CN114399095A - Dynamic vehicle distribution path optimization method and device based on cloud-edge-device collaboration - Google Patents

Abstract

The invention discloses a dynamic vehicle distribution path optimization method and device based on cloud edge-end cooperation, and relates to the field of spare part logistics distribution. According to spare part requests of all demand positions and distribution resources of enterprises, before distribution starts, a genetic algorithm is used, and according to optimization targets such as distribution timeliness and overall consumed cost, corresponding path planning is conducted on all distribution demands. By the edge computing equipment at the edge of the road, the road condition can be monitored in real time, and whether the passing time of the current road section is changed greatly or not can be judged. The method of the invention adopts the edge end to process the road condition change, and adopts a dynamic passing time table and an improved A-x algorithm according to the actual situation to optimize and adjust the distribution path of the remaining distribution tasks of the vehicle in real time.

Description

Dynamic vehicle distribution path optimization method and device based on cloud-edge-device collaboration

technical field

The invention relates to the field of logistics distribution of industrial spare parts, in particular to a dynamic vehicle distribution path optimization method and device based on cloud-side-end collaboration.

Background technique

Industrial spare parts are the guarantee materials for the normal operation of large-scale machinery and equipment. Effective and timely supply of spare parts must be ensured. Efficient logistics and distribution methods of spare parts are of great significance to improving the economic benefits of enterprises.

Rapid response to the demand for industrial spare parts is a key factor to ensure stable production of enterprises. Therefore, it is necessary to effectively shorten the stagnation time of spare parts in the logistics system through reasonable planning of logistics distribution paths, thereby reducing logistics costs and achieving economic benefits. maximization.

With the rapid development of cloud computing, Internet of Things, and e-commerce, logistics distribution is combined with new technologies such as cloud computing, Internet of Things, and intelligent transportation, which is different from the cloud distribution model of the traditional logistics distribution model. The cloud distribution mode solves the problem that the traditional logistics distribution mode is difficult to adapt to the needs of modern logistics distribution. However, with the rapid growth of the number of various mobile devices and computing demands, the traditional cloud computing centralized processing mode with cloud data centers as the core faces problems such as large network transmission delay, high data transmission cost, computing security and privacy risks. This is unable to effectively meet the demands of mobile users, especially users who need instant response, for computing services. Cloud collaboration offers a new approach to this problem.

As people's requirements for the timeliness of logistics and distribution are getting higher and higher, how to ensure that various logistics and distribution tasks are completed on time has become one of the key issues to effectively save distribution costs, and it is also an urgent need to solve the current logistics and transportation industry. Through the analysis of each link of vehicle delivery, the delivery route of the vehicle is optimized, and the delivery route is dynamically adjusted after the departure of the delivery vehicle, considering that uncertain factors such as changes in road conditions may affect the timeliness of delivery. Reducing the impact of road conditions, weather and other factors on the overall delivery timeliness through optimization is an important means to improve logistics distribution efficiency and reduce overall distribution costs. Due to several constraints and uncertainties in the logistics distribution of spare parts, it is particularly important to comprehensively consider the impact of dynamic uncertain factors such as traffic, weather, and road conditions on the distribution process.

An edge node contains computing equipment and communication equipment, which can process a certain range of computing tasks and return the results to a specific location. Therefore, edge devices can be used to judge changes in road condition information, and timely adjust delivery routes based on these changes.

The existing technologies mainly include the following: improving the shortcomings of the traditional genetic algorithm's early convergence time, from the traditional genetic algorithm's lack of local search ability, optimizing the search ability for high-quality solutions by integrating the hill-climbing algorithm, using fuzzy The mathematical evaluation method quantifies the road risk, and considers the factors affecting the choice of transportation routes from the perspective of mathematical estimation.

By analyzing the prior art, it is found that there are some deficiencies as follows. The current optimization is mostly in the path planning stage before the delivery starts, by improving the inherent defects of the algorithm itself, or by considering the static fixed road conditions, to improve the rationality of the planned delivery path. However, factors such as emergencies in the period from the start of the delivery to the completion of the delivery were not considered and dealt with. However, these unpredictable sudden changes in road conditions will have varying degrees of impact on the delivery effect. Therefore, if the delivery route of the vehicle is not dynamically adjusted according to the change of the actual road conditions during the delivery process, it will lead to an increase in delivery cost and delivery time.

SUMMARY OF THE INVENTION

The technical problem to be solved by the present invention is to provide a method and device for dynamic vehicle distribution path optimization based on cloud-side-end collaboration, in view of the deficiencies in the prior art, so as to reduce the risk that the time spent on completing the distribution task is greatly increased due to changes in road conditions.

In order to solve the above-mentioned technical problems, the technical solution adopted in the present invention is: a method for optimizing a dynamic vehicle distribution path based on cloud-side-terminal collaboration, comprising the following steps:

S1. Qualitatively analyze the distribution vehicle path planning problem, and determine the distribution time and distribution cost as the optimization goals of the first-stage path planning;

S2. Under the constraints of the demand for goods and the time window proposed according to the demand point, considering the distribution capacity of the distribution center, before the vehicle departs, according to the optimization goal of the first stage path planning, use the genetic algorithm to carry out the overall distribution path in the cloud in advance. optimization, and obtain the initial distribution path planning scheme, the overall distribution time of the initial distribution path planning scheme is the least, and the distribution cost is the lowest;

S3. After the delivery vehicle departs, the road condition data is sensed and judged, and the road network information of the corresponding road section in the initial delivery route planning scheme is updated;

S4. According to the dynamically updated road network information and the location of the delivery vehicle, dynamically adjust the remaining delivery routes until all delivery tasks are completed, and calculate the delivery time;

S5. Optimizing and adjusting the route according to changes in road conditions, and updating the cost of the current updated delivery route;

S6. Taking the distribution center as the starting point of the vehicle, and ending with the last delivery vehicle completing the delivery task, in the entire delivery process, dynamically adjust the delivery vehicle path in the entire delivery area according to steps S4 and S5.

Aiming at the deficiencies of the prior art, the present invention provides a dynamic vehicle distribution path optimization method based on cloud-edge-terminal collaboration. Through the collaborative cooperation of cloud, edge and terminal, the road conditions are processed to dynamically adjust the vehicle distribution path. The influence of various influencing factors in the distribution process on the overall distribution effect is reduced, thereby ensuring the time spent in distribution and saving distribution costs. Before the delivery starts, according to the distribution resources and distribution requirements, the genetic algorithm is used in the cloud to make the first-stage planning of the distribution path. After the delivery starts, the processing capability of the edge terminal can sense the changes in road conditions in real time and make judgments. Based on this, the remaining delivery routes can be dynamically optimized and adjusted, and sent to the vehicle terminal. Through the cooperation between the cloud, the side and the terminal, various dynamic factors affecting the road traffic conditions after the start of distribution will be well processed, and the vehicle distribution path will be dynamically optimized, which will improve the rationality of the overall distribution plan and ensure that Reduced delivery time and reduced distribution costs.

In step S1, the delivery cost is the sum of the vehicle transportation cost TC, the time window penalty cost PC and the vehicle cost; wherein,

Among them, K is the total number of vehicles, N is the total number of demand points to be visited, c _ij is the unit transportation cost between demand point i and demand point j, and d _ij is the distance between demand point i and demand point j , the value of x _ijk is 0 or 1, the value of 1 means that the vehicle k leaves the demand point i and goes to the demand point j, otherwise the value is 0; a and b are the delivery time window penalty coefficients, and w _ik is the vehicle k Waiting time at demand point i, t _ik is the time when vehicle k arrives at delivery point i, l _i is the latest service time window of demand point i; x _0jk is 0 or 1, when its value is 1, it means that k departs from distribution center 0 to demand point j.

By analyzing the required vehicle distribution route planning problem, a corresponding mathematical model is constructed, and the abstract concept becomes a concrete model, which can be measured and evaluated by numerical indicators. By clarifying the composition of each part of the cost in the distribution process, and establishing a mathematical formula accordingly, it provides a direct basis for multi-objective optimization.

The specific implementation process of step S2 includes:

其中，x代表相应的种群个体编号，T是正数，D_{u_max}表示第u种车型的最大行驶距离；N为待访问的需求点总数目；1) Construct the penalty function p(x):

Among them, x represents the individual number of the corresponding population, T is a positive number, D _{u_max} represents the maximum driving distance of the u-th vehicle type; N is the total number of demand points to be visited;

2) decoding the penalty function p(x), and constructing a chromosome corresponding to the penalty function p(x);

3) Randomly generate initial populations according to spare parts demand data, road network data and distribution resources;

4) Through selection, crossover and mutation operators, cyclic evolution is performed on the initial population if the termination conditions are not met, until the optimal solution is generated, and the initial distribution path planning scheme is obtained.

By using the genetic algorithm, on the one hand, the time of the optimization process is controlled within a reasonable range; is optimal. The construction of the penalty function can ensure that in the iterative optimization process, each individual can obtain the corresponding genetic probability according to its own advantages and disadvantages, and participate in the selection, crossover and mutation operator stages with an appropriate probability, thus ensuring the algorithm's performance. Convergence.

The specific implementation process of step S3 includes:

A) Calculate the change of the travel time of the road section to which the edge node belongs according to the road condition change information collected by each edge node, and update the current travel time of each road section in the road traffic schedule according to the calculated result;

B) Update the current position of the delivery vehicle according to the vehicle speed, and combine with the position of the next demand point to be headed to, obtain the road section that the travel route in the initial delivery route planning scheme needs to pass between the two positions, and complete the initial The update of the road network information of the corresponding road section in the delivery route planning scheme;

C) According to the current location of all delivery vehicles and the next delivery location to go to, query the road traffic schedule, and determine whether the changed road condition information will have a bad impact on the current delivery tasks of each delivery vehicle: if the vehicle is about to go to the road section changes The later road travel time is greater than N times the original travel time (N can be set according to the actual situation), that is, it is considered that the change of the road condition information has caused a bad impact on the delivery task, and the process goes to step S4.

By judging the changes of road conditions, it is clear which road conditions changes will have a bad impact on the delivery results, and can adapt to different situations by changing the threshold (that is, the size of N), which ensures the sensitivity of the proposed method.

The specific implementation process of step S4 includes:

I) Take the node with the smallest value of f(n) as the next node on the optimal path, f(n)=g(n)+h(n); g(n) is the actual passage from the starting node to the current node cost, h(n) is the estimated value of the travel cost from the current node to the end point;

II) Operate the P table and Q table maintained by the A* algorithm, including:

i) The P table and Q table are empty, the starting point S is added to the P table, its g(n) value is set to 0, the parent node is empty, and the g(n) value of other nodes in the road network is set to infinity;

ii) If the P table is empty, the algorithm fails, otherwise select the node with the smallest f(n) value in the P table, denote it as BT, and add it to the Q table; judge whether BT is the end point T, if so, go to step iii );

Otherwise, find each adjacent node NT of BT according to the road network topology attributes and traffic rules, and perform the following steps:

①Calculate the heuristic value of NT

f(NT)=g(NT)+h(NT);

g(NT)=g(BT)+cost(BT,NT);

Among them, cost(BT, NT) is the traffic cost from BT to NT;

②If NT is in the P table, and the toll cost value calculated by the formula g(NT)=g(BT)+cost(BT,NT) is smaller than that of NT, then update the toll cost value of NT to g (NT)=g(BT)+cost(BT, NT), and set the parent node of NT as BT;

③ If NT is in the Q table, and the travel cost value calculated by g(NT)=g(BT)+cost(BT,NT) is smaller than the travel cost value of NT, then update the travel cost value of NT to g( NT)=g(BT)+cost(BT, NT), set the parent node of NT as BT, and move NT out to the P table;

④ If NT is neither in the P table nor in the Q table, set the parent node of NT as BT, and move NT to the P table;

⑤ Return to step ii);

iii) Backtracking from the end point T, find the parent node in turn, and add it to the optimization path until the start point S, that is, the optimization path is obtained;

利用下式计算车辆通过路段k的通行时间Tk：III) Calculate the travel time required for the vehicle to pass through the optimized path. The optimized path includes multiple road segments, and the multiple road segments are numbered as 1, 2, 3, ..., k; [tk, tk'] indicates that the vehicle passes through The travel time Tk of the road segment _k , then Tk _{= tk'} - _tk ; the travel time spent by the vehicle passing through multiple road segments corresponds to _T'k1 , _T'k2 , _T'k3 ...; _fk represents The time period corresponding to the moment when the vehicle passes the starting point of road segment k,

Use the following formula to calculate the travel time Tk of the vehicle passing through the road segment k:

The value of m satisfies the following constraints:

ΔT represents the length of the period, and t ₀ represents the start time;

IV) Use the following formula to calculate the time it takes to process changes in road conditions: g _n (r)=b _n (r)+c _n (r)+k _n (r)+h _n (r);

θ _{n, n'} represents the number of vehicles through which data is transmitted from vehicle V _n to vehicle V _n' ;

M represents the number of edge nodes, D represents the set of edge computing devices, D={d ₁ , d ₂ , ..., d _m }, V represents the set of vehicles, V={v ₁ , v ₂ , ..., v _n }, λ _V2V represents the data transmission rate based on the V2V method, λ _V2I represents the data transmission rate based on the V2I method, ω _n represents the data amount of the transmitted computing task, v _n represents the delivery vehicle numbered n, and d _m represents the number of m edge computing device, ln indicates whether computing task _n is processed on an edge computing device, ln is 0 or 1, _p indicates the processing capability of each edge computing device, u _n is the request of computing task n the amount of resources;

V) Use the following formula to calculate the time required to complete the overall delivery task:

K表示编号为r_s的路径所经过的路段总数；R表示至完成配送时，所处理过的由路况变化而对道路通行时间造成恶劣影响的道路事件总数。

K represents the total number of road segments passed by the route numbered r_s; R represents the total number of road events that have been processed by the change of road conditions and have a bad impact on the road transit time until the delivery is completed.

The improved A* algorithm is used to re-plan the local paths that need to be adjusted, which not only ensures the solution speed of re-optimization, controls the calculation time within a certain range, but also makes the results of re-optimization conform to the actual situation, ensuring that all Reasonableness to make adjustments. The contradiction between the time consumption and the quality of the solution is reconciled as a whole. Using V2I and V2V methods ensures the transmission speed and transmission efficiency of road condition data and adjustment results between edge nodes and between edge nodes and vehicles, ensuring the reliability of traffic.

The specific implementation process of step S5 includes:

a) All delivery vehicles start delivery according to the initial route planning;

b) Divide the demand points that need to be distributed into multiple distribution areas according to r*r, and number each of the distribution areas, and r is the unit length;

c) The edge nodes monitor changes in road conditions, and record the events that affect road conditions in the events list according to the length of time. Each event in the events list has different sizes for the travel time of the road leading to the demand point in each delivery area. The influence of, that is, according to the influence degree of the event type corresponding to each event, the road travel time of the road section in the distribution area where the event occurs is increased to the original e _i times (e _i can be set according to actual use needs);

d) Change the road travel time of the road section in the corresponding delivery area to the affected value;

e) Judging whether the original planned route needs to be adjusted in the current time segment, that is, judging whether the changed road condition information will have a bad impact on the current delivery task of each delivery vehicle: if the vehicle is about to go to the road section after the change, the road travel time after the change is greater than the original N times the travel time (N can be set according to the actual situation), that is, it is considered that the change of the road condition information has caused a bad impact on the delivery task, and it needs to be adjusted; re-planning;

f) For each other time segment, repeat steps a) to f) until all distribution tasks are completed, that is, all vehicles have traversed all the demand points in the initial planning and return to the distribution center.

By dividing the distribution area, the monitoring range of each edge device is clarified, and the calculation difficulty of road condition processing is reasonably reduced, so that each edge device can perceive the changes in road conditions in their respective responsible areas and process them without As for the computing power beyond the edge devices, the computing load is reasonably shared.

The present invention also provides a computer device, comprising a memory, a processor and a computer program stored in the memory; it is characterized in that the processor executes the computer program to implement the steps of the method of the present invention.

The present invention also provides a computer program product, comprising a computer program/instruction; when the computer program/instruction is executed by a processor, the steps of the method of the present invention are implemented.

The present invention also provides a computer-readable storage medium on which computer programs/instructions are stored; when the computer programs/instructions are executed by a processor, the steps of the method of the present invention are implemented.

Compared with the prior art, the present invention has the following beneficial effects: the present invention takes into account many factors that affect road traffic conditions in the distribution area during the period from the start of the distribution to the completion of the distribution, and reduces the number of completed distribution. The risk of the time spent on the task is greatly increased due to changes in road conditions, which reduces the cost of distribution, brings economic benefits to the logistics distribution of the enterprise, and reduces energy consumption.

Description of drawings

1 is a flowchart of a method according to an embodiment of the present invention;

FIG. 2 is a schematic diagram of a cloud communication process according to an embodiment of the present invention.

Detailed ways

The embodiment of the present invention includes the following steps:

Step 1: Qualitatively analyze the distribution vehicle path planning problem, and determine the distribution time and cost as the optimization goals of the first-stage path planning. The distribution cost consists of three parts: vehicle transportation cost TC, time window penalty cost PC and vehicle cost.

Among them, TC includes transportation fuel consumption and vehicle maintenance costs, etc. This part is proportional to the length of the distribution path, and the calculation method is as follows:

c _ij is the unit transportation cost between demand point i and demand point j, d _ij is the distance between demand point i and demand point j, x _ijk is 0 or 1, and a value of 1 means that vehicle k from After the demand point i leaves, go to the demand point j, otherwise the value is 0. PC represents the penalty cost that needs to be increased when the vehicle exceeds the time window requirement. The calculation method is as follows:

a and b are the delivery time window penalty coefficients, w _ik is the waiting time of vehicle k at delivery point i, t _ik is the time when vehicle k arrives at delivery point _i , and li is the latest service time window of delivery point i.

The vehicle cost is proportional to the number of scheduled delivery vehicles, VN, and is calculated as follows:

The value of x _0jk is 0 or 1. When the value is 1, it means that the vehicle k departs from the distribution center 0 to the demand point j.

Step 2: Under the constraints of the demand for goods and the time window proposed according to the demand point, and considering the distribution capacity of the distribution center, before the vehicle departs, according to the existing data, use the genetic algorithm to optimize the overall distribution route in the cloud in advance.

Step 2.1: Formulate a multi-objective optimization function according to the requirements of the time window and the lowest total distribution cost;

This method uses the penalty function method to deal with the constraints. The basic idea of the penalty function method is: for an individual whose solution is infeasible in the solution space, if its fitness is to be calculated, it is necessary to assign a penalty function to the fitness function of the individual to reduce the fitness, so that the probability of inheritance to the next generation is made. reduced, and even automatically eliminated. According to the characteristics of the model in this paper, the vehicle capacity constraints and the maximum travel distance constraints can be implemented in the form of penalty functions. The formula is expressed as follows:

In the formula, p(x) is the penalty function, x represents the corresponding population individual number, and T is a large positive number.

Step 2.2: Encode the solution of the problem described in the previous step, and construct the corresponding chromosome;

The method uses natural number symbols to encode the solution to the problem. Natural number encoding By encoding the solution of the problem (ie, the set of vehicle paths) as an array of natural numbers of length k+m+1, each encoded chromosome represents an initial solution in the solution space of the problem. An example is as follows: Consider {0, 4, 7, 9, 11, 0, 1, 3, 5, 2, 0, 6, 8, 10, 0} chromosome coding, indicating that three vehicles are arranged to complete 11 spare parts demand points The distribution tasks of the three vehicles are: path 1: 0->4->7->9->11->0; path 2: 0->1->3->5->2 ->0; path 3: 0->6->8->10->0; among them, the number 0 represents the distribution center, 1-11 represents the number of 11 demand points, each vehicle starts from the distribution center, and then Return to the distribution center after completing the delivery task.

Step 2.3: Randomly generate initial populations according to spare parts demand data, road network data and distribution resources;

Step 2.4: Through selection, crossover and mutation operators, cyclic evolution is performed on the initial population without satisfying the termination conditions until the optimal solution is generated;

When applying this method to solve the problem of vehicle distribution path planning for spare parts, the parameters are set as: population size Popsize=80, maximum iteration algebra Maxgen=500, crossover rate Pc=0.88, and mutation rate Pm=0.05. The termination conditions are set as follows: if the number of iterations reaches 500 generations; or if the fitness value of a certain generation of chromosomes reaches 0.9 times the initial optimal chromosome fitness value, the search is terminated and the optimal solution is output.

Step 2.5: Obtain the initial distribution route planning scheme through decoding.

Step 3: After the delivery vehicle departs, collect data related to changes in road conditions through the edge device: The edge device senses and makes judgments on the road condition data, updates the road network data of the corresponding road section, and provides dynamic adjustment of local delivery routes. in accordance with.

Step 3.1: Process the road condition change information collected by the edge device, and calculate the change of the travel time of the road section to which it belongs;

Step 3.2: Dynamically update the road network data and the traffic schedule of the road. Taking 0.5h as the unit time ΔT, update the current position of the delivery vehicle according to the vehicle speed, and update the current travel time of all road sections in the road traffic schedule according to the data collected and processed from the edge nodes;

Step 3.3: Use the edge device to judge the road condition change information and the current location and next delivery location of all delivery vehicles to provide a basis for the next adjustment.

Step 4: Dynamically adjust the predetermined path based on the improved A* algorithm: By running the improved A* algorithm on the edge device, according to the dynamically updated road network information provided by the edge device and the location of the delivery vehicle, the remaining The delivery route is dynamically adjusted until all delivery tasks are completed.

Step 4.1: When it is determined that the driving route needs to be changed between the current delivery point and the next delivery point (in the original route planning), find the shortest path between the two points to realize it. Use the A* algorithm to re-plan the paths of the current demand point and the next demand point that need to be adjusted, that is, to find the shortest path between the two points. The specific implementation is as follows:

A* algorithm is a typical heuristic search algorithm, which is based on Diikstra's algorithm and is widely used to find the shortest path between two points.

The most important thing about the A* algorithm is to maintain a heuristic evaluation function:

f(n)=g(n)+h(n)

Among them, f(n) is the corresponding heuristic function when the algorithm searches for each node. It consists of two parts, the first part g(n) is the actual travel cost from the starting node to the current node, and the second part h(n) is the estimated value of the travel cost from the current node to the end point. Each time the algorithm expands, the node with the smallest f(n) value is selected as the next node on the optimal path.

Step 4.2: Operate the P table and Q table maintained by the A* algorithm. The A* algorithm maintains two sets: the P table and the Q table. The P table stores those nodes that have been searched but not yet added to the optimal path tree; the Q table maintains those nodes that have been added to the optimal path tree.

(1) The P table and Q table are empty, the starting point S is added to the P table, its g value is set to 0, the parent node is empty, and the g value of other nodes in the road network is set to infinity.

(2) If the P table is empty, the algorithm fails. Otherwise, select the node with the smallest f value in the P table, record it as BT, and add it to the Q table. Determine whether BT is the end point T, if so, go to step (3); otherwise, find each adjacent node NT of BT according to road network topology attributes and traffic rules, and perform the following steps:

①Calculate the heuristic value of NT

f(NT)=g(NT)+h(NT)

g(NT)=g(BT)+cos t(BT,NT)

Among them, cost(BT, NT) is the pass cost from BT to NT.

②If NT is in the P table, and the g value calculated by g(NT)=g(BT)+cos t(BT,NT) is smaller than the original g value of NT, then update the g value of NT to formula (3 ) result, and set the parent node of NT to BT.

③ If NT is in the Q table, and the g value calculated by g(NT)=g(BT)+cos t(BT,NT) is smaller than the original g value of NT, then update the g value of NT to g(NT )=g(BT)+cos t(BT,NT) As a result, the parent node of NT is set to BT, and NT is moved out of the P table.

④ If NT is neither in P table nor in Q table, set the parent node of NT as BT, and move NT to P table.

⑤ Go to step (2) to continue.

(3) Backtrack from the end point T, find the parent nodes in turn, and add them to the optimization path until the start point S, and then the optimization path can be obtained.

Step 4.3: Calculate and process the dynamics of transit time. In order to calculate the travel time of a certain road in real time, a structure of road travel schedule is adopted, as shown in Table 1. The table stores the travel time of the road at the current moment and the predicted value of the travel time in the next few moments.

The starting time is represented by t ₀ , the future period of time is divided into several time periods, and ΔT is used to represent the length of a time period, and the moment when the system starts to work belongs to the first time period. These periods are then numbered like 1, 2, 3, 4, .... Similarly, also number each road segment as 1, 2, 3, 4, .... T _ij is used to denote the travel time of road segment i in time period j. In this way, the travel time of different road sections at different times can be obtained.

Table 1 Road dynamic traffic schedule

The re-planning system will determine whether the current delivery route needs to be changed when the road conditions change, that is, when different road events occur. If the changed road travel time is greater than N times the original travel time, it means that route re-planning is required. The optimized path may contain multiple segments, number them 1, 2, 3, ..., k, .... By [t _k , t _k ′] representing the travel time T _k of the vehicle passing through the road section k, then T _k =t _k ′-t _k . A vehicle may take several time periods to pass through the road segment k, and these time periods correspond to the travel times T′ _k1 , T′ _k2 , T′ _k3 , . . .

First, calculate the time period f _{k corresponding to the moment when the vehicle passes the starting point of road segment k} :

Correspondingly, it can be concluded that:

According to the different values of the period length ΔT, the road length L and the road speed, it may happen that the vehicle only needs to pass through the road section in one period, or it may take multiple periods to pass. Therefore, the specific formula for the vehicle passing through the road segment k can be obtained as follows:

Among them, the value of m satisfies the following constraints:

Step 5: The computing edge device processes the road information change, and transmits the result to the time of the destination node through the communication network between vehicles in the area.

Step 5.1: Calculate the time required for the transmission of road condition change information to the edge server as b _n (i). The calculation method is as follows:

Among them, N represents the number of vehicles in the current section, M represents the number of edge computing devices erected, D represents the set of edge computing devices, D={d ₁ , d ₂ , ..., d _m }, V represents the set of vehicles, V={v ₁ , v ₂ , ..., v _n }, λ _V2V represents the data transmission rate based on V2V technology, λ _V2I represents the data transmission rate based on V2I technology, ω _n represents the data amount of the transmitted computing task, and v _n represents the number is the delivery vehicle of n, and d _m represents the edge computing device numbered m;

Step 5.2: The time when the computing task is sent to the edge server is c _n (i), and the calculation method is as follows:

Step 5.3: Calculate the time to perform path re-planning at the edge server as k _n (i). The calculation method is as follows:

Step 5.4: Calculate the time to send the replanned path back to the vehicle as h _n (i), calculated as follows:

Step 5.5: Find the overall time spent. The calculation method is as follows:

g _n (i)=b _n (i)+c _n (i)+k _n (i)+h _n (i)

Step 6: According to the optimization adjustment made to the route according to the change of road conditions, the cost of updating the updated delivery route.

Step 7: The vehicle starts from the distribution center, and the last delivery vehicle completes the delivery task before it ends. During the entire delivery process, according to the methods described in steps 3, 4, and 5, the routes of the delivery vehicles in the entire delivery area are analyzed. Make dynamic adjustments. For the convenience of processing, before implementing the following steps, we propose the following two assumptions:

(1) The logistics distribution center and the customer nodes and between the customer nodes are all reachable through the combination of roads in the road network;

(2) The change of road conditions occurs in a grid area, and within this time period, the travel time of the road leading to the demand point in the area will be affected to varying degrees (depending on the specific event type).

Step 7.1: All delivery vehicles start delivery according to the initial route planning;

Step 7.2: Divide and number the demand points to be distributed according to r*r, where r is the unit length, and the specific value can be determined according to the computing power of the edge device;

Step 7.3: Changes in road conditions are monitored by edge nodes, and events that affect road conditions are recorded in the events list according to segments with a time length of 0.5h. Influence of different sizes: according to the degree of influence of the type of event corresponding to each event, increase the road travel time of the road section in the area where the event occurs by e _i times the original (e _i can be set according to the actual situation) , and the road events that are specifically considered are listed in Table 2;

Table 2 Road Incident Table

Step 7.4: Change the road travel time of the road section in the corresponding area to the affected value;

Step 7.5: Determine whether the original planned route needs to be adjusted in this time segment, that is, determine whether the changed road condition information will have a bad impact on the current distribution task of each distribution vehicle: if the vehicle is about to go to the road section after the change of the road travel time is greater than N times of the original travel time, that is, it is considered that the change of the road condition information has caused a bad impact on the delivery task, and it needs to be adjusted;

Step 7.6: If adjustment is required, re-plan according to the A* algorithm that finds the shortest path between two points; do nothing without adjustment;

Step 7.7: Repeat the above steps according to the 0.5h time segment until all the distribution tasks are completed, that is, all vehicles have traversed all the distribution points in the initial planning and return to the distribution center.

Claims (9)

1. a dynamic vehicle distribution path optimization method based on cloud-edge-terminal collaboration, is characterized in that, comprises the following steps: S1. Qualitatively analyze the distribution vehicle path planning problem, and determine the distribution time and distribution cost as the optimization goals of the first-stage path planning; S2. Under the constraints of the demand for goods and the time window proposed according to the demand point, considering the distribution capacity of the distribution center, before the vehicle departs, according to the optimization goal of the first stage path planning, use the genetic algorithm to carry out the overall distribution path in the cloud in advance. optimization, and obtain the initial distribution path planning scheme, the overall distribution time of the initial distribution path planning scheme is the least, and the distribution cost is the lowest; S3. After the delivery vehicle departs, the road condition data is sensed and judged, and the road network information of the corresponding road section in the initial delivery route planning scheme is updated; S4. According to the dynamically updated road network information and the location of the delivery vehicle, dynamically adjust the remaining delivery routes until all delivery tasks are completed, and calculate the delivery time; S5. Optimizing and adjusting the route according to changes in road conditions, and updating the cost of the current updated delivery route; S6. Taking the distribution center as the starting point of the vehicle, and ending with the last delivery vehicle completing the delivery task, in the entire delivery process, dynamically adjust the delivery vehicle path in the entire delivery area according to steps S4 and S5. 2. The dynamic vehicle distribution path optimization method based on cloud-side-terminal collaboration according to claim 1, wherein in step S1, the distribution cost is the sum of vehicle transportation cost TC, time window penalty cost PC and vehicle cost; wherein ,

Among them, K is the total number of vehicles, N is the total number of demand points to be visited, c _ij is the unit transportation cost between demand point i and demand point j, and d _ij is the distance between demand point i and demand point j , the value of x _ijk is 0 or 1, the value of 1 means that the vehicle k leaves the demand point i and goes to the demand point j, otherwise the value is 0; a and b are the delivery time window penalty coefficients, and w _ik is the vehicle k Waiting time at demand point i, t _ik is the time when vehicle k arrives at delivery point i, l _i is the latest service time window of demand point i; x _0jk is 0 or 1, when its value is 1, it means that k departs from distribution center 0 to demand point j. 3. The dynamic vehicle distribution path optimization method based on cloud-side-terminal collaboration according to claim 1, wherein the specific implementation process of step S2 comprises: 1) Construct the penalty function p(x):

Among them, x represents the individual number of the corresponding population, T is a positive number, D _{u_max} represents the maximum driving distance of the u-th vehicle type; N is the total number of demand points to be visited; 2) decoding the penalty function p(x), and constructing a chromosome corresponding to the penalty function p(x); 3) Randomly generate initial populations according to spare parts demand data, road network data and distribution resources; 4) Through selection, crossover and mutation operators, cyclic evolution is performed on the initial population under the condition that the termination conditions are not met, until the optimal solution is generated, and the initial distribution path planning scheme is obtained. 4. The dynamic vehicle distribution path optimization method based on cloud-side-terminal collaboration according to claim 1, wherein the specific implementation process of step S3 comprises: A) Calculate the change of the travel time of the road section to which the edge node belongs according to the road condition change information collected by each edge node, and update the current travel time of each road section in the road traffic schedule according to the calculated result; B) Update the current position of the delivery vehicle according to the speed of the vehicle, and combine with the position of the next demand point to be headed to, obtain the road section that the travel route in the initial delivery route planning scheme needs to pass between the two positions, and complete the initial The update of the road network information of the corresponding road section in the delivery route planning scheme; C) According to the current location of all delivery vehicles and the next delivery location to be headed to, query the road traffic schedule, and determine whether the changed road condition information will have a bad impact on the current delivery tasks of each delivery vehicle: if the vehicle is about to go to the road section changes The later road travel time is greater than N times the original travel time, that is, it is considered that the change of the road condition information has caused a bad impact on the delivery task, and the process proceeds to step S4. 5. The dynamic vehicle distribution path optimization method based on cloud-side-terminal collaboration according to claim 1, wherein the specific implementation process of step S4 comprises: I) Take the node with the smallest value of f(n) as the next node on the optimal path, f(n)=g(n)+h(n), g(n) is the actual passage from the starting node to the current node cost, h(n) is the estimated value of the travel cost from the current node to the end point; II) Operate the P table and Q table maintained by the A* algorithm, including: i) The P table and Q table are empty, the starting point S is added to the P table, its g(n) value is set to 0, the parent node is empty, and the g(n) value of other nodes in the road network is set to infinity; ii) If the P table is empty, the algorithm fails, otherwise select the node with the smallest f(n) value in the P table, denote it as BT, and add it to the Q table; judge whether BT is the end point T, if so, go to step iii ); otherwise, find each adjacent node NT of BT according to the road network topology attributes and traffic rules, and perform the following steps: ①Calculate the heuristic value of NT f(NT)=g(NT)+h(NT); g(NT)=g(BT)+cost(BT,NT); Among them, cost(BT, NT) is the traffic cost from BT to NT; ②If NT is in the P table, and the traffic cost calculated by the formula g(NT)=g(BT)+cost(BT,NT) is smaller than the traffic cost of NT, then Update the passable cost value of NT to g(NT)=g(BT)+cost(BT, NT), and set the parent node of NT as BT; ③ If NT is in the Q table, and the pass cost value calculated by g(NT)=g(BT)+cost(BT,NT) is smaller than the pass cost value of NT, then update the pass cost value of NT to g( NT)=g(BT)+cost(BT, NT), set the parent node of NT as BT, and move NT out to the P table; ④If NT is neither in P table nor in Q table, then set the parent node of NT to BT, and move NT to P table; ⑤ Return to step ii); iii) Backtracking from the end point T, find the parent node in turn, and add it to the optimization path until the start point S, that is, the optimization path is obtained; III) Calculate the travel time required for the vehicle to pass through the optimized path. The optimized path includes multiple road segments, and the multiple road segments are numbered as 1, 2, 3, . . . , k; represented by [t _k , t _k′ ] The travel time Tk of the vehicle passing through the road section _k , then Tk _{= tk′} - _tk ; the travel time spent by the vehicle passing through multiple road sections corresponds to _T′k1 , _T′k2 , _T′k3 . . . f _k represents the time period corresponding to the time when the vehicle passes the starting point of road segment k,

Use the following formula to calculate the travel time T _{k of the vehicle passing through the road segment k} :

The value of m satisfies the following constraints:

ΔT represents the length of the period, and t ₀ represents the start time; IV) Use the following formula to calculate the time it takes to process changes in road conditions: g _n (r)=b _n (r)+c _n (r)+k _n (r)+h _n (r);

θ _{n, n'} represents the number of vehicles through which data is transmitted from vehicle V _n to vehicle V _n' ;

M represents the number of edge nodes, D represents the set of edge computing devices, D={d ₁ , d ₂ , ..., d _m }, V represents the set of vehicles, V={v ₁ , v ₂ , ..., v _n }, λ _V2V represents the data transmission rate based on the V2V method, λ _V2I represents the data transmission rate based on the V2I method, ω _n represents the data amount of the transmitted computing task, v _n represents the delivery vehicle numbered n, and d _m represents the number of m edge computing device, ln indicates whether computing task _n is processed on an edge computing device, ln is 0 or 1, _p indicates the processing capability of each edge computing device, u _n is the request of computing task n number of resources. V) Use the following formula to calculate the time required to complete the overall delivery task:

K represents the total number of road segments passed by the route numbered r_s; R represents the total number of road events that have been processed by the change of road conditions and have a bad impact on the road transit time until the delivery is completed. 6. The dynamic vehicle distribution path optimization method based on cloud-side-terminal collaboration according to claim 1, wherein the specific implementation process of step S5 comprises: a) All delivery vehicles start delivery according to the initial route planning; b) Divide the demand points that need to be distributed into multiple distribution areas according to r*r, and number each of the distribution areas, and r is the unit length; c) The edge nodes monitor changes in road conditions, and record the events that affect road conditions in the events list according to the length of time. Each event in the events list has different sizes for the travel time of the road leading to the demand point in each delivery area. Impact; d) Change the road travel time of the road section in the corresponding delivery area to the affected value; e) Judging whether the original planned path needs to be adjusted in the current time segment; if adjustment is required, re-planning is performed according to the A* algorithm for finding the shortest path between two points; f) For each other time segment, repeat steps a) to f) until all distribution tasks are completed, that is, all vehicles have traversed all the demand points in the initial planning and return to the distribution center. 7. A computer device comprising a memory, a processor and a computer program stored in the memory; characterized in that, the processor executes the computer program to implement the steps of the method according to any one of claims 1 to 6. 8. A computer program product, comprising a computer program/instruction; characterized in that, when the computer program/instruction is executed by a processor, the steps of the method according to any one of claims 1 to 6 are implemented. 9. A computer-readable storage medium on which computer programs/instructions are stored; characterized in that, when the computer programs/instructions are executed by a processor, the steps of the method according to any one of claims 1 to 6 are implemented.

Images