Research on Rural Logistics Terminal Distribution Efficiency Improvement in Rural Revitalization Strategy Assisted by Artificial Intelligence

With the promotion of the national rural revitalization strategy, the rural logistics industry has experienced rapid development nationwide. Rural logistics refers to all activities that provide logistics services to meet the needs of rural and agricultural production [1–2]. The rapid development of rural logistics industry not only brings the opportunity to improve the living and production conditions in rural areas, but also brings new opportunities for national economic development. The application of artificial intelligence in rural logistics can effectively improve the distribution efficiency of logistics [3–6].

With the rise of Internet logistics, intelligent logistics has become the new favorite of the logistics and distribution industry. The continuous maturity and development of artificial intelligence technology also provides great help for the high efficiency of the logistics and distribution industry. The application of artificial intelligence in logistics distribution has brought great changes to the logistics industry [7–9]. Through the application of technologies such as intelligent route planning, road condition prediction, drone delivery and logistics information tracking and management, more intelligent, efficient and accurate logistics distribution services can be realized. The evaluation of delivery efficiency needs to consider multiple indicators such as delivery time, error rate, cost control and user satisfaction [10–12]. The introduction of artificial intelligence technology will further enhance the efficiency and service quality of logistics and distribution, and provide customers with a better logistics experience. In the development trend of intelligent logistics, despite the rapid development of intelligent logistics system in urban areas, the construction of intelligent logistics in rural areas still exists many difficulties such as high cost and lack of talents, which need to be further improved in order to promote the revitalization of the countryside to promote the development of rural economy [13–16].

Literature [17] studied the application of ant colony algorithm in rural e-commerce logistics model. The drawbacks of the third-party distribution model were analyzed by constructing a model, and the shortest path and cost were calculated based on the ant colony algorithm using MATLAB software, which solved the difficulties of the third-party distribution model at the urban and rural ends. The results point out that the method not only improves the distribution efficiency, but also reduces the logistics cost. Literature [18] emphasized that intelligence is an inevitable trend in the development of intelligent logistics. It discusses the impact of artificial intelligence technology on the field of supply chain logistics, emphasizes the difficulties that exist in the intelligent development of this field, and puts forward targeted strategies aimed at promoting the development of modern industrial chain logistics towards intelligence. Literature [19] points out that artificial intelligence in the field of logistics not only improves the decision-making process, but also optimizes the use of resources and reduces environmental pollution. It also discusses the prominent challenges in sustainable logistics, trends in AI-driven logistics optimization, and future research directions. Literature [20] collected data through a structured questionnaire and analyzed the data using AMOS v25 and Structural Equation Modeling, and the results illustrated that AI improves the efficiency of logistics, and that cooperation between logistics companies reaches a key factor in this efficiency. Literature [21] explored the application of AI and lot technology in logistics and distribution route optimization. By collecting logistics data such as delivery records and real-time traffic, an optimization model based on AI and lot was established, and the effectiveness of the proposed scheme in practical applications was verified based on the analysis of cost-effectiveness and challenge response strategies. Literature [22] introduces the theoretical framework of IWD algorithm and discusses the constrained multi-vehicle logistics and distribution problem. The GWO algorithm is used to optimize the IWD algorithm appropriately, and the intelligent traffic information cloud is created by using digital intelligence AI technology, and the effectiveness of the cited algorithm is proved by example simulation. Literature [23] describes the use of intelligent lACA to solve the logistics and distribution path planning and multiple distribution center problems. Simulation experiments were carried out with the help of intelligent lACA and Cat CGA, and the results emphasized that intelligent lACA can effectively plan logistics distribution paths, improve distribution efficiency, and reduce distribution costs. Literature [24] describes the current status and development prospects of AI application in the logistics industry, pointing out that with the application of AI technology in logistics enterprises, the logistics link is optimized and the logistics efficiency is improved.

In this paper, based on the current situation of rural logistics distribution, a model of rural logistics terminal distribution based on crowdsourcing service is proposed. Then, by adjusting the traditional LRP model, the LRP optimization model of rural logistics terminal distribution based on the hybrid mode of delivery-to-home + self-pickup is constructed, and the improved potential field ant colony algorithm is used for solving and case analysis. Finally, the structural equation model is used to investigate the relevant factors and the interaction mechanism between the factors affecting the improvement of the distribution efficiency of rural logistics terminals, so as to propose the optimization path to improve the distribution efficiency of rural logistics terminals.

2

Current situation and mode of terminal distribution in rural logistics

2.1

Current situation of rural logistics and distribution

Rural logistics is the “last kilometer” of the logistics process, which is directly facing the rural consumer. According to the survey, rural logistics there are inherent transportation disadvantages and population distribution is too dispersed and other issues, the current rural logistics infrastructure is not perfect, the vast majority of rural logistics is used in the agency point collection mode. Agency point is generally a store in the town, the courier company directly with the agency to deal with the express delivery in the region, and then generally by the consumer to pick up and send, special circumstances can also ask the agency to send people to pick up and send and pay the delivery cost, so theoretically speaking, the agency has become the “end” of the rural logistics. This collection mode is led by the e-commerce platform, which combines individual merchants to provide consumers with collection services. Major courier companies set up express points in this model.

For the courier company, the agency point collection mode is the most efficient operation, rural logistics, if the real express delivery company to each household is basically impossible, but the logistics end point in the agency does not mean that the end of the logistics, but only a helpless early “end”, in this case, rural residents can not get the Maximum satisfaction. Currently, there is basically no effective organization in rural areas for the distribution of “to the home”, because the agency point of the individual business is completely different from the courier company, can not be shaped logistics delivery management model. And the cooperation between the two and the practice will inevitably produce a variety of problems, so for the solution of the “last kilometer” of the rural terminal logistics distribution problems have great limitations.

2.2

Crowdsourcing Model for Rural Logistics Terminal Distribution

In order to change the status quo of rural logistics terminal distribution and improve the satisfaction of rural residents with logistics distribution, this paper proposes a crowdsourcing service model for logistics terminal distribution in order to promote the development of rural logistics industry on the basis of meeting the distribution needs of the residents, improve the vitality of the rural economy, and add bricks to the realization of the strategy for the revitalization of the countryside.

The logistics crowdsourcing model involves transferring the distribution work carried out by employees of the original enterprise to a large group outside the enterprise to complete. The rural logistics terminal crowdsourcing mode is a third-party logistics operation mode that is independent of producers and consumers, and can provide services to both at the same time. Rural residents and express delivery enterprises share an O2O platform, which is built in the form of cell phone APP or WeChat public number, with a complete operational mechanism and standardized management policy, in order to maximize the guarantee that both as a service object and as a participant in the main body of the rural residents can be docked with the platform. When the customer puts forward the distribution service demand on the platform, if the courier enterprise can meet or need to pay a lower additional cost can be satisfied, you can choose the door-to-door self-supporting distribution service mode. If the courier company can not meet the needs of customers through the lower cost of all-round, the recipient can be released on the O2O platform, the crowdsourcing task will be sent to the willingness to take over the contract and have the qualifications of the contractor.

When rural residents want to participate in the platform services, they can submit their service wishes to the platform and upload their personal real information, and the platform authenticates their identity information according to the user registration information and establishes relevant credit mechanisms, so that they can become the contractors of the rural O2O platform, and the contractors can use their spare time to complete the task of contracting on a part-time basis, and they can independently select the delivery routes. The O2O platform also conducts real-time credit assessment, qualifications and qualifications of contractors, and provides a comprehensive service for contractors. The O2O platform also conducts real-time credit assessments for contractors, and the level of qualification and credit will directly affect the assignment of tasks.

3

Artificial intelligence-assisted rural logistics terminal distribution optimization model

This paper constructs a mixed-mode LRP model for rural logistics terminal distribution to home pickup, and proposes a potential field ant colony algorithm to optimize and solve the model, so as to realize the improvement of rural logistics terminal distribution efficiency with the assistance of artificial intelligence.

3.1

LRP Optimization Model for Rural Logistics Terminal Distribution

3.1.1

Model description

In this paper, the rural logistics terminal distribution to home pickup hybrid model LRP model is some changes to the traditional LRP model, the same point is that both have to carry out the study of site selection and transportation routing arrangements, the difference is that the traditional LRP site selection is the object of the distribution center, and the rural logistics terminal distribution pickup site selection path is the next level of the pickup service point for the number and location of the determination of the number and location of the service point [25]. The traditional LRP model and this paper’s hybrid model for rural logistics terminal distribution-to-home self-pickup are schematically shown in (a) and (b) in Fig. 1, respectively. The description of the LRP mathematical model for the rural logistics terminal hybrid distribution mode studied in this paper is specified as follows:

G = (A,P) is an undirected network graph, A = {0,1, 2,…, n} denotes the set of vertices including distribution centers N_d, pickup service points N_o and customer point sets N_c, and P ={(a_i,a_j) : a_i,a_j ∈ A} is the set of edges denoting the distance or distance cost between two connected vertices. That is, the distribution system consists of three types of nodes: distribution centers, self-pickup service stations, and customers. In this case, the location of the customer’s point and their demand are known, and the customer specifies a delivery time window. After the vehicle is issued from the distribution center, when the goods are delivered from the distribution center to the customer, there are two optional services for any customer: self-pickup and direct delivery. By determining the location and number of pickup points, customers requiring pickup and delivery are planned based on their time window and proximity to the pickup service point. We serve pickup customers within the range and dispatch vehicles from the distribution center to deliver to pickup points and customers who require delivery. With the goal of minimizing the total cost, in the transportation process, the vehicle’s carrying capacity and the arrival of the self-pickup point can not exceed its maximum capacity limit, for the time window limit customers, if not delivered on time, the enterprise needs to bear a certain amount of time to penalize the cost.

3.1.2

Basic assumptions

Since the rural logistics terminal distribution to home self-pickup hybrid model LRP model designed in this paper is an abstraction of the real problem, considering the complexity of the actual problem, this paper makes the following assumptions to ensure the operability of the model: 1)

It is assumed that the location of each node and the demand is fixed and known, the customer has certain requirement limitations on the time window, and the self-pickup service point and vehicle have maximum capacity constraint limitations.

2)

Assume that each customer can only visit one pickup service point at a time to pick up goods.

3)

Assume that a customer can only be served by one vehicle and a self-pickup service point can only be served by one vehicle.

4)

If the distribution object is fresh products, assume that the distribution center and the self-pickup service station have perfect temperature-controlled cold storage facilities, and that the freshness of fresh products in the distribution center and self-pickup equipment is 100%, and there is no decline in freshness wear and tear and so on. At the same time, the transportation vehicles are equipped with certain cold storage equipment such as insulation boxes, but the freshness of fresh products will diminish over a certain period of time.

5)

Vehicles in the transportation process are traveling at a constant speed, without considering the impact of traffic jams on the distribution of goods.

6)

It is assumed that there is only one delivery each time at the self-pickup service station, and there is no delivery failure or secondary delivery.

3.1.3

Description of symbols

To facilitate the specific description of the model, the parameters and variables used in this paper are defined as follows:

N_D –––Distribution Center Collection N_D = {1}.

N_O –––Meet at the Pickup Service Station, N_O = {2,3,4,…, n + 1}.

N_C –––Client point collection, N_C = {n + 2, n + 3, n + 4,…, n + 1 + m}.

A –––The full set of nodes, A = N_D ∪ N_o ∪ N_C.

K –––Vehicle assembly, K = {1,2,3,…,k}.

d_ij –––Distance from node i to node j.

q_i –––Demand at customer point i.

Q_k –––Capacity Limit for Distribution Vehicle k.

Q_o –––Capacity limitations for Pickup Service Station o.

e_i –––The earliest arrival time expected by Customer i.

l_i –––Customer i desired latest arrival time.

ee_i –––Earliest acceptable arrival time for Customer i.

ll_i –––The latest acceptable arrival time for Customer i.

S_ik –––Service time of vehicle k at node i.

α –––A cost factor for vehicles arriving earlier than the customer expects.

β –––Cost factor for vehicles arriving later than the customer’s desired arrival time.

c –––Unit transportation costs.

w –––Unit origination costs.

C_o –––Fixed costs of opening and operating a pickup kiosk.

f_ds –––Distance penalty function unit cost.

D –––The furthest pickup distance acceptable to the customer.

d –––The optimal pickup distance acceptable to the customer.

T_i –––The time at which quality loss occurs in fresh produce.

T_n –––The time when fresh products become unavailable for sale.

M –––Very large positive numbers.

Δg –––Unit loss per unit of shipment during delivery from a distribution center.

SA_ik –––Time for vehicle k to reach node i.

SL_ik –––The time at which vehicle k leaves node i.

T_ij –––The travel time for a vehicle to reach node j from node i.

u_ij –––Auxiliary variables to prevent subcircuits in transportation routes.

θ_i –––Cargo quality spoilage rate at vehicle arrival at customer i.

F(SA_ik) –––Time penalty cost of vehicle k in the distribution network in reaching node i.

DS(d_ij) –––Distance penalty costs for firms in the distribution network that exceed the customer’s desired pickup distance.

O_i –––0-1 variable, 1 if the self-service point is open, 0 otherwise.

x_i,k –––0-1 variable, 1 if vehicle k accesses node j from node i, 0 otherwise.

y_ij –––0-1 variable, 1 if customer i is served j by a self-service pickup station, 0 otherwise.

3.1.4

Objective function analysis

The objective function in this paper is divided into four components: the cost of using pick-up kiosks, the transportation cost of delivery, the cost of using refrigerated logistics delivery trucks, and the penalty cost of customer service, which minimizes the total cost. The specific cost components are as follows: 1)

Cost of opening and operating a self-pickup service station

The cost of opening and operating a self-pickup point of rural terminal logistics distribution is the most important investment cost in its site selection stage. Among them, such as placing refrigeration equipment, rent, geographic conditions and labor costs will affect the cost of opening and operating the self-pickup service station. Taken together, the total cost of opening and operating a self-pickup service station determined in this paper is as follows: 1 $Z_{p i c k - u p} = \sum_{i \in N_{o}} C_{o} O_{i}$ where C_o is the cost of opening and operating a pickup service point, and O_i is a 0-1 variable indicating whether or not the pickup point is open.

2)

The cost of dispatching vehicles

In the transportation of the terminal, for the fresh products in the goods to ensure their freshness, the distribution vehicle should have certain refrigerated temperature control equipment, and thus the vehicle has a certain departure cost: 2 $Z_{v e h i c l e} = \sum_{i \in N_{D}} \sum_{j \in N_{o} \cup N_{c}} \sum_{k \in K} w x_{i j k}$

3)

Transportation costs in the distribution process

The costs incurred by vehicles in the transportation process are generally related to the length of the path. Factors affecting transportation costs are generally distance, transportation time, fuel consumption and so on. And time, distance, etc. and transportation costs are generally proportional. In the optimization study of vehicle path, the most common objective function is the shortest distance or the lowest cost. Therefore, the transportation costs required for distribution using logistics vehicles are as follows: 3 $Z_{t r a n s p o r t} = \sum_{i \in A} \sum_{\begin{array}{l} j = A \\ i \neq j \end{array}} \sum_{k = K} c \cdot d_{i j} \cdot x_{i j k}$

4)

Customer Service Penalty Costs

In this paper, three types of penalty costs related to customer service are stipulated to discipline firms and used to improve customer satisfaction, which are the penalty cost for violating the time window, the penalty cost for decreasing freshness, and the penalty cost for the distance of self-pickup.

(1)

Penalty cost of violating time window

According to the hybrid time window constraint, the time window penalty function established in this paper is as follows: 4 $F (S A_{i k}) = {\begin{array}{l} I N & S A_{i k} < e e_{i} \\ α \cdot (e_{i} - S A_{i k}) & e e_{i} \leq S A_{i k} < e_{i} \\ 0 & e_{i} \leq S A_{i k} \leq l_{i} \\ β \cdot (S A_{i k} - l_{i}) & l_{i} < S A_{i k} \leq l l_{i} \\ I N & S A_{i k} > l l_{i} \end{array}$

Where: α is the cost factor if the vehicle arrives before the customer’s allowed time window. β is the cost factor when the vehicle arrives later than the customer’s allowed time window.

The total time penalty cost after distribution to the customer’s point is then: 5 $Z_{t i m e} = \sum_{i \in N_{o} \cup N_{c}} \sum_{k \in K} F (S A_{i k})$

(2)

Freshness penalty cost

Freshness penalty cost refers to the enterprise in the distribution process, the enterprise will bear because of the product freshness decline caused by the loss of demand or customer satisfaction decline generated by the penalty cost. To calculate the penalty cost of decreased freshness, it is first necessary to be clear about the degree of change in fresh product quality over time. A curve reflecting the stage of product quality change over time is shown in Figure 2.

where θ_i is the spoilage rate of the product. The graph shows that when the vehicle arrives at customer i in the range of [0,T], the product spoilage rate of customer i is θ_i = 0, i.e., the product is intact and there is no rotting and deterioration. When the vehicle arrives at customer i in the range of [T_l,T_n], the cold chain product quality spoilage rate $θ_{i} = 1 - \frac{T_{n} - S A_{i k}}{T_{n} - T_{l}}$ When the vehicle arrives at customer i in a time greater than T_n, the cold chain product quality spoilage rate θ_i = 1, because at this time the product distribution is meaningless, and thus the penalty is maximum. The product quality spoilage rate is expressed as: 6 $θ_{i} = {\begin{matrix} 0 & 0 \leq S A_{i k} < T_{l} \\ 1 - \frac{T_{n} - S A_{i k}}{T_{n} - T_{i}} & T_{l} \leq d_{i j} < T_{n} \\ 1 & S A_{i k} \geq T_{n} \end{matrix}$

The resulting penalty cost for a product’s decline in freshness during distribution is expressed as follows: 7 $Z_{f r e s h} = \sum_{i \in N_{o} \cup N_{e}} Δ g \cdot q_{i} \cdot θ_{i}$ Δg is the unit loss value of cold chain products per unit shipment from the distribution center to the pickup point or customer, q_i is the customer i demand, and θ_i is the customer i quality spoilage rate.

(3) Self-pickup distance penalty cost

Distance penalty cost means that the location of the pickup service point should be prioritized by considering the geographical location of the customer, because the customer will measure the convenience and travel cost to decide whether to need the pickup service or to choose the service of delivery to the house. The optimal pickup distance acceptable to the customer is d, and the farthest pickup distance acceptable to the customer is D. If the distance from the pickup point i to the customer j is less than d, then DS(d_ij) = 0. If the distance from the pickup point i to the customer j is between the optimal pickup distance d and the farthest pickup distance D, then $D S (d_{i j}) = 1 - \frac{D - d_{i j}}{D - d}$ . If the distance from the pickup point is greater than the maximum distance that the customer can tolerate, then D, DS(d_ij) = 1. i.e., the penalty function of the pickup distance is as follows: 8 $D S (d_{i j}) = {\begin{array}{l} 0 & 0 \leq d_{i j} < d \\ 1 - \frac{D - d_{i j}}{D - d} & d \leq d_{i j} < D \\ 1 & d_{i j} \geq D \end{array}$

The total distance penalty cost to the firm is thus: 9 $Z_{d i s} = \sum_{i \in N_{C}} \sum_{j e N_{O}} f_{d s} \cdot D S (d_{i j})$

Based on the above description, the total penalty cost is shown in the following equation: 10 $\begin{array}{l} Z_{p e n d l y} & = \sum_{i \in N_{o} \cup N_{c}} \sum_{k \in K} F (S A_{i k}) + \sum_{i \in N_{c}} \sum_{j \in N_{o}} f_{d s} \cdot D S (d_{i j}) \\ + \sum_{i \in N_{o} \cup N_{c}} Δ g \cdot q_{i} \cdot θ_{i} \end{array}$

3.1.5

Mathematical models

11

\min Z = Z_{p i c k - u p} + Z_{v e h i c l e} + Z_{t r a n s p o r t} + Z_{p e n a l t y}

12

\sum_{i \in A} \sum_{\begin{matrix} j \in N_{o} \cup N_{c} \\ i \neq j \end{matrix}} x_{i j k} \cdot q_{j} \leq Q_{k}, \forall k \in k

13

\sum_{i \in N_{c}} y_{i j} \cdot q_{i} \leq Q_{v}, \forall j \in N_{o}

14

y_{i j} = \max {\frac{- \min {(d_{i j} - \min_{\begin{matrix} j = n + 2 n + 3 \\ n + 1 + m \end{matrix}} (d_{i j})), 0}}{d_{i j} - \min_{\begin{matrix} j = n + 2, n + 3 \\ - n + 1 + m \end{matrix}} (d_{i j})} + 1, 0}, \forall i \in N_{c}, j \in N_{o}

15

\sum_{j \in N_{o} \cup N_{c}} x_{i j k} = \sum_{j \in N_{o} \cup N_{c}} x_{i j k} \leq 1, \forall i \in N_{D}, k \in k

16

\sum_{k \in k} \sum_{i \in A} x_{i k} = \sum_{k \in k} \sum_{i \in A} x_{j i k} = 1, \forall j \in N_{o} \cup N_{C}

17

\sum_{\begin{matrix} i \in A \\ i \neq j \end{matrix}} x_{i j k} = \sum_{\begin{matrix} i \in A \\ i \neq j \end{matrix}} x_{j i k}, \forall j \in N_{o} \cup N_{c}, k \in K

18

\sum_{k \in K} \sum_{i \in N_{D} \cup N_{o}} x_{i j k} + \sum_{i \in N_{o}} y_{i j} = 1, \forall j \in N_{c}

19

x_{i j k} = 0, \forall i \in N_{O}, k \in K

20

\sum_{k \in K} x_{i j k} = 0, \forall i \in N_{o}, j \in N_{o}

21

\sum_{k \in k} \sum_{\begin{matrix} i \neq A \\ i \neq j \end{matrix}} x_{i j k} \leq M \cdot O_{j}, \forall j \in N_{o}

22

\sum_{j \in N_{o}} y_{i j} \leq 1, \forall i \in N_{c}

23

u_{i k} - u_{j k} + n \cdot x_{i j k} \leq n - 1, \forall i \in A, j \in N_{o} \cup N_{c}, k \in K

24

u_{i k} \geq 0, \forall i \in A, k \in K

25

S L_{i k} = S A_{i k} + S_{i k}, \forall i \in N_{o} \cup N_{C}, k \in K

26

S A_{k} = \sum_{i \in A} x_{i j k} (S L_{i k} + T_{i j}), \forall i \in N_{o} \cup N_{C}, k \in K

27

e e_{i} \leq S A_{i k} \leq l l_{i}, \forall i \in N_{o} \cup N_{C}, k \in K

28

O_{i} = {0, 1}, \forall i \in N_{o}

29

x_{i j k} = {0, 1}, \forall i \in A, j \in A, k \in K

30

y_{i j} = {0, 1}, \forall i \in N_{b}, j \in N_{o} \cup N_{c}

Equation (11) represents the objective function that minimizes the total cost on the routes of this network, where the total cost includes the fixed cost of the pickup service point, the departure cost of the refrigerated delivery vehicle, the transportation cost during delivery, and the customer service penalty. The customer service penalties include the penalty cost for violating the time window, the freshness reduction penalty, and the pickup distance penalty cost.

Constraint (12) is the capacity constraint of the vehicle, which indicates that the amount of goods loaded on the vehicle cannot exceed the maximum capacity limit of the vehicle.

Constraint (13) is the capacity constraint of the pickup service point, which indicates that the amount of goods demanded by the customer to be delivered to the pickup service point cannot exceed the maximum capacity limit of the pickup service point.

Constraint (14) is to determine the service relationship between the self-pickup point and the self-pickup customer, if the customer i is within the maximum acceptance range of the self-pickup point j and is the closest to the self-pickup point, the customer will be categorized as served by the self-pickup point j.

Constraint (15) ensures that each path formed by the refrigerated logistics vehicle used for distribution is a closed loop, and the vehicle has to return to the corresponding origin after completing the distribution from the distribution center.

Constraint (16) guarantees that each customer and pickup service station is visited by only one vehicle and only once.

Constraint (17) denotes site flow balance, i.e., vehicles arrive and leave a node the same number of times and with the same amount of goods.

Constraint (18) denotes that for any customer i, it can only be delivered by a distribution center vehicle or picked up directly at a pickup point.

Constraints (19) and (20) denote that a self-pickup point cannot reach a self-service point and no path exists between the same nodes.

Constraint (21) denotes that only self-pickup kiosks are open to serve customers.

Constraint (22) denotes that a customer can only go to one self-pickup service point at a time to pick up the goods.

Constraints (23) and (24) denote constraints that prevent sub-loops from occurring, where u_ik is a non-negative auxiliary variable.

Constraint (25) denotes the time at which the vehicle leaves point i.

Constraint (26) denotes the time when the vehicle arrives at point j.

Constraint (27) denotes the time at which the vehicle arrives at point j to be within the customer’s time window.

Constraints (28) to (30) are 0-1 decision variables that define the range of the variables.

3.2

Logistics terminal distribution path optimization based on improved ant colony algorithm

3.2.1

Steps of the Ant Colony Algorithm

Ant Colony Algorithm (ACO) is an optimization algorithm developed based on simulating the behavior of ants searching for food and is mainly used to solve combinatorial optimization problems [26]. Its basic steps are as follows: 1)

Initialize pheromone: the pheromone on each path is initialized to a small positive number, which indicates that all paths have a certain possibility to be taken in the initial state.

2)

Ant selection of paths: each ant chooses one of the paths that can be traveled at the current position at any moment according to a certain probability, and the probability is related to the pheromone concentration on the paths and the length of the paths. Its probability selection is shown in equation (31): 31 $P_{i j}^{k} (t) = {\begin{array}{l} \frac{{[τ_{i j} (t)]}^{α} \cdot {[η_{i j} (t)]}^{β}}{\sum_{i v n} {[τ_{i j} (t)]}^{α} \cdot {[η_{i j} (t)]}^{β}}, j \in a_{k} \\ 0, O t h e r \end{array}$

In the above equation, $P_{i j}^{k} (t)$ denotes the state transfer probability of the knd ant transferring from grid i to grid j at the moment of t ; α denotes the pheromone importance factor, which indicates the importance of being influenced by the pheromone at the time of path selection. β is the expected heuristic information importance factor, indicating the magnitude of being influenced by the heuristic information at the time of path selection. τ_ij (t) is the pheromone concentration of ants moving from node i to node j at moment t. a_k denotes the set of feasible regions. η_ij (t) denotes the expected heuristic information of ants transferring from node i to node j at the t th moment, which is expressed as shown in Equation (32): 32 $η_{i j} (t) = \frac{1}{d_{i j}}$

Where, d_ij denotes the Euclidean distance between node i and node j. According to equation (32) it is found that η_ij (t) is inversely proportional to the distance from node i to node j. The Euclidean distance is shown in equation (33): 33 $d_{i j} = \sqrt{{(x_{i} - χ_{i})}^{2} + {(y_{i} - y_{i})}^{2}}$

3)

Updating pheromone: when all ants have finished traversing the entire path, the pheromone concentration on each path needs to be updated. Generally, at the end of each iteration, the pheromone volatilizes part of the pheromone while another part is added due to the pheromone left behind by the ants. This is usually done using the pheromone updating equation shown in (34): 34 $τ_{i j} (t + 1) = (1 - ρ) \cdot τ_{i j} + Δ τ_{i j}$

In the above equation, ρ is the volatilization factor of pheromone. 1 – ρ is the retention factor of the pheromone after volatilization and takes the value range (0,1). Δτ_ij is the pheromone added on the path after one iteration of the ants, which is evaluated as shown in Eq. (35): 35 $Δ τ_{i j} (t) = {\begin{array}{l} \begin{array}{l} \frac{Q}{L_{k}}, T h e p a t h 〈 i, j 〉 t h r o u g h w h i c h a n t k \\ p a s s e s t h i s i t e r a t i o n \end{array} \\ 0, O t h e r \end{array}$

Where Q is the pheromone intensity coefficient and L_k is the total length of the path traveled by the k rd ant to reach the end point. In summary, it can be seen that the smaller the length of the path L_k that the ant passes through in the path planning, the more pheromone is released in that path, and the greater the possibility that other ants are attracted to it, playing a positive feedback effect.

4)

Judge the termination conditions: repeat steps (2)~(3) until the pre-set stopping conditions are satisfied, such as reaching the maximum number of iterations, convergence to the threshold, etc., and output the optimal solution.

3.2.2

Steps of the artificial potential field algorithm

Artificial potential field algorithm (APF) is a typical model-free and optimization-free control method, which is widely used in robot path planning, obstacle avoidance and other fields [27]. Its basic steps are as follows: 1)

Build a robot model: first, a model of the robot needs to be built, including the size, shape and initial position of the robot and other information.

2)

Construct a two-dimensional or three-dimensional potential field model: including elements such as obstacles and target points. Among them, the obstacle generates repulsive force and the target point generates gravitational force.

3)

Calculate the gravitational potential field: Based on the distance between the target point and the robot position, the gravitational potential field can be calculated to represent the force with which the robot is attracted toward the target point. Assuming that the current node of the robot is X = (x, y) and the target node is X_g = (x_g, y_g). The gravitational potential field function is defined as shown in equation (36): 36 $U_{a} = \frac{1}{2} k (X - X_{g})$

Where: k is the gravitational field constant greater than zero. x is the robot node volume. X_g is the target node volume. The negative gradient of the gravitational potential field is the gravitational force of the target point on the robot as shown in equation (37): 37 $F_{a} = - g r a d (U_{a}) = - k (X - X_{g}) = k (X_{g} - X)$

4)

Calculate the repulsive potential field: based on the distance between the robot and the obstacle, the repulsive potential field can be calculated to represent the repulsive force on the robot, at which time the robot will be pushed away from the obstacle. The repulsive potential field function is defined as shown in equation (38): 38 $U_{r} = {\begin{array}{l} \frac{1}{2} m {(\frac{1}{ρ} - \frac{1}{ρ_{0}})}^{2}, ρ \leq ρ_{0} \\ 0, ρ > ρ_{0} \end{array}$ where m is the repulsive field constant greater than zero, ρ is the distance between the current node of the robot and the obstacle node, and ρ₀ is the maximum influence range of the repulsive potential field of the obstacle node. When ρ > ρ₀, the repulsive potential field of the obstacle node has no effect on the robot at this time. The negative gradient of the repulsive potential field is the repulsive force of the obstacle on the robot, as shown in equation (39): 39 $F_{a} = - g r a d (U_{r}) = {\begin{array}{l} m (\frac{1}{ρ} - \frac{1}{ρ_{0}}) \frac{1}{ρ^{2}}, ρ \leq ρ_{0} \\ 0, ρ > ρ_{0} \end{array}$

5)

Calculate the combined force on the robot at the current position: the gravitational and repulsive fields are merged to obtain the total potential field, which describes the robot’s state and direction of action in the environment. The robot will move towards the minimum value of the total potential field to reach the target point. The formula is shown in equation (40): 40 $F_{1} = F_{a} + F_{r}$

6)

Update robot position: the robot position is updated based on the current position of the robot and the gradient direction of the total potential field. If the robot encounters an obstacle, the repulsive field and the total potential field need to be recalculated.

7)

Check if the robot has reached the goal point or collided with an obstacle.

8)

If the robot has not reached the goal point, go back to step (5) and re-execute. If the robot has reached the goal point, the algorithm ends.

The workflow of the artificial potential field algorithm is shown in Fig. 3.

3.2.3

Potential Field Ant Colony Algorithm

In this paper, we propose an adaptive pheromone regulation to improve the ant colony algorithm by adding artificial potential field gravity to the heuristic information function, and we propose an optimization algorithm to improve the Ant Colony Algorithm for Potential Fields (AACO) [28]. 1)

Improved pheromone updating rule

In the traditional ant colony algorithm, as the number of ant colony iterations increases, the pheromone retained on the map gradually increases, resulting in the later ants walking more and more towards the paths with large concentration. With the volatilization of pheromones, the paths with few ants gradually become devoid of ants, which is prone to local optimization. In order to prevent the original ACO algorithm from falling into local optimality, an adaptive pheromone regulation is proposed: It is to add the difference between the sum of pheromone constants carried by all ants on the optimal path derived from the previous generation and the sum of pheromone constants carried by all ants on the worst path derived from the original update pheromone increment to get the adaptive regulation pheromone τ_ij (t)*. The update method is as shown in Eq. (41): 41 $τ_{i j} {(t + 1)}^{*} = (1 - ρ) \cdot τ_{i j} + Δ τ_{i j} (t) + Δ τ_{i j}^{k} (t)$ where ρ is the volatilization factor of pheromone in the ACO algorithm. $Δ τ_{i j}^{k} (t)$ is the difference between the sum of pheromone carried by the ants on the optimal path and the sum of pheromone carried by the ants on the worst path, which is shown in (42): 42 $\begin{matrix} Δ τ_{i j}^{k} (t) = \sum_{k = 1}^{n} Δ^{*} τ_{i j}^{k} (t) \\ Δ^{*} τ_{i j}^{k} (t) = M \cdot \frac{Q}{L_{\min}} - N \cdot \frac{Q}{L_{\max}} \end{matrix}$

Where: $Δ^{*} τ_{i j}^{k} (t)$ denotes the pheromone increment left by the knd ant on path 〈i, j〉. M is the number of ants on the optimal path of this loop. N is the number of ants on the worst path of this iteration. L_min is the length of the optimal path for this iteration and L_max is the length of the worst path for this iteration.

2)

Improvement of the heuristic pheromone function

The original ant colony has the same pheromone concentration on the path at the beginning of the search. Because the ants don’t know which path is optimal yet, they will search globally at the same time, which is the advantage of the ant colony algorithm, i.e., global search ability. At this time, the main difference lies in the inspirational information, because the original formula path selection node inspirational is not strong, so it will lead to the path search has blindness, spend a lot of time and may search once the path is not the optimal, but due to the ant colony algorithm has a positive feedback, in the path is not the optimal path instead of the pheromone is the most, resulting in the traditional ant colony algorithm in the early stage of the search is slow and the search path may fall into the local optimization problem. Problems.

In order to solve the above problems, this paper adds artificial potential field gravity into the heuristic information function, which is equivalent to giving the ants a target point, which will not carry out a large area of useless search, enhances the targeting of the heuristic function in the search process, and improves the algorithm search efficiency. Based on the ant colony algorithm, the improved formula of heuristic information $η_{\bar{i}} {(t)}^{*}$ in this paper is shown in (43): 43 $η_{i j} {(t)}^{*} = \frac{α}{d_{i j}}$

Where: η_ij (t)* is the heuristic information improved by adding the gravitational factor of the potential field. d_ij denotes the Euclidean distance from node i to node j. α is the added gravitational factor, whose formula is shown in (44): 44 $α = φ^{g}$ where φ is a constant and takes values in the range of (0,1). g denotes the gravitational force of the artificial potential field from the current node j to the target node, which is given by Eq: 45 $g = β \cdot d (j, M)$ where β is the positive scale factor and d( j,M) denotes the distance between node j to the target node M.

Substituting Eqs. (43)~(45) into the improved heuristic information Eq. η_ij (t)* yields: 46 $η_{i j} {(t)}^{*} = \frac{φ^{β \cdot d (j, M)}}{d_{i j}}$

According to the improvement heuristic, it is known that the larger the distance to the target node, the larger the value of potential energy suffered by the unmanned vehicle. The smaller the distance, the smaller the value of potential energy suffered by the unmanned vehicle.

3)

Path optimization and update

Since the searched path may have a large number of turning points, the planned path not only increases in length, but also the logistics robot needs to make many turns and change the direction of movement during traveling, which is not conducive to fast movement. Due to the search rules of the ant colony algorithm in the raster graph, the path search can only be carried out along the direction of integer multiples of π/4, and it is impossible to break the limitation of the search step length. In order to solve the above problems, the connecting method can effectively remove the redundant turning points and folding lines, and its core idea is: connect the original path adjacent turning nodes, and if the path does not pass through the obstacles after the connection, remove the redundant turning between the two points and update the path. If the connected path passes through an obstacle, the path is taken as it was originally.

Let the unmanned vehicle walks through the set of turning points that need to be checked as N, M ={m₁,m₂ …,m_n} is the set of turning points of the path to be optimized, Q is the set of all points on the line connecting the turning points of the point to be checked N_i, M_i, and W is the set of obstacles. Add the turn point M_i to be optimized to the set of check points, and then sequentially connect the set of turn points in M to the first element in set N, and add the points on the connecting line to set Q. If W ∩ Q = N, this connection is used. If W ∩ Q ≠ N, the current concatenation is not used. And add the elements in set M to the first position of set N sequentially for verification, until all the elements in M are added to set N and the verification is completed. After judging all the turn nodes in turn, the path optimization is completed.

3.2.4

Steps of the potential field ACO algorithm

The steps of the potential field ant colony algorithm are shown below: 1)

Establish a raster environment model: use the raster method to establish a raster model of the unmanned vehicle distribution environment, and set up free grids and obstacle grids. The free grid is a white grid, indicating that it can be passed. The obstacle grid is a black grid, indicating that it can not be passed.

2)

Initialization parameters: initialize the maximum number of iterations of the algorithm K, the number of ants of the colony M, the pheromone importance factor α, the expected heuristic information importance factor β, the pheromone volatility coefficient ρ, and other parameters initialization settings.

3)

Selection of path nodes: ant k selects the next feasible node of the path according to the transfer probability formula, adds the visited nodes to the taboo table, and judges whether ant k reaches the target point until all m ants complete the search in this iteration.

4)

Record the length of each path and its number of ants: the paths traveled by each ant for each transfer and the length of the paths traveled by each ant in this iteration are recorded, and the number of ants on each path from the starting point of this iteration to the target point is recorded.

5)

Pheromone update: Based on the taboo table and the data recorded in step 4), find out the optimal and worst paths of this iteration and the number of mom ants on their paths, and update the pheromone concentration according to equations (41) and (42).

6)

End iteration: determine whether the algorithm reaches the maximum number of iterations, if it reaches the maximum number of iterations, output the optimal path. Otherwise, empty the taboo table and return to step 3) for another iteration.

7)

Secondary optimization: for the optimal route derived by the unmanned vehicle through the program, connect the redundant turning points to get the streamlined path.

4

Analysis of model applications

In order to verify the effectiveness of the constructed rural logistics terminal distribution optimization model, i.e., whether the model can effectively improve the distribution efficiency, this paper firstly analyzes the potential field ant colony algorithm used in the arithmetic case experiments, and then specifically explores the influencing factors of rural logistics terminal distribution.

4.1

Calculated experiments and analysis of results

4.1.1

Data setup

The examples of simulation experiments are selected from the Solomon benchmark to test the advantages and disadvantages of the established models and algorithms. In the case analysis, some of the data in the random C1 and R1 types are selected, because the customer positions in the C and R classes belong to the heap distribution and uniform distribution, which are representative. Depending on the number of customers selected, questions will be divided into small (less than 40 customers) and large (more than 40 customers). The coordinates of the warehouse center are (50, 50), this section is a simple example, and the maximum number of candidate distribution centers is 5, and the opening costs are 21000, 23000, 24000, 25000, and 27000 yuan respectively. The transportation capacity of the vehicle from the warehouse to the distribution center is 500, the enabling cost is 320 yuan/vehicle, and the driving speed is 45km/h. The average unit transportation cost in the first-level distribution system is 16 yuan/km.

4.1.2

Analysis of results

The potential field ant colony algorithm (AACO) proposed in this paper is used to solve the algorithm for 1000 iterations. When solving the algorithm, the computational results are compared with the ordinary ant colony algorithm (ACO) to verify the advantages and disadvantages of the algorithm. The validation results are shown in Tables 1 and 2, where Z* denotes the optimal solution of the objective function, the problem size denotes the number of candidate distribution centers and the number of customers, and the number of activations denotes the number of centers open and the number of vehicles activated for candidate distribution. Each algorithm is run 10 times and the difference between the optimal and worst solutions is calculated. 1)

Small-scale problem

In this paper, the number of customers in the small-scale problem is 10, 25 or 40, and the numerical results of the small-scale example are shown in Table 1.

As can be seen from Table 1, as the number of customers increases from 10 to 25 and then to 40, the error in the solution increases gradually, and the number of open distribution centers increases gradually, which is more in line with the actual situation. And R101, R102, R103 in the two algorithms, the difference is more obvious, in the case of the same number of customers, using the unimproved ACO algorithm needs to open the number of distribution centers is more, the cost is also larger, the maximum value of Z* reaches 88944. At the same time, from the overall can be seen that the CIA with the potential field ACO algorithm to solve the optimal solution between the worst solution is also less error in 61.54% of the cases of error in the best solution, the error in the worst solution is less. 61.54% of the cases the error is 0, while 92.31% of the solutions with the unimproved ACO algorithm have errors, which indicates that the improved potential field ACO algorithm is more stable and suitable for solving the rural logistics terminal distribution optimization problem.

2)

Large-scale problem

In this paper, the number of customers in the large-scale problem is 50 and the number of candidate distribution centers is increased from 3 to 5. The numerical results for the large-scale example are shown in Table 2.

It can be seen from Table 2 that with the increase of the number of candidate distribution centers in the case of a certain number of customers, the cost is gradually increasing in the R1 class and C1 class examples, and the error between the optimal solution and the worst solution of the two algorithms in solving is also gradually increasing. And in Table 2, we can see two special cases, i.e., when the number of candidate distribution centers is 5 and the number of customers is 50, the errors of C103 and C104 in the potential field ACO algorithm are 9.97% and 10.68%, respectively, and in the unimproved ACO algorithm the errors are 21.04% and 22.15%, respectively, because these two algorithms take longer time during the solution process, which leads to an increase in the error between the optimal solution and the worst solution. The error between the optimal solution and the worst solution increases. Meanwhile, it can also be observed from Table 2 that in the case of the same number of candidate distribution centers and the same number of customers, the number of open distribution centers obtained by using the two algorithms for the calculation is the same, but there is a difference in the number of vehicles enabled. Therefore, it can be concluded that the solution using the improved potential field ACO algorithm can effectively reduce distribution costs and optimize the distribution path.

Table 1.

Numerical results of small instances

Problem name	Problem size	AACO			ACO
Problem name	Problem size	Z*	Enabled quantity	Error/%	Z*	Enabled quantity	Error/%
R101	3-10	36997	1-4	0.00	66263	2-7	0.02
R102	3-10	35145	1-3	0.03	57981	2-7	0.29
R103	3-10	35909	1-3	0.49	60625	2-8	0.30
R104	3-10	30767	1-2	0.00	33135	1-5	0.01
R105	3-10	33962	1-5	0.00	37681	1-10	1.23
R106	3-10	32044	1-4	0.00	38927	1-8	0.30
R107	3-10	31960	1-4	0.00	39681	1-8	1.03
R108	3-10	31818	1-3	0.34	36388	1-6	0.58
R109	3-10	32791	1-2	0.00	34148	1-5	0.27
R110	3-10	31485	1-2	0.00	35882	1-6	0.00
R112	3-10	31425	1-2	0.00	35493	1-5	0.32
R105	3-25	55172	2-5	0.00	60608	2-8	0.83
R106	3-25	55870	2-4	0.00	58718	2-8	0.21
R107	3-40	82155	3-5	0.22	86005	3-7	0.37
R108	3-40	83961	3-5	2.33	87197	3-8	3.31
R110	3-40	83814	3-6	0.90	88944	3-9	1.20
R112	3-40	82437	3-5	1.04	84390	3-9	1.99
C101	3-25	58139	2-3	0.00	58977	2-6	0.24
C102	3-25	57810	2-3	0.00	60318	2-6	0.07
C103	3-25	57338	2-3	0.00	58974	2-7	0.37
C104	3-25	58042	2-3	0.00	59697	2-6	0.21
C105	3-25	58612	2-3	0.25	59619	2-8	0.00
C106	3-25	58259	2-3	0.00	59410	2-8	0.28
C107	3-25	59336	2-3	0.87	58724	2-7	0.26
C108	3-25	57823	2-3	0.05	58363	2-6	0.37
C109	3-25	57185	2-3	0.00	60336	2-8	0.07

Table 2.

Numerical results of large-scale instances

Problem name	Problem size	AACO				ACO
Problem name	Problem size	Z*	Enabled quantity	Error/%	Z*	Enabled quantity	Error/%
R101	3-50	54471	2-11	0.02	55454	2-13	0.27
R103	3-50	44567	2-8	0.03	50626	2-11	0.35
R105	3-50	57161	2-8	3.57	63124	2-13	1.58
R107	3-50	59978	2-6	0.04	63383	2-13	0.21
R109	3-50	55578	2-7	0.02	60524	2-10	0.25
R111	3-50	51209	2-6	0.23	53704	2-9	0.01
R101	4-50	79985	3-11	0.01	80423	3-14	0.09
R103	4-50	69772	3-8	0.00	72509	3-12	0.19
R105	4-50	73655	3-8	0.2	74794	3-11	0.14
R107	4-50	71807	3-6	0.00	75552	3-10	0.03
R109	4-50	73031	3-7	0.00	74737	3-12	0.39
R111	4-50	72617	3-6	0.02	73512	3-9	0.58
R101	5-50	102625	4-11	0.01	116441	4-13	1.25
R103	5-50	99020	4-8	0.02	104661	4-10	0.01
R105	5-50	100330	4-8	0.47	107958	4-11	1.25
R107	5-50	97618	4-6	0.56	104461	4-9	1
R109	5-50	99952	4-7	0.82	104891	4-11	0.59
R111	5-50	98772	4-6	0.58	102784	4-9	1.22
C101	3-50	79301	3-5	0.03	80340	3-8	0.51
C102	3-50	78063	3-5	0.05	81809	3-7	0.21
C103	3-50	79639	3-6	0.03	80458	3-8	0.33
C104	3-50	79459	3-5	0.02	79781	3-8	0.67
C105	3-50	80344	3-7	0.01	82754	3-9	0.27
C106	3-50	79538	3-8	0.00	81385	3-9	0.02
C107	3-50	81384	3-10	0.01	83391	3-12	0.78
C108	3-50	79543	3-7	0.00	81144	3-10	0.41
C109	3-50	80238	3-8	0.01	81064	3-9	0.82
C101	5-50	98822	4-6	0.01	101709	4-8	0.19
C102	5-50	99237	4-6	0.02	110225	4-9	0.42
C103	5-50	122854	5-6	9.97	138197	5-9	21.04
C104	5-50	124010	5-7	10.68	135480	5-11	22.15
C105	5-50	98611	4-6	0.01	102968	4-9	2.88
C106	5-50	98117	4-6	0.02	100435	4-10	1.5
C107	5-50	99178	4-6	0.03	100968	4-9	0.87
C108	5-50	100515	4-8	0.03	101638	4-11	0.57
C109	5-50	100958	4-9	2.26	101614	4-13	1.23

4.2

Analysis of factors affecting the efficiency of rural logistics terminal distribution

4.2.1

Questionnaire design

The comprehensive implementation of the rural revitalization strategy should be promoted in six aspects: industrial revitalization, talent revitalization, cultural revitalization, ecological revitalization and organizational revitalization. Accordingly, this paper combines the previous studies and takes into account the needs of rural residents for convenient logistics services and the employment and development needs of rural logistics terminal distribution point employees, and designs a questionnaire to measure the development level of rural logistics terminal distribution in terms of six latent variables, namely, economic efficiency, informationization level, service quality, industry development, ecological efficiency, and organizational efficiency, as well as the corresponding 20 observable variables.

Measurable variables include industrial development (X1), job creation (X2), agricultural trade (X3), safe delivery (X4), on-time delivery (X5), door-to-door delivery (X6), disinfection and sterilization (X7), customer satisfaction (X8), information updating (X9), information feedback (X10), smart access (X11), green packaging materials (X12), packaging recycling (X13), Green Transportation Planning (X14), Income Status (X15), Employment Prospects (X16), Willingness to Further Study (X17), Policy Support (X18), Business Cooperation (X19), and Station Management (X20).

The questionnaire was designed with a total of 40 questions, including 6 latent variables and 20 measurable variables, and the scoring was based on a 5-point Likert scale, whereby the filler scored the questions on a scale of 1-5.

4.2.2

Reliability and Validity Tests

In this paper, Cronbach’s alpha coefficient is used to measure the reliability of the designed questionnaire, and the results of the reliability test are shown in Table 3. The Cronbach’s alpha reliability coefficient takes the value range of [0,1], and the closer the value of 1 the better, if the reliability coefficient α of the scale ∈ [0.8,0.9], it means that the reliability of the scale is acceptable. If the scale’s reliability coefficient α < 0.7, it means that some items of the scale need to be discarded.

Table 3.

Reliability test results

Latent variable	Measurable variable (Code)	Mean	Standard deviation	Factor load	Cronbach’s Alpha
Economic benefits	X1	3.42	0.758	0.792	0.826
	X2	3.94	1.164	0.743
	X3	3.97	1.325	0.825
Quality of service	X4	3.81	1.417	0.763	0.948
	X5	3.92	1.453	0.849
	X6	3.95	1.215	0.726
	X7	3.93	1.402	0.794
	X8	3.95	1.379	0.858
Informatization level	X9	3.11	1.024	0.737	0.862
	X10	3.10	0.976	0.728
	X11	3.83	1.421	0.794
Ecological benefits	X12	3.98	0.957	0.735	0.837
	X13	3.93	1.203	0.783
	X14	3.84	1.215	0.821
Industry development	X15	3.92	1.248	0.805	0.894
	X16	3.85	1.302	0.837
	X17	3.86	1.294	0.729
Organizational efficiency	X18	3.77	1.259	0.796	0.861
	X19	3.54	1.326	0.747
	X20	3.71	1.137	0.772
Total					0.952

As shown in Table 3, the Cronbach’s alpha coefficient of the rural logistics terminal distribution questionnaire is 0.952>0.7, and the Cronbach’s alpha coefficient α of the questionnaire’s dimensions is greater than 0.7, which indicates that the reliability is overall preferred and the questionnaire is relatively reliable and stable.

Meanwhile, the validity of the questionnaire was tested in this paper and the obtained KMO value was 0.874>0.8, which indicates that the validity passed the test, and the significance probability of the Bartlett’s test of sphericity, Sig. was 0.000<0.05.Therefore, the questionnaire has good validity as a whole.

4.2.3

Evaluation of overall model fitness

In this paper, a structural equation model is constructed based on six latent variables and 20 measurable variables, namely, economic efficiency, informationization level, service quality, industry development, ecological efficiency, and organizational efficiency, and the model is evaluated for fitness. The fitting coefficients are used to assess the fitness of the structural equation model, and then the model fit index is used to evaluate the fit quality of the structural equation model. The results of the model fitness evaluation are shown in Table 4.

Table 4.

Model suitability evaluation results

Fitness indicator	Statistical measure	Recommended value	Actual fitting value	Adaptation evaluation results
Absolute fitting index	CHI/DF	(1,3)	2.357	Ideal
	GFI	>0.9	0.889	Passable
	RMSEA	<0.08	0.068	Qualified
Relative fitting index	IFI	>0.9	0.876	Passable
Relative fitting index	CFI	>0.9	0.893	Passable
Simplicity adaptation index	PGFI	>0.5	0.561	Qualified
Simplicity adaptation index	PNFI	>0.5	0.574	Qualified

As can be seen from Table 4, by comparing the actual fit values with the recommended values of the fitness indicators, the actual fit values of each fitness indicator are greater than or close to the recommended values, while the CHI/DF indicator achieves the desired results, which indicates that the setting of this theoretical model is acceptable.

4.2.4

Analysis of model run results

Using Amos 25.0 as an analytical tool, the model was tested and corrected one by one, and finally the final structural model and the path relationship between the variables were obtained, as shown in Figure 4.

According to the structural equation model in Fig. 4, the path coefficient estimation results of the model can be derived as shown in Table 5. Among them, economic efficiency, informationization level, service quality, industry development, eco-efficiency, and organizational efficiency are denoted by EB1, IL, QS, ID, EB2, and OE, respectively, while X1~X20 denote the corresponding measurable variables. This paper analyzes the estimation results of model path coefficients from two aspects: structural equation and measurement equation. 1)

Table 5 shows that in the structural equation model, there is a path of mutual influence between the six latent variables, that is, there is a correlation between the six factors affecting the terminal distribution of rural logistics: there is a significant correlation between the quality of service and the level of informatization, and ecological efficiency. Organizational efficiency and industry development are also correlated with economic efficiency. In addition, the service quality of rural logistics terminal distribution can also have a positive impact on the economic efficiency of rural logistics terminal distribution, which means that the development of rural logistics can be pushed forward with regard to the above six aspects.

2)

The path coefficients of the measurement model indicate the extent to which each measurable variable can be explained by its latent variable. It can be seen from Table 5 that under the latent variable of economic benefits, X2 employment creation (0.925) has the highest explanatory strength, and it can also be found that rural employment groups are the core force of rural logistics development, which is consistent with the theoretical significance of rural talent revitalization strategy, and X3 agricultural trade (0.792) and X1 industrial development (0.661) can also explain the economic benefits better.

Table 5.

Model path coefficient estimation

Model type	Variable relation	Normalized regression coefficient	The standard error of the estimation parameter	Critical ratio	Significance
Structural equation	QS←EB2	0.457	0.085	1.128	**
	QS←IL	0.958	0.129	11.205	***
	EB1←QS	0.982	0.084	13.247	***
	EBI←OE	0.264	0.213	1.436	**
	EBI←ID	0.323	0.232	1.752	**
Measurement equation	X1←QS	0.661	0.061	8.414	***
	X2←QS	0.925	0.072	12.438	***
	X3←QS	0.792	-	-	-
	X4←QS	0.851	-	-	-
	X5←QS	0.928	0.086	14.203	***
	X6←QS	0.794	0.079	11.251	***
	X7←QS	0.879	0.088	11.987	***
	X8←QS	0.943	0.080	14.125	***
	X9←IL	0.844	-	-	-
	X10←IL	0.796	0.097	11.127	***
	X11←IL	0.883	0.142	12.340	***
	X12←EB2	0.589	-	-	-
	X13←EB2	0.824	0.259	7.256	***
	X14←EB2	0.922	0.288	8.315	***
	X15←ID	0.837	-	-	-
	X16←ID	0.928	0.087	14.023	***
	X17←ID	0.795	0.103	11.208	***
	X18←OE	0.902	-	-	-
	X19←OE	0.774	0.089	11.415	***
	X20←OE	0.808	0.071	12.468	***

Under the latent variable of service quality, X5 on-time delivery (0.928) and X8 customer satisfaction (0.943) have higher explanatory strengths, followed by X7 disinfection and sterilization (0.879), X4 safe delivery (0.851), and X6 door-to-door delivery (0.794), which suggests that the service attributes of rural logistics terminal delivery are gradually becoming more and more important to rural permanent residents. Through the research, it is found that rural residents usually attach great importance to the timeliness of rural logistics terminal delivery, which is due to the fact that the residence is far away from the rural terminal delivery point, and the delivery point does not provide door-to-door service, resulting in the high cost of picking up the package in rural areas, which often makes the rural residents plan the time of picking up according to the information of picking up and picking up several packages in a single pickup, and this pickup strategy does reduce the residents’ pickup time to a certain degree. Such a pickup strategy to a certain extent reduces the cost of residents’ pickup time, but also directly causes the terminal distribution point of warehouse management costs.

In the expression of the latent variable of informatization level, X11 intelligent access (0.883), X9 information updating (0.844) and X10 information feedback (0.796) can make a better measure of the informatization level. Combined with the directional interviews for the practitioners of the rural logistics terminal distribution points, it can be seen that improving the status of the terminal distribution information updating, feedback, and intelligent management can improve the level of informationization of the rural logistics terminal distribution and thus enhance the operational efficiency of the rural logistics system. It can be seen that improving terminal distribution information updating, feedback, intelligent management and other conditions can improve the level of rural logistics terminal distribution informatization, and thus improve the operational efficiency of the rural logistics system.

From the perspective of eco-efficiency dimension, the path coefficient of X14 Green Transportation Planning (0.922) is the largest, which reflects that the planning of eco-logistics distribution paths is particularly important, but considering the complexity of the actual road conditions in rural areas, Green Transportation Planning is still a pain point that is difficult to overcome for practitioners of rural logistics terminal distribution points, and the path coefficients of X13 Packaging Recycling (0.824) and X12 Green Packaging Materials ( 0.589), according to the targeted interviews, it is found that the implementation of the eco-logistics concept has begun to bear fruit, and packaging recycling has been commonly practiced, but some residents still do not have correct knowledge of green packaging materials.

From the perspective of industry development dimension, the path coefficient of X16 Employment prospect (0.928) is the largest, and the coefficient values of X15 Income status (0.837) and X17 Willingness to pursue further education (0.795) are also higher, which reflects the demands of the employees of the rural logistics terminal distribution points.

From the perspective of organizational efficiency dimension, X18 government support (0.902) has the largest path coefficient, which means that for the practitioners at the terminal distribution points, to improve the organizational efficiency of the rural logistics terminal distribution viewpoints, it is reasonable to get support from the government and related policies. X20 Station Management (0.808) and X19 Business Cooperation (0.774) focus on the management and operational capabilities of practitioners at the terminal distribution points. Effective station management requires station operators to rationally arrange station operations, and efficient business cooperation requires station operators to be rooted in the local agricultural products industry to provide logistics support for “agricultural products to the city”. All these are necessary conditions to improve the efficiency of rural logistics terminal distribution.

5

Conclusion and path to optimization

5.1

Conclusion

This paper constructs a rural logistics terminal distribution LRP optimization model based on the mixed mode of delivery to home and self-pickup, optimizes the distribution path using the improved potential field ant colony algorithm, verifies the optimization effect through the arithmetic example, and explores the influencing factors of rural logistics terminal distribution efficiency improvement. The main conclusions obtained are as follows: 1)

As the number of customers increases, the error of the solution gradually increases, and the number of open distribution centers gradually increases. In the case of a certain number of customers, as the number of candidate distribution centers increases, the cost gradually increases, and the error between the optimal solution and the worst solution of the algorithm gradually increases. In the case of the same number of candidate distribution centers and the number of customers, the improved potential field ant colony algorithm is used for solving, which can effectively reduce the distribution cost and optimize the distribution path.

2)

This paper combines the rural revitalization strategy, existing researches, and the relevant needs of rural permanent residents and practitioners of rural logistics terminal distribution points, and identifies 6 latent variables and 20 observable variables that affect the improvement of terminal distribution efficiency, including economic efficiency, informationization level, service quality, industry development, ecological efficiency, and organizational efficiency. Through structural equation modeling, it can be seen that the informationization level and eco-efficiency of logistics terminal distribution have a positive effect on service quality, industry development and organizational efficiency have a positive effect on economic efficiency, and service quality has a significant positive effect on economic efficiency.

5.2

Optimization of pathways

Based on the above conclusions, this paper proposes the following optimization paths to enhance the efficiency of rural terminal distribution: 1)

Integrate resources and improve the quality of terminal services. Based on the principle of “government guidance, enterprise participation, market operation and resource sharing”, a comprehensive logistics service platform with multi-level resource sharing should be established in rural areas to form a “county-village-village” three-level distribution network. The three-level distribution network should be established in rural areas to form a “county-village-village” three-level distribution network, and optimize the station management mode of distribution points, pay attention to the sterilization of parcels, and pay attention to the distribution needs of rural residents.

2)

Optimize facilities and promote the construction of terminal informationization. Construct an information platform for rural logistics terminal distribution, use modern artificial intelligence technology to assist the terminal distribution link, reasonably plan the distribution time and distribution path, so that the information is updated in time and the information feedback is accurate, and mobilize the “market-government-logistics industry” to participate. Multi-party main body participation.

3)

Specialized management and emphasis on terminal personnel training. The career planning of rural logistics terminal distribution point practitioners should be fully considered, and relevant talents should be attracted through industry benefits, while local talents should be developed to realize professional management of logistics terminals by training professionals.

4)

Promote policies emphasizing government support and market cooperation. The government should increase the support of the logistics industry, and from the perspective of the local market, combined with the strategy of rural revitalization and relevant logistics policies, promote the long-term cooperation between rural logistics terminal distribution points and local industries, so as to improve the local industrial chain, and then realize the optimization of the distribution efficiency of rural logistics terminals.

Language:: English

Publication timeframe:: 1 times per year
Journal Subjects:: Life Sciences, Life Sciences, other, Mathematics, Applied Mathematics, General Mathematics, Physics, Physics, other

Journal RSS Feed

Research on Rural Logistics Terminal Distribution Efficiency Improvement in Rural Revitalization Strategy Assisted by Artificial Intelligence

Luping Song

Published Online: Mar 19, 2025

Received: Oct 31, 2024

Accepted: Feb 15, 2025

DOI: https://doi.org/10.2478/amns-2025-0432

KeywordsLRP model, Ant colony algorithm, Artificial potential field algorithm, Structural equation modeling

© 2025 Luping Song, published by Sciendo

This work is licensed under the Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International License.

Keywords
LRP model, Ant colony algorithm, Artificial potential field algorithm, Structural equation modeling