A low-carbon dynamic path planning method and system for a garbage vehicle
By employing Q-learning hyperheuristic particle swarm optimization and dynamic response mechanisms, the problem of low-cost and low-carbon emissions under dynamic perturbations in waste collection route planning was solved, achieving rapid and effective route optimization.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- NANJING UNIV OF INFORMATION SCI & TECH
- Filing Date
- 2022-11-02
- Publication Date
- 2026-06-26
Smart Images
Figure CN116415745B_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of vehicle scheduling technology, specifically a method and system for dynamic route planning of garbage trucks. Background Technology
[0002] Many countries have begun implementing Integrated Solid Waste Management (ISWM). ISWM encompasses waste generation, resource separation, storage, collection, and treatment. Waste collection, as a crucial link between front-end waste generation and end-stage waste treatment, acts as a "bridge." It can be described as vehicles departing from depots, following specific routes to collect waste from all disposal points, a process that incurs significant costs. Given the various problems caused by current urban waste, and based on both environmental and economic considerations, the rational scheduling of waste collection vehicles is of paramount importance to ISWM.
[0003] Since the garbage collection problem is NP-hard, exact algorithms cannot obtain an exact solution in polynomial time. Therefore, most scholars use heuristic or metaheuristic algorithms to solve this problem. Although existing research on the garbage collection problem has achieved some results, related studies all adopt static scheduling methods, which usually assume that information such as the number of garbage collection points and vehicle conditions in the environment can be predicted in advance and remain constant. However, in the actual garbage collection process, various dynamic disturbances inevitably occur, such as the need for cleaning new garbage collection points and sudden vehicle malfunctions, which make the original transportation plan unreasonable or even infeasible. Therefore, this paper proposes a dynamic route planning method and system for garbage vehicles. Summary of the Invention
[0004] The purpose of this invention is to provide a low-carbon dynamic vehicle route planning method and system for waste collection. The planning method can dynamically derive vehicle routes with lower transportation costs and lower carbon emissions even under uncertain environments. It uses a multi-trip vehicle routing problem as its basic model, considering practical factors such as low-carbon indicators, maximum driver working hours, and vehicle capacity constraints. Simultaneously, it introduces two types of dynamic events closely related to waste collection: one is the emergence of new collection needs at waste disposal points over time; the other is vehicle malfunctions during transportation. To solve this model, a method is proposed... Q The Q-learning-based Hyper-heuristic Particle Swarm Optimization (QLHPSO) algorithm employs eight low-level heuristic search operators as evolutionary mechanisms for individuals, and is based on the hyperheuristic algorithm framework. Q Learning high-level strategies to control low-level heuristic search operators, i.e., through... QThrough training, the population in different states gradually learns to choose the learning strategy most suitable for itself, thereby generating higher-quality offspring particles. Furthermore, to respond to dynamic events promptly and efficiently, QLHPSO designed a dynamic response mechanism. By memorizing historical optimal information, utilizing heuristic information from dynamic events, and random initialization, a high-quality initial population is constructed after each dynamic event, providing a good starting point for the algorithm and accelerating its convergence speed.
[0005] The objective of this invention can be achieved through the following technical solutions:
[0006] A low-carbon dynamic path planning method for garbage trucks, the dynamic path planning method comprising the following steps:
[0007] Step 1: Read the input information of the instance, define the optimization objective, and set the constraints.
[0008] Step Two: Q Initialize the parameters of the hyperheuristic particle swarm optimization algorithm for learning.
[0009] Step 3: Generate an initial candidate population and calculate fitness to determine individual extreme values and global extreme values.
[0010] Step 4: Select the underlying heuristic search operator LLH for the population through the high-level learning strategy HLS, and generate a new population. Calculate the target value for each new individual and increase the Eva value by PS.
[0011] Step 5: Update the individual extrema and global extrema in each iteration.
[0012] Step Six: If Eva > Eva If the value is maxed out, the iteration terminates and the individual with the best fitness is output. This individual represents the planned vehicle scheduling scheme. Otherwise, proceed to step four.
[0013] Furthermore, in step one, input information is first obtained, and then the information is input into the pre-constructed low-carbon dynamic path planning model of the Q-learning hyperheuristic particle swarm algorithm. The Q-learning hyperheuristic particle swarm algorithm introduces a dynamic response mechanism, and finally determines the optimal scheduling scheme under the current environment at the moment when the dynamic event occurs.
[0014] Furthermore, the Q-learning hyperheuristic particle swarm optimization algorithm employs integer encoding and generates data based on a dynamic response mechanism. PS Individual, each individual X The encoding is a string consisting of 3~ n ( tl A sequence of integers between )
[0015]
[0016] in, i Indicates the number of the garbage disposal point. i = [3, n ( tl Given that the optimization objective is the transportation cost generated in the path, i.e., the lower the transportation cost, the higher the fitness and the better the planned solution, then the individual fitness is defined as:
[0017]
[0018] Determine the individual extreme value and the global extreme value.
[0019] Furthermore, the dynamic response mechanism includes using dynamic event-inspired information to repair the original solution, reusing historical optimal information, and random initialization.
[0020] Furthermore, the process of selecting the low-level heuristic search operator LLH for the population through the high-level learning strategy HLS and generating a new population specifically includes:
[0021] (1) If the population is using it for the first time Q The high-level learning strategy (QHLS) will... Q Learned Q All elements in the table are initialized to 0; otherwise, the table is updated based on the reward value obtained from performing the corresponding action according to the previous group state. Q The table and update formula are as follows;
[0022] (2) Perceive the current population state based on the current optimal solution.
[0023] (3) Calculate the selection probability of each low-level heuristic search operator in the current state according to the formula, as follows;
[0024] .
[0025] (4) Determine the optimal LLH under the current population state by using the roulette wheel selection strategy.
[0026] Furthermore, the low-level heuristic search operator includes a combination of four learning operators and two local search operators. The four learning operators include a greedy learning strategy, a multi-faceted learning strategy, an exploratory learning strategy, and an exploitative learning strategy. The two local search operators include an enhanced local search operator 1 and an enhanced local search operator 2.
[0027] Furthermore, the specific steps for determining the current population state are as follows:
[0028] (1) Assumption x best( t )and x best ( t -1) are respectively represented as the first t generation and first t The optimal solution of the -1 generation, ∆ f ( x best ) is defined as x best ( t )and x best ( t The increment of fitness values between -1) and Δ f ( x best )= f ( x best ( t ))− f ( x best ( t -1). Obviously, ∆ f ( x best )>0 indicates an improvement in population convergence, ∆ f ( x best )=0 indicates that the population has stagnated. (The sequence length is...) n Establish a line segment connection matrix for each individual. If the individual sequence contains points i Time j The line segment, then x ij =1, otherwise, x ij =0.
[0029]
[0030] (2) Obtain the connection matrix between two volumes using the following formula. G 1 and G 2. Set of identical line segments D ( G 1, G 2), then calculate using the formula. G 1 and G 2 similarity .
[0031]
[0032]
[0033] (3) Defined the similarity of populations. Wherein is the sum of all individuals in the population and the current best individual. x The average similarity between the best ones. similarpop = [0,1]. .
[0034] A low-carbon dynamic route planning system for garbage trucks, characterized in that the acquisition module is used to acquire input information, including: at the dispatch point t l The total number of garbage collection points, parking lots, and transfer stations that garbage collection vehicles need to serve. n ( t l ), coordinates of garbage collection points, garbage yards, garbage transfer stations, the amount of garbage at each collection point, and the capacity of garbage trucks. Q and the driver's maximum working hours T max。
[0035] The determination module is used to input information into a pre-built... Q In the low-carbon dynamic path planning model of garbage collection vehicles, the hyperheuristic particle swarm optimization algorithm is studied. Q The hyperheuristic particle swarm optimization algorithm introduces a dynamic response mechanism to determine the optimal scheduling scheme for the current environment at the moment a dynamic event occurs.
[0036] The Q The optimization objective of the low-carbon dynamic path planning model for garbage collection vehicles, which learns from the hyperheuristic particle swarm optimization algorithm, is to optimize the path planning at the scheduling point. t l The planned scheduling scheme has lower transportation costs and lower carbon emissions; Q The constraint condition for the low-carbon dynamic path planning model of garbage collection vehicles based on the hyperheuristic particle swarm optimization algorithm is that at the scheduling point... t l All vehicles at the dispatch point t l All departures originate from the last station visited before the dynamic event occurs and depart only once, at the scheduling point. t l Each unserved garbage collection point is allowed to be served by only one vehicle once; when a garbage collection point is served, a vehicle must travel from a certain location to that garbage collection point and leave from that location; all vehicles must empty all garbage at the garbage transfer station; the amount of garbage loaded by each vehicle during a single trip must not exceed its capacity limit; and the working time of each driver must not exceed the prescribed maximum working time limit.
[0037] A computer-readable storage medium for storing one or more programs, characterized in that the one or more programs include instructions that, when executed by a computing device, cause the computing device to perform any one of the methods according to claims 1 to 7.
[0038] A computing device, characterized in that it comprises:
[0039] One or more processors, a memory, and one or more programs, wherein the one or more programs are stored in the memory and configured to be executed by the one or more processors, the one or more programs including instructions for performing any of the methods according to claims 1 to 7.
[0040] The beneficial effects of this invention are:
[0041] 1. The dynamic path planning method of this invention adopts a low-carbon dynamic path planning method for garbage collection vehicles using a Q-learning super-heuristic particle swarm algorithm. It designs a decoding method that eliminates time and capacity constraints, so that all decoded solutions are feasible solutions, thereby improving the solution efficiency of the algorithm.
[0042] 2. The dynamic path planning method of the present invention Q The hyperheuristic particle swarm optimization algorithm introduces a dynamic response mechanism. By memorizing historical optimal information, utilizing the heuristic information of dynamic events, and random initialization, a high-quality initial population is constructed after each dynamic event, providing a good search starting point for the algorithm and accelerating the convergence speed.
[0043] 3. The dynamic path planning method of this invention selects the low-level heuristic search operator LLH for the population through the high-level learning strategy HLS and generates a new population. Eight low-level heuristic search operators are designed as the evolutionary mode of individuals, and it is based on the hyperheuristic algorithm framework. Q Learning high-level strategies to control low-level heuristic search operators, i.e., through... Q Through training, populations in different states gradually learn to choose the learning strategy that best suits them, thereby generating higher-quality offspring particles. Attached Figure Description
[0044] Figure 1 is a flowchart of the main process of the hyperheuristic particle swarm algorithm used in this invention.
[0045] Figure 2 shows the scheme repair strategy when introducing new garbage disposal points using the hyperheuristic particle swarm algorithm in this invention;
[0046] Figure 3 illustrates the repair strategy for vehicle faults introduced by the present invention using a hyperheuristic particle swarm algorithm. Detailed Implementation
[0047] The present invention will be further described below with reference to the accompanying drawings. The following embodiments are only used to more clearly illustrate the technical solution of the present invention, and should not be used to limit the scope of protection of the present invention.
[0048] This embodiment takes Nanjing Jiangbei New Area as an example, obtaining the coordinates of 54 residential communities, 1 warehouse and 1 garbage treatment station, as well as the daily garbage volume of each community. The capacity of garbage trucks is limited to 10 tons, and the maximum driving time is limited to 8 hours.
[0049] A low-carbon dynamic path planning method for garbage trucks, such as Figure 1 As shown, the dynamic path planning method includes the following steps:
[0050] Step 1: Read the input information of the instance, define the optimization objective, and set the constraints.
[0051] Plane coordinate information of the point ( Ax 1, Ay 1), ( Ax 2, Ay 2), ..., ( Axn ( tl ), Ayn ( tl The size of a problem indicates the number of elements it contains. n ( tl If there are 1 point (parking lot number 1, garbage transfer station number 2), then the distance between different points is calculated using the Euclidean distance formula, which is defined as:
[0052]
[0053] in, dij Point i With point j The distance between them;
[0054] The optimization objective, "transportation costs generated along the route," is defined as follows:
[0055]
[0056] in, f ( X This includes fixed costs (Cfixed). tl ), fuel cost Cfuel tl ) and carbon emission costs Ccarbon ( tl ).
[0057] (a) Fixed costs:
[0058] During the garbage collection process, once a vehicle is used, corresponding maintenance costs and driver salaries will be incurred. The fixed cost Cfixed for completing a garbage collection operation is... tl )as follows:
[0059]
[0060] (b) Fuel costs
[0061] Fuel costs arise during vehicle operation due to fuel consumption, and fuel efficiency is often affected by factors such as driving speed and road conditions. This paper assumes that road conditions are stable and speed is uniform during transportation, and that the fuel cost per unit distance traveled is fixed. Therefore, the total fuel cost Cfuel incurred after the vehicle completes its entire journey is calculated. tl )as follows:
[0062]
[0063] (c) Carbon emission performance:
[0064] Due to the worsening greenhouse effect, many countries have implemented carbon tax systems to control CO2 emissions. A carbon tax is a fee levied on CO2 emissions based on the amount emitted. The carbon emission cost (Ccarbon) of all vehicles using the vehicle for the entire journey is as follows:
[0065]
[0066]
[0067] in, FE Here are the fuel emission parameters: Ce is the carbon tax, Cm is the fuel cost per unit distance traveled, and Cf is the fixed operating cost per vehicle. Bk ( tl ) indicates at the scheduling point tl ,vehicle k A collection of all itineraries. dij For point i With point j The distance between them FCijk For vehicles k From point i Arrive at the destination j The amount of fuel consumed is calculated as follows:
[0068]
[0069] in, αij and β These are parameters related to road conditions and vehicle type, calculated as follows:
[0070]
[0071] in, a For vehicle acceleration, g Let gravitational acceleration be constant. θij From point i Time j The road surface slope of this section is... Cr The rolling resistance coefficient, Cd This is the traction coefficient. A This refers to the frontal surface area of the vehicle. ρ This indicates air density.
[0072] The constraints are defined as follows:
[0073] (1) Ensure that all vehicles are at the dispatch point tl Each departure starts from the last station visited before the dynamic event occurs and departs only once.
[0074]
[0075] (2) Ensure that at the scheduling point tl Each remaining unserved waste collection point is only allowed one vehicle to serve it once, that is:
[0076]
[0077] (3) When each garbage collection point is served, a vehicle will always travel from a certain location to that garbage collection point and leave from that location, that is:
[0078]
[0079] (4) Ensure that all vehicles empty their garbage completely at the garbage transfer station, i.e.:
[0080] (5) This ensures that the amount of garbage loaded by each vehicle during a single trip does not exceed its capacity limit, i.e.:
[0081]
[0082] (6) Ensure that the working time of each vehicle does not exceed the prescribed maximum working time limit, that is:
[0083]
[0084] in, qi Indicates each garbage disposal point i The amount of garbage, lijk Indicates vehicle k From point i Drive toj The load capacity, tij Indicates the vehicle starts from point i Arrive at the destination j Time, Tmax This indicates the driver's maximum daily working hours.
[0085] Step Two: Q Initialize the parameters of the hyperheuristic particle swarm optimization algorithm for learning.
[0086] Step 3: Generate an initial candidate population and calculate fitness to determine individual extreme values and global extreme values.
[0087] Q The hyperheuristic particle swarm optimization algorithm is learned using integer encoding, for those containing n ( tl The problem at point 1 (garage yard number 1, garbage transfer station number 2) is generated based on a dynamic response mechanism. PS Individual, each individual X The encoding is a string consisting of 3~ n ( tl A sequence of integers between )
[0088]
[0089] in, i Indicates the number of the garbage disposal point. i = [3, n ( tl Given that the optimization objective is the transportation cost generated in the path, i.e., the lower the transportation cost, the higher the fitness and the better the planned solution, then the individual fitness is defined as:
[0090]
[0091] Determine the individual extreme value and the global extreme value.
[0092] Step 4: Select the underlying heuristic search operator LLH for the population using the high-level learning strategy HLS, and generate a new population. Calculate the target value for each new individual, increasing the Eva value by PS, including:
[0093] S41. Set up a greedy learning strategy
[0094] Individuals generate mutated individuals through reverse mutation; secondly, mutated individuals and individual extrema are combined with a greedy crossover operator to generate crossover individual one; finally, crossover individual one is combined with the global extrema in the same way to generate a new individual after greedy learning.
[0095] S42. Setting up diverse learning strategies
[0096] The particle generates mutated individuals through a multivariate mutation operator; secondly, the generated mutated individuals are partially mapped and crossed with the individual extrema to obtain cross-entity one; finally, cross-entity one is partially mapped and crossed with the global extrema to generate a new individual after multivariate learning.
[0097] S43. Set up an exploratory learning strategy:
[0098] Particles generate mutated individuals through multivariate mutation operators; secondly, the generated mutated individuals are partially mapped and crossed with the individual extreme values of any particle in the population with better fitness than itself, generating new individuals after exploration and learning.
[0099] S44. Set up an exploitation-based learning strategy, where particles generate mutated individuals through a multivariate mutation operator; then, perform partial mapping and cross-mapping between the generated mutated individuals and the individual extreme values to generate new individuals after exploitation learning.
[0100] S45. Set up enhanced local search operator 1 and enhanced local search operator 2, and combine the learning operator and the local search operator into a low-level heuristic search operator.
[0101] The specific implementation steps of the enhanced local search operator 1 are as follows:
[0102] S451. By decoding individuals using a decoding method that eliminates time and capacity constraints, a scheduling scheme for garbage collection vehicles is obtained.
[0103] S452. Find the vehicle index and trip index corresponding to the trip containing the most garbage disposal points;
[0104] S453. Perform 2-opt optimization on the route with the most garbage disposal points to open up intersecting routes and effectively find new routes with lower costs;
[0105] S454. Reorganize the optimized trip with the remaining unselected trips to form a new individual. During the reorganization process, there are two cases based on the destination of the vehicle trip: when the destination of the trip is a transfer station, it means that the vehicle will not go to the parking lot again in this trip, and only the starting point and the garbage transfer station in the trip need to be removed; when the destination of the trip is the parking lot, the vehicle must be going to the garbage transfer station to unload and then return to the parking lot, and in this case, the starting point, the garbage transfer station and the parking lot in this trip need to be removed.
[0106] S455, Output the new individual after implementing Enhanced Local Search Strategy 1, if the original individual X If the target value is worse than the new individual, then the new individual is used to replace the original individual.
[0107] The operation of Enhanced Local Search Operator 2 is similar to that of Enhanced Local Search Operator 1. The difference is that Enhanced Local Search Operator 2 no longer uses the 2-opt operator to optimize the route containing the most garbage disposal points. Instead, it applies the point interpolation method to the route, which helps to find the optimal solution when the routes do not intersect.
[0108] S46. If the population is using it for the first time Q The high-level learning strategy (QHLS) will... Q Learned Q All elements in the table are initialized to 0; otherwise, the table is updated based on the reward value obtained from performing the corresponding action according to the previous group state. Q The table and update formula are as follows;
[0109] S47. Perceive the current population state based on the current optimal solution.
[0110] The specific steps for determining the current population state are as follows:
[0111] S471, Assumption x best( t )and x best( t -1) are respectively represented as the first t generation and first t The optimal solution of the -1 generation, ∆ f ( x (best) is defined as x best( t ) and x best( t The increment of fitness values between -1) is ∆ f ( x best)= f ( x best( t ))− f ( x best( t -1)) Obviously, ∆ f ( x best)>0 indicates an improvement in population convergence, ∆ f ( x best)=0
[0112] This indicates that the population has stagnated, and the sequence length is [value missing]. n Establish a line segment connection matrix for each individual. ,
[0113] If the individual sequence contains a line segment from point i to point j
[0114]
[0115] S472, Obtain the connection matrix between two volumes using the following formula. G 1 and G 2. Set of identical line segments D ( G 1, G 2), then calculate using the formula. G 1 and G 2 similarity .
[0116]
[0117] S473 defines the similarity of populations. Wherein is the sum of all individuals in the population and the current best individual. x The average similarity between the best ones. similarpop = [0,1].
[0118] The criteria for classifying population states are shown in the table above.
[0119] S48. Calculate the selection probability of each low-level heuristic search operator in the current state according to the formula. The formula is as follows;
[0120]
[0121] S49. Determine the optimal LLH under the current population state using a roulette wheel selection strategy.
[0122] Step 5: Update individual and global extrema: Update individual and global extrema in each iteration according to the survival of the fittest rule;
[0123] Step Six: If Eva > Eva If the value of the vehicle scheduling scheme is maxed out, the iteration terminates and the individual with the best fitness is output. Otherwise, proceed to step four.
[0124] A low-carbon dynamic route planning system for garbage trucks, the planning system comprising:
[0125] Obtain input information, including at the scheduling point. tl The total number of garbage collection points, parking lots, and garbage transfer stations that garbage collection vehicles need to serve. n ( tl), coordinates of garbage collection points, garbage yards, garbage transfer stations, the amount of garbage at each collection point, and the capacity of garbage trucks. Q and the driver's maximum working hours Tmax ;
[0126] Input information into a pre-built... Q In the low-carbon dynamic path planning model of garbage collection vehicles, the hyperheuristic particle swarm optimization algorithm is studied. Q The hyperheuristic particle swarm optimization algorithm introduces a dynamic response mechanism to determine the optimal scheduling scheme in the current environment at the moment a dynamic event occurs.
[0127] The aforementioned Q The optimization objective of the low-carbon dynamic route planning model for garbage collection vehicles, based on the hyperheuristic particle swarm optimization algorithm, is to achieve low transportation costs and low carbon emissions in the planned scheduling scheme. The constraint condition for this model is that all vehicles at the scheduling point... tl All departures originate from the last station visited before the dynamic event occurs and depart only once, at the scheduling point. tl Each unserved garbage collection point is allowed to be served by only one vehicle once; when a garbage collection point is served, a vehicle must travel from a certain location to that garbage collection point and leave from that location; all vehicles must empty all garbage at the garbage transfer station; the amount of garbage loaded by each vehicle during a single trip must not exceed its capacity limit; and the working time of each driver must not exceed the prescribed maximum working time limit.
[0128] A computer-readable storage medium storing one or more programs, the one or more programs including instructions that, when executed by a computing device, cause the computing device to perform any of the methods.
[0129] A computing device, comprising:
[0130] One or more processors, a memory, and one or more programs, wherein the one or more programs are stored in the memory and configured to be executed by the one or more processors, and the one or more programs include instructions for performing any of the methods.
[0131] The dynamic path planning system and method of this invention are further illustrated by the following simulation experiments:
[0132] 1. Experimental conditions:
[0133] Simulation was performed using MATLAB R2017b on an Intel(R) Core(TM) i5-5500U CPU@2.40GHz, 8 GB of memory, and Windows 10 system.
[0134] 2. Experiment Content:
[0135] We conducted an on-site investigation of Green Ring Company in Jiangbei New Area, Nanjing, China, and obtained a real-world example of urban household waste collection and transportation. This example includes 54 residential communities, 1 warehouse, and 1 waste disposal site in Jiangbei New Area.
[0136] The coordinates of the processing station, and the daily garbage volume for each residential area. Garbage trucks have a capacity limit of 10 tons, and the driver...
[0137] The maximum duration is limited to 8 hours.
[0138] 3. Experimental Results
[0139] The experimental results of this invention and existing algorithms for solving dynamic vehicle routing problems were compared on a waste collection and transportation example at Lvhuan Company in Jiangbei New Area, Nanjing.
[0140] To verify Q This study examines the performance of the hyperheuristic particle swarm optimization algorithm in solving the low-carbon dynamic vehicle routing problem for waste collection. Table 1 lists four existing algorithms and their performance on a waste collection example at Lvhuan Company in Jiangbei New Area, Nanjing. Q Comparative results of learning hyperheuristic particle swarm optimization algorithms.
[0141] This experiment reproduced four existing algorithms for solving the dynamic vehicle routing problem (DVRP): GA_TS, SAEA, PSO, and 2MPSO. These algorithms were applied to a newly established low-carbon dynamic vehicle routing problem for waste collection and compared with the proposed algorithm QLHPSO. GA_TS is a hybrid algorithm combining genetic algorithm and tabu search algorithm; SAEA adaptively evolves the configuration of the evolutionary algorithm (parameter values, search operator numbers, and search operator calling order) to generate high-quality solutions in an efficient evolutionary manner; PSO is a particle swarm optimization algorithm for solving DVRP using a continuous encoding method; and 2MPSO includes a heuristic initialization phase and a particle swarm optimization phase. Here, Best and mean represent the optimal and average transportation cost target values found in 20 runs, respectively, with the best values of Best and mean highlighted in bold. The experimental results show that the proposed algorithm... QThe QLHPSO hyperheuristic particle swarm optimization algorithm outperforms the four comparative algorithms in terms of overall performance for waste collection and transportation. It has high solution accuracy and can effectively solve low-carbon dynamic vehicle routing problems. It can quickly plan a vehicle scheduling scheme with low transportation costs and low carbon emissions in dynamic environments.
[0142] Table 1
[0143]
[0144] When a new garbage collection point W requests collection, all vehicles are instructed to serve the existing garbage collection points according to the previous dispatch plan. For the newly added garbage collection point W, the following measures are taken: Figure 2 The handling method shown is as follows: When a vehicle malfunctions, the order in which normally functioning vehicles serve the stations remains unchanged. For waste collection points that cannot continue to be served due to vehicle malfunctions, the following measures are taken: Figure 3 The processing method shown is to improve the original solution by using dynamic event-inspired information, taking advantage of the dynamic characteristics of the low-carbon dynamic vehicle routing problem for waste collection, thereby improving the search efficiency of the algorithm.
[0145] In the description of this specification, references to terms such as "an embodiment," "example," "specific example," etc., indicate that a specific feature, structure, material, or characteristic described in connection with that embodiment or example is included in at least one embodiment or example of the invention. In this specification, illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples.
[0146] The foregoing has shown and described the basic principles, main features, and advantages of the present invention. Those skilled in the art should understand that the present invention is not limited to the above embodiments. The embodiments and descriptions in the specification are merely illustrative of the principles of the invention. Various changes and modifications can be made to the invention without departing from its spirit and scope, and all such changes and modifications fall within the scope of the claimed invention.
Claims
1. A low-carbon dynamic path planning method for garbage trucks, characterized in that, The dynamic path planning method includes the following steps: Step 1: Read the input information of the instance, define the optimization objective, and set constraints; Step Two: Q Initialize the parameters of the hyperheuristic particle swarm optimization algorithm for learning; Step 3: Generate an initial candidate population, calculate fitness, and determine individual extreme values and global extreme values; Step 4: Select the underlying heuristic search operator LLH for the population through the high-level learning strategy HLS, and generate a new population. Calculate the target value for each new individual and increase the Eva value by PS. Step 5: Update the individual extrema and global extrema in each iteration; Step Six: If Eva > Eva If the fitness value is maxed out, the iteration terminates and the individual with the best fitness is output. This individual is the planned vehicle scheduling scheme. Otherwise, proceed to step four. The first step involves acquiring input information and then inputting it into a pre-built low-carbon dynamic path planning model based on the Q-learning hyperheuristic particle swarm algorithm. The Q-learning hyperheuristic particle swarm algorithm introduces a dynamic response mechanism and finally determines the optimal scheduling scheme under the current environment at the moment a dynamic event occurs. The Q-learning hyperheuristic particle swarm optimization algorithm uses integer encoding and is based on a dynamic response mechanism for generation. PS Individual, each individual X The encoding is a string consisting of 3~ n ( tl A sequence of integers between ) in, i Indicates the number of the garbage disposal point. i = [3, n ( tl Given that the optimization objective is the transportation cost generated in the path, i.e., the lower the transportation cost, the higher the fitness and the better the planned solution, then the individual fitness is defined as: Determine individual extreme values and global extreme values; The dynamic response mechanism includes using dynamic event-inspired information to repair the original solution, reusing historical optimal information, and random initialization; The process of selecting the low-level heuristic search operator LLH for the population through the high-level learning strategy HLS and generating a new population specifically includes: (1) If the population is using it for the first time Q High-level learning strategies will Q Learned Q All elements in the table are initialized to 0; otherwise, the table is updated based on the reward value obtained from performing the corresponding action according to the previous group state. Q The table and update formula are as follows; (2) Perceive the current population state based on the current optimal solution; (3) Calculate the selection probability of each low-level heuristic search operator in the current state according to the formula, as follows; ； (4) Determine the optimal LLH under the current population state by using the roulette wheel selection strategy.
2. The low-carbon dynamic path planning method for garbage trucks according to claim 1, characterized in that, The low-level heuristic search operators include a combination of four learning operators and two local search operators. The four learning operators include a greedy learning strategy, a multi-faceted learning strategy, an exploratory learning strategy, and an exploitative learning strategy. The two local search operators include an enhanced local search operator 1 and an enhanced local search operator 2.
3. The low-carbon dynamic path planning method for garbage trucks according to claim 2, characterized in that, The specific steps for determining the current population state are as follows: (1) Assumption x best ( t )and x best ( t -1) are respectively represented as the first t generation and first t The optimal solution of the -1 generation, ∆ f ( x best ) is defined as x best ( t )and x best ( t The increment of fitness values between -1) and Δ f ( x best )= f ( x best ( t ))− f ( x best ( t −1));∆ f ( x best )>0 indicates an improvement in population convergence, ∆ f ( x best )=0 indicates that the population has stagnated, and the sequence length is 0. n Establish a line segment connection matrix for each individual. If the individual sequence contains points i Time j The line segment, then x ij =1, otherwise, x ij =0; (2) Obtain the connection matrix between two volumes using the following formula. G 1 and G 2. Set of identical line segments D ( G 1, G 2), then calculate using the formula. G 1 and G 2 similarity ; (3) Defined the similarity of populations. Wherein is the sum of all individuals in the population and the current best individual. x The average similarity between the best ones. similarpop = [0,1].
4. A low-carbon dynamic path planning system for garbage trucks, characterized in that, The acquisition module is used to acquire input information, including: at the scheduling point t l The total number of garbage collection points, parking lots, and garbage transfer stations that garbage collection vehicles need to serve. n ( t l ), coordinates of garbage collection points, garbage yards, garbage transfer stations, the amount of garbage at each collection point, and the capacity of garbage trucks. Q and the driver's maximum working hours T max ; The determination module is used to input information into a pre-built... Q In the low-carbon dynamic path planning model of garbage collection vehicles, the hyperheuristic particle swarm optimization algorithm is studied. Q The hyperheuristic particle swarm optimization algorithm introduces a dynamic response mechanism to determine the optimal scheduling scheme in the current environment at the moment a dynamic event occurs. The Q The optimization objective of the low-carbon dynamic path planning model for garbage collection vehicles, which learns from the hyperheuristic particle swarm optimization algorithm, is to optimize the path planning at the scheduling point. t l The planned scheduling scheme has lower transportation costs and lower carbon emissions; Q The constraint condition for the low-carbon dynamic path planning model of garbage collection vehicles based on the hyperheuristic particle swarm optimization algorithm is that at the scheduling point... t l All vehicles at the dispatch point t l All departures originate from the last station visited before the dynamic event occurs and depart only once, at the scheduling point. t l Each unserved garbage collection point is allowed to be served by only one vehicle once; when a garbage collection point is served, a vehicle must travel from a certain location to that garbage collection point and leave from that location; all vehicles must empty all garbage at the garbage transfer station; the amount of garbage loaded by each vehicle during a single trip must not exceed its capacity limit; and the working time of each driver must not exceed the prescribed maximum working time limit.
5. A computer-readable storage medium for storing one or more programs, characterized in that, The one or more programs include instructions that, when executed by a computing device, cause the computing device to perform the method of any one of claims 1 to 3.
6. A computing device, characterized in that, include: One or more processors, a memory, and one or more programs, wherein the one or more programs are stored in the memory and configured to be executed by the one or more processors, the one or more programs including methods for performing any one of claims 1 to 3.