meeting the customer's service time window. The time window penalty cost is related to factors such as the vehicle's arrival time, the customer's service time window, and the penalty coefficient, and can be calculated using the following formula:

[0125]

[0126] Among them, H4 represents the time window penalty cost, x ik The decision variable indicating whether the k-th vehicle serves the i-th customer, t ik represents the arrival time of the kth vehicle when serving the ith customer, a i and b i represents the start and end time of the service time window of the i-th customer, ρ i and σ i represents the penalty coefficient for early arrival and late arrival of the i-th customer;

[0127] Combining the costs of all the above aspects, the formula for the total transportation cost can be obtained as follows:

[0128] H=H1+H2+H3+H4

[0129] Among them, H represents the total transportation cost, H1 represents the fixed cost, H2 represents the transportation cost, H3 represents the cost of cargo damage and deterioration, and H4 represents the time window penalty cost;

[0130] In karst areas, due to the complexity of the terrain, the driving speed and fuel consumption of vehicles are affected to varying degrees. Therefore, in the calculation of transportation costs and cargo damage and deterioration costs, terrain factors are introduced to consider the impact of terrain on transportation costs and cargo quality, providing a more reasonable and efficient transportation route solution for cold chain logistics in karst areas. Specifically, the terrain factor is defined as:

[0131]

[0132] where f ij represents the terrain factor between the i-th customer and the j-th customer, vij represents the actual driving speed between the i-th customer and the j-th customer, v0 represents the ideal driving speed on flat terrain, and e ij represents the actual fuel consumption between the i-th customer and the j-th customer, e0 represents the ideal fuel consumption on flat terrain, and r ij represents the actual cargo spoilage rate between the i-th customer and the j-th customer, r0 represents the ideal cargo spoilage rate on flat terrain, and s ij is the actual slope impact between the i-th customer and the j-th customer, s0 is the road slope impact on flat terrain, the larger the terrain factor, the greater the impact of the terrain on the cost, and vice versa, so the transportation time, energy consumption model and temperature control calculation formula are modified:

[0133]

[0134] Among them, t ij is the transportation time, v0·f ij is the actual speed after terrain correction, t delay It is an additional delay time to reflect the traffic conditions, sharp bends or steep slopes that may exist in karst areas;

[0135] E ij =γ k ·q i ·d ij ·s ij ·f ij

[0136] Among them, E ij is the energy consumption model, γ k Indicates the energy consumption coefficient per unit load of the vehicle, indicates the slope factor, s ij Reflects the impact of slope on energy consumption, f ij is the terrain factor, which comprehensively reflects the impact of complex terrain on vehicle energy consumption;

[0137] T j (t) = T0 + α·t·f ij -β·d ij ·f ij

[0138] Among them, T j (t) is temperature control, α is the time coefficient of cargo temperature rise, reflecting the impact of external temperature and terrain on temperature rise, and β represents the cooling capacity of vehicle cold chain equipment, which is affected by terrain and transportation distance;

[0139] The calculation formula for transportation cost, cargo damage and deterioration cost, and time window penalty cost is modified to:

[0140]

[0141]

[0142]

[0143] Among them, p j represents the unit price of goods for the jth customer, f ij It is the terrain factor, which reflects the impact of complex terrain on the transportation path cost.

[0144] The final multi-objective optimization objective function is:

[0145]

[0146] λ1, λ2, and λ3 are the weight coefficients of energy consumption, temperature control, and time, respectively;

[0147] Due to the complex terrain in the karst area, the driving speed and fuel consumption of vehicles are affected to varying degrees. In the route distance constraint, route cost constraint, cargo temperature constraint and customer time window constraint, terrain factors are introduced to reflect the impact of terrain on the constraints. Specifically, the constraints of route distance constraint, transportation time constraint, cargo temperature constraint and customer time window constraint are modified as follows:

[0148]

[0149]

[0150] T min ≤T j (t) = T0 + α·t·f ij -β·d ij ·f ij ≤T max

[0151]

[0152] Among them, α and β are two constants, representing the sensitivity of cargo temperature and vehicle arrival time to terrain factors, respectively;

[0153] The basic model was modified and supplemented in combination with terrain factors, and a cold chain logistics transportation path optimization model suitable for karst landforms was constructed, providing a more reasonable and efficient transportation path solution for karst landforms.

[0154] S2, based on the fireworks algorithm, simulates the phenomenon of fireworks explosion in nature to solve the optimization problem, where each firework individual generates new sparks according to its fitness, and these sparks explore the solution space around the original individual to find potential improvement opportunities;

[0155] During the operation of the algorithm, each firework represents a candidate solution, and the sparks produced by the explosion of the firework represent new candidate solutions. These sparks propagate in the solution space according to their fitness values, thereby exploring new possible solutions. The fireworks algorithm can be used to solve optimization problems of the following form:

[0156] minf(x)stx∈Ω

[0157] Where x=(x1,x2,...,x d ) is the decision variable vector, f(x) is the objective function, Ω is the feasible domain, and d is the dimension of the problem;

[0158] A certain number of fireworks are randomly generated in the feasible domain Ω, each firework represents a candidate solution, and the fitness value of each firework, that is, the objective function value, is calculated;

[0159] According to the fitness value of each firework, the number of sparks and explosion radius generated by each firework are determined to ensure sufficient global search capability and good local search capability. The calculation formulas for the number of sparks and explosion radius are as follows:

[0160]

[0161]

[0162] Among them, S i is the number of sparks produced by the ith firework, A i is the explosion radius of the ith firework, M and a m are two constants used to control the total number of sparks and the explosion range. is the maximum value of the objective function in the current fireworks population, that is, the worst value, is the minimum value of the objective function in the current fireworks population, that is, the optimal value, f(x i ) is the objective function value of the ith firework, N is the number of fireworks, and ∈ is a small positive number to avoid division by zero errors;

[0163] In order to avoid the number of sparks and explosion radius being too large or too small, S i and A i Restrictions are made, namely:

[0164]

[0165]

[0166] Among them, a and b are two constants, satisfying 0<a<b<1, and Respectively represent floor and ceiling rounding, a m and b mare two constants, satisfying 0<a m <b m ;

[0167] In the spark generation process, heuristic factors are used to evaluate the quality of each potential solution. The heuristic factor can be a comprehensive score based on the route's historical transportation efficiency, cost, safety record, etc.

[0168] Calculation of heuristic factor,The calculation of heuristic factor can be performed by the following formula:

[0169] H(i)=ω1E(i)+ω2C(i)+ω3S(i)

[0170] Where H(i) is the heuristic factor of the i-th solution, E(i) is the efficiency score, C(i) is the cost score, S(i) is the safety score, ω1, ω2, ω3 are the corresponding weight coefficients, which can be adjusted according to actual conditions;

[0171]

[0172]

[0173]

[0174] Solutions with higher heuristic factors will have a greater probability of generating more sparks and thus be explored more in subsequent searches. The relationship between the number of sparks and the heuristic factor can be expressed as:

[0175] N sparks (i) = N base +α·H(i)

[0176] Dynamically adjust the heuristic factor. In order to adapt to the changing environment and data, the heuristic factor needs to be dynamically adjusted:

[0177] H(i,t+1)=H(i,t)·(1-β·ΔR(i))

[0178] If a route becomes infeasible due to an unexpected event, the heuristic factor of the associated solution will be reduced, and vice versa;

[0179] The heuristic factor will make the Fireworks algorithm more efficient and accurate in solving the cold chain logistics transportation path optimization problem of H enterprise, especially in complex terrain conditions such as karst landforms;

[0180] S3. Newly generated sparks will be evaluated according to their fitness, and the next generation population will be determined based on the fitness value. The best individuals are selected to maintain the progress and effectiveness of the algorithm.

[0181] For each firework, according to its number of sparks and explosion radius, a corresponding number of sparks are randomly generated around it. Each spark also represents a candidate solution. The fitness value of each spark is calculated. The spark generation formula is as follows:

[0182]

[0183] in, is the kth spark of the ith firework, x i is the i-th firework, A i is the explosion radius of the ith firework, is a random vector, each component of which follows a uniform distribution between -1 and 1.

[0184] In order to increase the diversity of the population, for each firework, a certain number of dimensions are randomly selected to perform Gaussian mutation operations to generate Gaussian sparks. Each Gaussian spark also represents a candidate solution. The fitness value of each Gaussian spark is calculated. The generation formula of Gaussian sparks is as follows:

[0185]

[0186] in, is the Gaussian spark of the ith firework, x i is the ith firework, σ is a constant used to control the intensity of Gaussian variation, g i is a random vector whose components follow a standard normal distribution;

[0187] Among all fireworks, sparks and Gaussian sparks, select the one with the best fitness value as the current optimal solution and keep it for the next generation. Then, according to the distance from each candidate solution to other candidate solutions, select a certain number of candidate solutions as the fireworks for the next generation.

[0188] In order to maintain the diversity of the population, not only the fitness value of the individual is considered, but also the differences between individuals. Even if the fitness value of an individual is not the highest, but it can bring new genetic information to the population, it may also be selected to increase the genetic diversity of the population and avoid premature convergence. The improved formula is:

[0189]

[0190] f(i) is the fitness value of the individual, δ(i) is the genetic diversity contribution value of the individual, and N is the size of the population;

[0191] Regardless of the diversity of the population, the elite retention strategy retains a certain proportion of the best individuals to ensure that the algorithm can continue to evolve towards excellent solutions. This method can ensure that the best individuals will not be lost during the evolution process, thereby improving the overall performance of the algorithm. The specific formula is:

[0192]

[0193] The roulette selection mechanism is adopted to allocate the selection probability according to the individual fitness value, and introduce random factors to increase the randomness of the selection process. This can introduce more uncertainty in the selection process and avoid falling into the local optimal solution, thereby improving the global search ability. The selection probability of the roulette selection is calculated as follows:

[0194]

[0195] Introducing a competitive selection mechanism, comparing the fitness values ​​of two randomly selected individuals, and selecting the individual with a higher fitness value. This method can increase the competitiveness of the population while retaining excellent individuals, thereby promoting the improvement of overall fitness.

[0196] The selection operation can guide the algorithm to approach excellent solutions while maintaining population diversity, thereby improving the algorithm's global search capability and convergence speed. The combination of these strategies improves the performance of the algorithm and also enhances its adaptability and flexibility in complex optimization problems.

[0197] S4. The selected individuals will undergo crossover and mutation operations. The crossover operation generates offspring individuals by combining the characteristics of the parent individuals, while the mutation operation randomly changes certain characteristics of the individuals to introduce new genetic information. The crossover and mutation operations work together on the population to increase its diversity and adaptability.

[0198] The crossover operation is improved as follows:

[0199] like Figure 2 As shown, two or more firework individuals are selected as parents, and some of their features are exchanged according to certain rules to generate new firework individuals as offspring:

[0200] O child =α·P1+(1-α)·P2

[0201] Among them, P1 and P2 are the characteristics of the two parent individuals, α is the crossover coefficient, which is used to control the weight of the parent characteristics. The formula generates offspring by weighted combination of the parent characteristics to maintain the diversity of solutions;

[0202] The selection of the crossover point can be random or based on some heuristic rules, and the crossover operation can be single-point crossover, multi-point crossover or uniform crossover;

[0203] The crossover probability determines the probability of the parent individual performing a crossover operation, which affects the algorithm's exploration ability. Usually, the crossover probability P crossover Set to a moderate value to ensure that the algorithm can maintain diversity and converge effectively;

[0204]

[0205] N crossover is the number of individuals to be crossovered, N total is the total population.

[0206] In order to adapt to the complex path optimization problem, the path length and quality are considered in the crossover operation. When performing the crossover operation, individuals with shorter path length and higher quality are preferentially selected as parents, in the hope that the offspring can inherit better characteristics. The parent fitness value F parent The calculation formula is:

[0207]

[0208] L path is the path length, C path is the comprehensive cost of the path, γ is the trade-off coefficient between path length and cost;

[0209] Compare the fitness of individuals before and after crossover to evaluate the effect of the crossover operation. If the crossover operation can produce offspring with higher fitness, the crossover operation is successful.

[0210] The crossover operation is improved by introducing the consideration of path length and quality to increase the exploration depth of the solution space. The crossover operation can guide the algorithm to approach excellent solutions while maintaining population diversity, thereby improving the algorithm's global search capability and convergence speed.

[0211] The formula for the mutation operation is as follows:

[0212]

[0213] O mutated (k) represents the individual characteristics after mutation, P(k) is the characteristic of the original individual, σ is the variation amplitude, N(0, 1) is the random noise introduced by the standard normal distribution, P mutation is the mutation probability;

[0214] The mutation probability determines the probability of an individual performing a mutation operation, which affects the diversity and innovation ability of the algorithm. Usually, the mutation probability is set to a lower value to ensure the stability of the algorithm and introduce enough new genetic information;

[0215] In order to adapt to complex path optimization problems, this study introduced an adaptive mutation strategy based on path quality in the mutation operation. When performing the mutation operation, individuals with poor path quality are preferentially selected for mutation in the hope of generating better solutions through mutation. The evaluation formula of path quality Qpath(i) is:

[0216]

[0217] Compare the fitness of individuals before and after mutation to evaluate the effect of the mutation operation. If the mutation operation can produce individuals with higher fitness, then the mutation operation is considered successful.

[0218] Improvements in mutation operations With the adaptive mutation strategy, mutation operations can guide the algorithm to explore new solution spaces while maintaining population diversity, thereby improving the algorithm's global search capability and convergence speed.

[0219] S5. The update of pheromones is based on individual performance. Excellent path solutions will increase pheromones, which will be explored more in subsequent searches, helping the algorithm remember and utilize past successful experiences.

[0220] The update of pheromone is no longer fixed, but dynamically adjusted according to the search state of the algorithm. The dynamic adjustment of pheromone volatilization rate can be expressed as:

[0221]

[0222] ρ(t) represents the pheromone volatilization rate at time t, ρ0 is the initial volatilization rate H current and H max Respectively represent the current solution set entropy value and the maximum entropy value;

[0223] Path reward and punishment mechanism: After each iteration, all generated paths are rewarded or punished according to their quality. High-quality paths will gain additional pheromones, increasing their probability of being selected in subsequent iterations; while poor-quality paths will have their pheromones reduced, reducing their probability of being selected.

[0224]

[0225] The pheromone update formula, the improved pheromone update formula is as follows:

[0226] τ ij (t+1)=(1-ρ)·τ ij (t)+Δτ ij (t)

[0227] where τ ij is the pheromone concentration on path (i, j) at time t, ρ is the pheromone volatility, Δτ ij (t) is the amount of pheromone added on the path (i, j) at time t, which is determined by the number of ants passing through the path and the path quality;

[0228] When the algorithm shows signs of convergence, the pheromone volatilization rate will increase to prevent premature convergence; when the algorithm falls into a local optimum, the pheromone volatilization rate will decrease to increase the exploration of the solution space;

[0229] With the improved pheromone update strategy, the Fireworks algorithm can better balance exploration and development when solving the cold chain logistics transportation path optimization problem, avoid falling into the local optimal solution, and accelerate the convergence speed of the algorithm and improve the quality of the solution;

[0230] S6, termination judgment, check whether the algorithm reaches the preset threshold of fitness value change. If the termination condition is met, the algorithm will stop, which ensures that the algorithm will not run indefinitely. If the termination condition is not met, return to S2 and continue iteration;

[0231] Specifically, first, define the convergence standard and set a fitness value change threshold fitness_threshold;

[0232] Secondly, record the fitness value. At the end of each iteration, record the current fitness value current_fitness and save it in a list fitness_history;

[0233] Finally, check whether the fitness value has converged. Traverse fitness_history and check whether the fitness value change in the most recent n_consecutive_iterations iterations is less than fitness_threshold. If the fitness value change in consecutive n_consecutive_iterations iterations is less than fitness_threshold, the algorithm is considered to have converged.

[0234] S7, output the optimal solution. Finally, after the algorithm ends, the individual with the highest fitness is selected from the current population, and the transportation path plan represented by the individual is output as the optimal solution, ensuring that the result of the algorithm is a practical and valuable solution that can be used by the end user.

[0235] The above description is only a preferred specific implementation manner of the present invention, but the protection scope of the present invention is not limited thereto. Any technician familiar with the technical field can make equivalent replacements or changes according to the technical scheme and inventive concept of the present invention within the technical scope disclosed by the present invention, which should be covered by the protection scope of the present invention.

Claims

1. A cold chain logistics vehicle planning system, comprising the following steps: Create initial population → Explosion and spark generation → Evaluation and selection → Genetic operation → Pheromone update → Termination judgment → Output optimal solution, including: S1. Create an initial population, each individual represents a potential transportation path solution. These individuals are randomly placed in the search space, and their fitness is evaluated based on the total cost and efficiency of the path solution they represent; Due to the complex terrain in the karst region, the driving speed and fuel consumption of vehicles are affected to varying degrees. Therefore, the cold chain transportation route optimization problem is subject to vehicle capacity constraints, route distance constraints, route cost constraints, cargo temperature constraints, customer time window constraints and customer service times constraints. Multiple constraints are analyzed and calculated. In the route distance constraint, route cost constraint, cargo temperature constraint and customer time window constraint, terrain factors are introduced to reflect the impact of terrain on the constraints; The goal of cold chain transportation route optimization is to minimize the total transportation cost, which is composed of multiple aspects, including fixed cost, transportation cost, cargo damage and deterioration cost, and time window penalty cost. Multiple aspects of cost are analyzed and calculated to establish the objective function. In karst areas, due to the complexity of the terrain, the driving speed and fuel consumption of vehicles are affected to varying degrees. Therefore, in the calculation of transportation costs and cargo damage and deterioration costs, terrain factors are introduced to consider the impact of terrain on transportation costs and cargo quality, providing a more reasonable and efficient transportation route solution for cold chain logistics in karst areas. Specifically, the terrain factor is defined as: where f ij represents the terrain factor between the i-th customer and the j-th customer, v ij represents the actual driving speed between the i-th customer and the j-th customer, v0 represents the ideal driving speed on flat terrain, and e ij represents the actual fuel consumption between the i-th customer and the j-th customer, e0 represents the ideal fuel consumption on flat terrain, and r ij represents the actual cargo spoilage rate between the i-th customer and the j-th customer, r0 represents the ideal cargo spoilage rate on flat terrain, and s ij is the actual slope impact between the i-th customer and the j-th customer, s0 is the road slope impact on flat terrain, the larger the terrain factor, the greater the impact of the terrain on the cost, and vice versa; S2, based on the fireworks algorithm, simulates the phenomenon of fireworks explosion in nature to solve the optimization problem, where each firework individual generates new sparks according to its fitness, and these sparks explore the solution space around the original individual to find potential improvement opportunities; During the operation of the algorithm, each firework represents a candidate solution, and the sparks produced by the explosion of the firework represent new candidate solutions. These sparks propagate in the solution space according to their fitness values, thereby exploring new possible solutions. The fireworks algorithm can be used to solve optimization problems of the following form: minf(x)stx∈Ω Where x=(x1,x2,...,x d ) is the decision variable vector, f(x) is the objective function, Ω is the feasible domain, and d is the dimension of the problem; A certain number of fireworks are randomly generated in the feasible domain Ω, each firework represents a candidate solution, and the fitness value of each firework, that is, the objective function value, is calculated; According to the fitness value of each firework, the number of sparks and explosion radius generated by each firework are determined, which can ensure both sufficient global search capability and good local search capability; During the spark generation process, the quality of each potential solution is evaluated by a heuristic factor, which can be a comprehensive score based on factors such as the route's historical transportation efficiency, cost, and safety record; Calculation of heuristic factor,The calculation of heuristic factor can be performed by the following formula: H(i)=ω1E(i)+ω2C(i)+ω3S(i) Where H(i) is the heuristic factor of the i-th solution, E(i) is the efficiency score, C(i) is the cost score, S(i) is the safety score, ω1, ω2, ω3 are the corresponding weight coefficients, which can be adjusted according to actual conditions; Solutions with higher heuristic factors will have a greater probability of generating more sparks and thus be explored more in subsequent searches. The relationship between the number of sparks and the heuristic factor can be expressed as: N sparks (i)=N base +α·H(i) Dynamically adjust the heuristic factor. In order to adapt to the changing environment and data, the heuristic factor needs to be dynamically adjusted: H(i,t+1)=H(i,t)·(1-β·ΔR(i)) If a route becomes infeasible due to an unexpected event, the heuristic factor of the associated solution will be reduced, and vice versa; S3. Newly generated sparks will be evaluated according to their fitness, and the next generation population will be determined based on the fitness value. The best individuals are selected to maintain the progress and effectiveness of the algorithm. For each firework, according to its number of sparks and explosion radius, a corresponding number of sparks are randomly generated around it. Each spark also represents a candidate solution. The fitness value of each spark is calculated. The spark generation formula is as follows: in, is the kth spark of the ith firework, x i is the i-th firework, A i is the explosion radius of the ith firework, is a random vector, each component of which follows a uniform distribution between -1 and 1. In order to increase the diversity of the population, for each firework, a certain number of dimensions are randomly selected to perform Gaussian mutation operations to generate Gaussian sparks. Each Gaussian spark also represents a candidate solution. The fitness value of each Gaussian spark is calculated. The generation formula of Gaussian sparks is as follows: in, is the Gaussian spark of the ith firework, x i is the ith firework, σ is a constant used to control the intensity of Gaussian variation, g i is a random vector whose components follow a standard normal distribution; Among all fireworks, sparks and Gaussian sparks, select the one with the best fitness value as the current optimal solution and keep it for the next generation. Then, according to the distance from each candidate solution to other candidate solutions, select a certain number of candidate solutions as the fireworks for the next generation. In order to maintain the diversity of the population, not only the fitness value of the individual is considered, but also the differences between individuals. Even if the fitness value of an individual is not the highest, but it can bring new genetic information to the population, it may also be selected to increase the genetic diversity of the population and avoid premature convergence. The improved formula is: f(i) is the fitness value of the individual, δ(i) is the genetic diversity contribution value of the individual, and N is the size of the population; Regardless of the diversity of the population, the elite retention strategy retains a certain proportion of the best individuals to ensure that the algorithm can continue to evolve towards excellent solutions. This method can ensure that the best individuals will not be lost during the evolution process, thereby improving the overall performance of the algorithm. The specific formula is: The roulette selection mechanism is adopted to allocate the selection probability according to the individual fitness value, and introduce random factors to increase the randomness of the selection process. This can introduce more uncertainty in the selection process and avoid falling into the local optimal solution, thereby improving the global search ability. The selection probability of the roulette selection is calculated as follows: Introduce a competitive selection mechanism to compare the fitness values ​​of two randomly selected individuals and select the individual with a higher fitness value; S4. The selected individuals will undergo crossover and mutation operations. The crossover operation generates offspring individuals by combining the characteristics of the parent individuals, while the mutation operation randomly changes certain characteristics of the individuals to introduce new genetic information. The crossover and mutation operations work together on the population to increase its diversity and adaptability. S5. The update of pheromones is based on individual performance. Excellent path solutions will increase pheromones, which will be explored more in subsequent searches, helping the algorithm remember and utilize past successful experiences. The update of pheromone is no longer fixed, but dynamically adjusted according to the search state of the algorithm. The dynamic adjustment of pheromone volatilization rate can be expressed as: ρ(t) represents the pheromone volatilization rate at time t, ρ0 is the initial volatilization rate H current and H max Represent the current solution set entropy value and maximum entropy value respectively; Path reward and punishment mechanism: After each iteration, all generated paths are rewarded or punished according to their quality. High-quality paths will gain additional pheromones, increasing their probability of being selected in subsequent iterations; while poor-quality paths will have their pheromones reduced, reducing their probability of being selected. The pheromone update formula, the improved pheromone update formula is as follows: t ij (t+1)=(1-ρ)·τ ij (t)+Δτ ij (t) where τ ij is the pheromone concentration on path (i, j) at time t, ρ is the pheromone volatility, Δτ ij (t) is the amount of pheromone added on the path (i, j) at time t, which is determined by the number of ants passing through the path and the path quality; When the algorithm shows signs of convergence, the pheromone volatilization rate will increase to prevent premature convergence; when the algorithm falls into a local optimum, the pheromone volatilization rate will decrease to increase the exploration of the solution space; S6, termination judgment, check whether the algorithm reaches the preset threshold of fitness value change. If the termination condition is met, the algorithm will stop, which ensures that the algorithm will not run indefinitely. If the termination condition is not met, return to S2 and continue iteration; Specifically, first, define the convergence standard and set a fitness value change threshold fitness_threshold; Secondly, record the fitness value. At the end of each iteration, record the current fitness value current_fitness and save it in a list fitness_history; Finally, check whether the fitness value has converged. Traverse fitness_history and check whether the fitness value change in the most recent n_consecutive_iterations iterations is less than fitness_threshold. If the fitness value change in consecutive n_consecutive_iterations iterations is less than fitness_threshold, the algorithm is considered to have converged. S7, output the optimal solution. Finally, after the algorithm ends, the individual with the highest fitness is selected from the current population, and the transportation path plan represented by the individual is output as the optimal solution, ensuring that the result of the algorithm is a practical and valuable solution that can be used by the end user.

2. A cold chain logistics vehicle planning system according to claim 1, characterized in that: In step S4, the crossover operation is improved as follows: two or more firework individuals are selected as parents, and their partial features are exchanged according to certain rules to generate new firework individuals as children: The child =α·P1+(1-α)·P2 Among them, P1 and P2 are the characteristics of the two parent individuals, α is the crossover coefficient, which is used to control the weight of the parent characteristics. The formula generates offspring by weighted combination of the parent characteristics to maintain the diversity of solutions; The selection of the crossover point can be random or based on some heuristic rules, and the crossover operation can be single-point crossover, multi-point crossover or uniform crossover; The crossover probability determines the probability of the parent individual performing a crossover operation, which affects the algorithm's exploration ability. Usually, the crossover probability P crossover Set to a moderate value to ensure that the algorithm can maintain diversity and converge effectively; N crossover is the number of individuals to be crossovered, N total is the total population. In order to adapt to the complex path optimization problem, the path length and quality are considered in the crossover operation. When performing the crossover operation, individuals with shorter path length and higher quality are preferentially selected as parents, in the hope that the offspring can inherit better characteristics. The parent fitness value F parent The calculation formula is: L path is the path length, C path is the comprehensive cost of the path, γ is the trade-off coefficient between path length and cost; Compare the fitness of individuals before and after crossover to evaluate the effect of the crossover operation. If the crossover operation can produce offspring with higher fitness, the crossover operation is successful. The crossover operation can guide the algorithm towards an excellent solution while maintaining the diversity of the population.

3. A cold chain logistics vehicle planning system according to claim 1, characterized in that: In step S4, the formula of the mutation operation is as follows: O mutated (k) represents the individual characteristics after mutation, P(k) is the characteristic of the original individual, σ is the variation amplitude, N(0, 1) is the random noise introduced by the standard normal distribution, P mutation is the mutation probability; The mutation probability determines the probability of an individual performing a mutation operation, which affects the diversity and innovation ability of the algorithm. Usually, the mutation probability is set to a lower value to ensure the stability of the algorithm and introduce enough new genetic information; In order to adapt to the complex path optimization problem, this study introduced an adaptive mutation strategy based on path quality in the mutation operation. When performing the mutation operation, individuals with poor path quality are preferentially selected for mutation in the hope of generating better solutions through mutation. The path quality Q path The evaluation formula for (i) is: Compare the fitness of individuals before and after mutation to evaluate the effect of the mutation operation. If the mutation operation can produce individuals with higher fitness, then the mutation operation is considered successful. Improvements in mutation operations With adaptive mutation strategies, mutation operations can guide the algorithm to explore new solution spaces while maintaining population diversity.

Images