Dangerous cargo transport vehicle path planning method and system

By constructing a multi-objective optimization model and a compliance repair operator to optimize the transportation routes of dangerous goods, the problem of compartmentalized loading and resource scheduling in refined oil distribution has been solved, thereby improving safety, resource utilization efficiency and operational synergy, and adapting to order changes and emergencies.

CN121810137APending Publication Date: 2026-04-07CHINA NAT PETROLEUM CORP +2
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202512028573.6
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2025-12-30
Publication Date
2026-04-07

AI Technical Summary

Technical Problem

Existing technologies fail to effectively model the compartmentalized loading structure of tank trucks in refined oil distribution, resulting in the inability to enforce oil compatibility and unloading port sequence constraints. This leads to fragmented resource scheduling, a singular objective, difficulty in achieving full-process compliance and maximizing overall resource utilization efficiency, and reliance on manual experience to handle order changes and emergencies, which is time-consuming and difficult to optimize.

Method used

A multi-objective optimization model is constructed, which combines transportation costs, time and safety risks, and incorporates constraints such as warehouse compatibility, driver qualifications and fatigue driving regulations. The optimal route plan is generated through a multi-objective optimization algorithm, and the scheduling scheme is optimized by using a compliance repair operator to ensure the coordinated optimization of routes, loading and manpower.

Benefits of technology

It achieves inherent safety in the transportation of dangerous goods, improves vehicle space utilization and loading rate, reduces empty mileage, lowers logistics costs, enhances supply chain collaboration efficiency, and enables rapid handling of large-scale orders and emergencies.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN121810137A_ABST
    Figure CN121810137A_ABST
Patent Text Reader

Abstract

The invention relates to the technical field of dangerous cargo transportation, and discloses a dangerous cargo transportation vehicle path planning method and system, and the method comprises the following steps: obtaining resource data; wherein the resource data comprises oil product data, vehicle data and driver data; constructing a path planning model with the transportation cost, the transportation time and the transportation safety risk as targets; setting constraint conditions of the path planning model; wherein the constraint condition comprises that the quantity of oil products distributed to the shipping space of the vehicle is lower than the rated capacity of the shipping space; the oil types of the adjacent storage bins of the vehicle are compatible; matching a driver with an oil product type according to qualification; and solving the path planning model, and taking a path corresponding to the solution of the path planning function when the target function value is minimum as an optimal path. According to the method and the device, the problem that generation of illegal scheduling schemes is difficult to eradicate from the source in the prior art is solved, and the problem that generation of illegal scheduling schemes is difficult to eradicate from the source in the prior art is solved.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] This invention relates to the field of dangerous goods transportation technology, specifically a method and system for route planning of dangerous goods transportation vehicles. Background Technology

[0002] Refined oil product distribution is a crucial link connecting oil depots, gas stations, and end customers. Its dispatch efficiency and accuracy directly impact a company's operating costs, service levels, and safety risks. Existing technological solutions largely focus on route planning for general hazardous chemicals, aiming to find the optimal route from origin to destination and balance transportation costs, time, and road safety risks. However, these solutions have significant limitations when addressing the specific scenario of refined oil product distribution: 1. Ignoring core physical constraints: The unique compartmentalized loading structure of the tanker truck was not modeled, as well as the resulting oil compatibility constraints and unloading port sequence constraints, which resulted in the planning scheme being unexecutable or inefficient during actual loading and unloading.

[0003] 2. Fragmented resource scheduling: Vehicle routing issues are usually handled separately from driver scheduling and qualification matching issues, lacking integrated collaborative optimization of "people-vehicle-goods-road", making it difficult to maximize the overall efficiency of resource utilization.

[0004] 3. Single objective: The main objective is to minimize the path or cost, with insufficient consideration given to indicators that affect overall operational efficiency, such as "arrival time coordination", "site operation efficiency" and "full-process compliance".

[0005] 4. Weak adaptability: It relies on human experience to handle massive orders and complex constraints. The scheduling is time-consuming and highly subjective, making it difficult to respond quickly to changes in orders, vehicle malfunctions or traffic emergencies, and it is difficult to guarantee the overall optimal solution. Summary of the Invention

[0006] To overcome the shortcomings of existing technologies, this invention provides a method and system for route planning of dangerous goods transport vehicles, which solves the problem that existing technologies cannot prevent the generation of illegal scheduling schemes from the source.

[0007] The technical solution of the present invention to solve the above-mentioned technical problems is as follows: A method for route planning of dangerous goods transport vehicles includes the following steps: Acquire resource data; the resource data includes: fuel data, vehicle data, and driver data; Construct a route planning model with transportation cost, transportation time, and transportation safety risks as objectives. The formula for the objective route planning model is: + + ; in, This represents the minimization operation. Represents the objective function value. Indicates the first weight. Indicates transportation costs, Indicates the second weight. Indicates the transportation time. Indicates the third weight. Indicates the total security risk; Define the constraints for the path planning model; The constraints include: The amount of fuel allocated to the vehicle's storage compartment is less than the compartment's rated capacity; The fuel types in adjacent compartments of the vehicle are compatible; Driver qualifications must be matched with fuel type; Solve the path planning model, and take the path corresponding to the solution of the path planning function that minimizes the objective function value as the optimal path.

[0008] By embedding hard constraints such as warehouse compatibility, driver qualifications, and fatigue driving regulations into the core of the optimization algorithm, the generation of illegal scheduling schemes is eliminated from the source, achieving inherent safety. Through precise matching optimization of "warehouse-order", the utilization rate of vehicle space and loading rate are significantly improved, empty mileage and empty warehouse waste are reduced, and the overall logistics cost can be reduced by more than 10%. With "arrival time optimization" as the driving force, routes, loading, and manpower are coordinated, reducing customer waiting time and site operation conflicts, and improving the overall supply chain collaboration efficiency. It overcomes the shortcomings of manual scheduling, such as long time consumption, reliance on experience, and difficulty in global optimization, and can quickly process large-scale orders and complex constraints, and dynamically respond to emergencies.

[0009] As a preferred technical solution, solving the path planning model and taking the path corresponding to the solution of the path planning function that minimizes the objective function value as the optimal path includes the following steps: Resource data is combined and encoded and input into a multi-objective optimization model to generate an initial solution for the path planning function; The resource data combination is changed and encoded as input into the multi-objective optimization model, and the initial solution of the initial path planning is iterated to generate a new solution; All solutions are scored for fitness, and the path corresponding to the solution with the highest fitness is selected as the optimal path.

[0010] For the repaired solution data, the complete scheduling scheme is derived from the encoding and decoding, substituted into the objective function for calculation, and the fitness score of the solution data is generated for subsequent selection operations.

[0011] As a preferred technical solution, resource data is combined and encoded and input into a multi-objective optimization model to iteratively solve the initial solution of the initial path planning and generate a new solution, including the following steps: Randomly combine resource data, encode the combined resource data, and generate path codes and / or load codes; Among them, the route code stores driver data and candidate route data, while the loading code stores the warehouse data of the order being assigned to the vehicle.

[0012] This facilitates the generation of new solutions through the selected method.

[0013] As a preferred technical solution, the process involves changing the combination of resource data and encoding it as input to a multi-objective optimization model, iterating through the initial solution of the initial path planning to generate a new solution, and including the following steps: Two solutions are randomly selected, and all or part of the resource data corresponding to the two solutions are swapped. The resource data is then combined and encoded and input into the multi-objective optimization model. The initial solution of the initial path planning is iterated to generate a new solution.

[0014] This facilitates the generation of new solutions through cross-validation.

[0015] As a preferred technical solution, the process involves changing the combination of resource data and encoding it as input to a multi-objective optimization model, iterating through the initial solution of the initial path planning to generate a new solution, and including the following steps: The resource data corresponding to a certain solution is randomly modified. The modified resource data is then combined and encoded and input into the multi-objective optimization model. The initial solution of the initial path planning is iterated to generate a new solution.

[0016] This facilitates the generation of new solutions through mutation.

[0017] As a preferred technical solution, the process involves changing the combination of resource data and encoding it as input to a multi-objective optimization model, iterating through the initial solution of the initial path planning to generate a new solution, and including the following steps: Check if all new solutions meet the constraints: if yes, accept the new solution; if no, verify the new solution. The methods for verifying the new solution include one or more of the following: If a new solution is detected that does not meet the constraint that the amount of oil in the compartment allocated to the vehicle is less than the compartment's rated capacity, then the remaining capacity is searched in other compartments of the same vehicle, the loading code is updated, and the excess oil data is migrated. If a new solution is detected that does not meet the constraint that the oil type is compatible with the adjacent compartments of the vehicle, the order is prohibited from being assigned to one or two of the adjacent compartments of the vehicle, and the loading code data is updated to prohibit the identification of the oil type assigned to the adjacent compartments. If a new solution is detected that does not meet the constraint of matching driver qualification with fuel type, driver data that meets the driver qualification requirements is selected from the available driver data pool, and the selected driver data replaces the driver data in the path code.

[0018] Examine the three constraints and handle any content that does not meet the constraints.

[0019] As a preferred technical solution, if a new solution is detected that does not meet the constraint that the oil types of adjacent compartments of the vehicle are compatible, the oil type of a compartment with compatible oil types is first searched within the same vehicle to place the oil.

[0020] This facilitates efficient adjustments when the new solution does not meet the constraint of oil type compatibility between adjacent compartments of the vehicle.

[0021] As a preferred technical solution, the following steps are also included: Output one or more of the following based on the optimal path: Personnel Shift Schedule: Each driver's assigned vehicle, task sequence, and estimated working hours; Vehicle loading diagram: Types, weights, and delivery order of oil in each compartment of each vehicle; Delivery route map: the route of each vehicle, the estimated arrival time at each station, and the estimated departure time at each station; Global Gantt Chart: Visualizes the collaborative operation of all resource data over time.

[0022] The above content allows for visualization of the optimal path solution that balances cost, efficiency, and safety.

[0023] As a preferred technical solution, the constraints also include one or more of the following: The driver's qualifications must match the type of vehicle assigned and the risk level of the roads traversed. Drivers must not exceed the legal limit for continuous driving time; When planning delivery routes, it is essential to ensure that the order of visiting stations matches the order of unloading oil at each storage location. The vehicle departs from the oil depot and eventually returns; Customers are served within the designated time window; Overloading is not allowed.

[0024] The above constraints further enhance the dynamic matching capability of multi-dimensional compliance for "people-vehicle-goods-road".

[0025] A dangerous goods transport vehicle route planning system includes a processor, a memory, and a computer program stored in the memory. When the processor executes the computer program, it implements the steps of the dangerous goods transport vehicle route planning method.

[0026] Compared with the prior art, the present invention has the following advantages: (1) Safety and compliance are prioritized: By embedding hard constraints such as compartment compatibility, driver qualifications, and fatigue driving regulations into the core of the optimization algorithm, the generation of illegal scheduling schemes is eliminated from the source, thus achieving inherent safety; (2) Maximize resource utilization efficiency: Through precise matching and optimization of "warehouse-order", the utilization rate and loading rate of vehicle warehouses are significantly improved, empty driving mileage and empty warehouse waste are reduced, and the overall logistics cost can be reduced by more than 10%; (3) Enhanced operational synergy: Guided by "delivery time optimization", the route, loading, and manpower were coordinated, reducing customer waiting time and site operation conflicts, and improving the overall supply chain synergy efficiency; (4) Intelligent and automated scheduling: It overcomes the shortcomings of manual scheduling, such as long time consumption, reliance on experience, and difficulty in global optimization. It can quickly process large-scale orders and complex constraints, and dynamically respond to emergencies. Attached Figure Description

[0027] Figure 1 This is a flowchart of the present invention. Detailed Implementation

[0028] The present invention will be further described in detail below with reference to the embodiments and accompanying drawings, but the embodiments of the present invention are not limited thereto.

[0029] The principles and features of the present invention are described below. The embodiments given are only for explaining the present invention and are not intended to limit the scope of the present invention.

[0030] Example 1 like Figure 1 As shown, this invention addresses the core pain points of "warehouse-based loading, driver-vehicle binding, and collaborative arrival" in the refined oil delivery scenario by proposing a data-driven integrated scheduling optimization method. Its core lies in constructing an optimization model integrating orders, vehicles (including warehouses), drivers, road networks, and regulatory constraints, and solving it through an intelligent algorithm embedding dedicated repair operators, ultimately outputting executable integrated scheduling instructions.

[0031] This invention provides an intelligent scheduling method for refined oil product distribution resources that deeply integrates physical constraints of compartmentalized loading with multi-dimensional compliance rules. Its core innovation lies not in simple path planning, but in constructing a five-in-one collaborative optimization model of "people-vehicle-goods-road-warehouse" and designing an embedded solution mechanism adapted to it.

[0032] This invention has the following features: 1. Dynamic loading optimization model and method based on digital compartment mapping: Establish a digital compartment model that accurately corresponds to physical vehicles, and implement a triple dynamic constraint mechanism in scheduling optimization, including oil capacity matching, adjacent compartment compatibility verification, and qualification matching verification; 2. Implementation method of "people-vehicle-goods-road" multi-dimensional compliance dynamic matching engine: Real-time association of driver qualification status, vehicle characteristics, cargo hazard attributes and route risk level to achieve linkage analysis and rigid matching of the four factors, ensuring compliance of the dispatching plan throughout the entire process; 3. Improved heuristic search algorithm with embedded compliance repair operator: In the standard heuristic algorithm framework, a special operator is designed and embedded that can actively detect and repair specific violations such as warehouse conflicts and qualification mismatch. Complex constraint satisfaction is internalized as part of the search process, which significantly improves the solution efficiency and feasibility of the solution.

[0033] The overall process is as follows Figure 1 As shown.

[0034] The specific steps are as follows: 1. Data Acquisition and Preprocessing Obtain order requirements (oil type, quantity, time window), vehicle data (bay structure, load capacity, qualification requirements), and personnel data (driver level, remaining working hours).

[0035] 2. Construct a multi-objective optimization model A multi-objective optimization model is constructed that comprehensively considers total transportation cost, vehicle utilization efficiency, and operational safety risks. The optimal scheduling scheme is sought through weighted aggregation and other methods.

[0036] This invention does not seek to simultaneously achieve the theoretical optimal values ​​of the three objectives, but rather seeks an optimal, executable "balanced solution" or "compromise solution" among multiple conflicting objectives by constructing a "multi-objective comprehensive optimization function." The specific details are as follows: I. Conflicts between objectives This is precisely the core challenge of multi-objective optimization problems. In the scenario of refined oil product distribution, these three objectives are usually conflicting and cannot be simultaneously optimized. For example: (1) In pursuit of the lowest total transportation cost, there may be a tendency to make vehicles fully loaded and travel long distances, but this may increase the exposure time of high-risk road sections, thus conflicting with the lowest safety risk coefficient.

[0037] (2) Pursuing the highest vehicle utilization rate may lead to continuous vehicle operation and tight driver schedules, which may violate the regulations on fatigue driving and increase safety risks. It may also violate the customer time window (soft constraint) due to rushing, resulting in time penalties.

[0038] (3) In pursuit of the lowest safety risk coefficient, the safest route may be chosen and the most experienced driver may be assigned, but this will undoubtedly increase mileage costs and reduce vehicle turnover.

[0039] 3. Setting Constraints (1) Hard constraints: vehicle compartment compatibility, maximum load capacity, driver qualification compliance, and prohibited time periods / areas for hazardous chemicals; (2) Soft constraints: customer time window preferences and driver scheduling balance.

[0040] 4. Solution Algorithm Design Multi-source data integration and feature modeling ① Order Characterization: Define the order set Each order Includes oil type demand Demand Destination coordinates Expected time window And attributes such as unloading port restrictions.

[0041] ② Vehicle characterization: Define the set of vehicles Its key feature is a digitalized warehouse structure. For example, vehicles... have Each position is an independent position. Capacity And an oil compatibility matrix is ​​used to define the oil combinations that adjacent storage locations are allowed to load.

[0042] ③Personnel profiling: Includes qualification level (e.g., A / B / C class driver's licenses correspond to different hazardous materials transport vehicle types), current cumulative driving hours Fatigue status and historical safety score .

[0043] Constructing multi-constraint, multi-objective optimization models ① Loading constraints for different compartments: The loading volume of any compartment must not exceed its capacity; the type of oil in the same compartment must be unique; according to the compatibility matrix, incompatible oils are prohibited from being loaded into adjacent compartments; ② Compliance matching constraints between people, vehicles, and roads: The driver's qualifications must match the type of vehicle assigned and the risk level of the roads traversed (such as whether they pass through tunnels or densely populated areas); the driver's continuous driving time must not exceed the legal limit; ③ Physical constraints on unloading sequence: Based on the actual unloading pipeline design of the vehicle, when planning the delivery route, it is necessary to ensure that the order of visiting the station is consistent with the unloading sequence of each warehouse to avoid ineffective vehicle relocation within the station; ④ Flow balance and time window constraints: Vehicles depart from the oil depot and eventually return; customers are served within their time window; overloading is not allowed.

[0044] Multi-objective optimization function Establish objective function + + ; This includes fuel costs, road and bridge tolls, vehicle depreciation, and labor costs. Transportation cost weighting; Penalties for deviations from the customer's expected delivery window due to excessively long overall delivery cycles; Penalty weight; Safety risk values ​​calculated based on road conditions and driver status; Security risk weight.

[0045] The approach here is: 1. Quantification and Normalization: Indirectly incorporating the impact of targets such as "vehicle utilization rate" into the calculation. Cost_total (such as wasted resources by running empty) and Time_penalty (e.g., waiting time). All sub-objectives are converted into calculable costs (e.g., waiting time). Cost_total ),punish( Time_penalty ) or risk value ( Risk_score ).

[0046] 2. Weighted summation: By setting different weight coefficients This reflects the degree of preference of the scheduling strategy for different objectives. The weights are set by decision-makers based on actual operational strategies (for example, during peak seasons, cost and efficiency are given more emphasis). (High weighting), safety is more valued during off-seasons or special periods. (High weighting).

[0047] 3. Optimizing a Single Objective: Ultimately, the algorithm minimizes this weighted sum F, rather than minimizing three independent objectives separately. This essentially involves finding the scheduling scheme that achieves the best overall performance given the preferences (weights).

[0048] (4) Design an improved heuristic solution algorithm with embedded compliance repair operators. Core idea: An algorithm is a data-driven, iterative optimization process. It receives raw data from multiple sources, encodes it into optimizable "solution" data, and iteratively applies genetic operations and dedicated compliance repair operators to correct and optimize the "solution" data, ultimately outputting compliant and efficient scheduling scheme data.

[0049] The detailed steps of the solution algorithm are as follows: Step 1: Preparation of multi-source input data Data content: (1) Order data O Each order The data packet contains "{ Oil type t_i, demand q_i, destination Coordinates loc_i, time window [ET_i, LT_i], ...}".

[0050] (2) Vehicle data V Each vehicle The core of the data packet is its digital warehouse structure data, such as "{ Position List [c1, c2,...], Capacity of each storage unit [C1, C2,...], Oil compatibility matrix, ...}".

[0051] (3) Number of personnel D Each driver Data packets, such as "{ Qualification level, cumulative driving hours H_d, fatigue Status, security score S_d, ...}".

[0052] (4) Road network and regulations data: road segment risk level, restricted time / area, regulations threshold (such as maximum driving time), etc.

[0053] Step 2: Generation and encoding of initial solution data (1) Data processing action: The above input data are randomly or heuristically combined to form an initial scheduling scheme.

[0054] (2) Data encoding (structuring of key data): ① First layer (path encoding): Generates a sequence of data, for example, " [v1: d3 ->o5 ->o2, v2: d1 - >o4, ...] This data represents vehicle v1 (driven by the driver). d3 The order of delivery customers (by driver) is as follows: o5, o2 .

[0055] ② Second layer (loading code): Generates a mapping data that precisely records the specific vehicle bay location assigned to each order. For example, "{ o5: (v1, c2, 3000L), o2: (v1, c1, 5000L), ...} indicates an order. o5 5,000 liters of fuel were distributed to the vehicles. v1 Warehouse No. 2.

[0056] Step 3: Iterative Optimization and Data Evolution This step is repeated during algorithm iteration, continuously generating and optimizing new "solution data".

[0057] (1) Selection: Fitness score based on all solution data from the previous round (determined by the objective function) (Calculation), select the better "parent" solution data to proceed to the next stage.

[0058] (2) Cross: ① Data manipulation: Randomly select two "parent" solution data (such as two different path encoding sequences) and swap a portion of their subsequences.

[0059] ② Data impact: New "child" path encoding data is generated, which inherits some features of the parent, but may disrupt the original loading matching relationship and constraints.

[0060] (3) Variation: ① Data manipulation: Randomly make small perturbations to the encoding of a solution data (such as randomly swapping the order of two customers, or randomly reassigning an order to another vehicle).

[0061] ② Data impact: Introducing new search possibilities may also create new data violation points.

[0062] (4) Embedded compliance repair operator: ① Data scanning and detection: Systematically scan the newly generated "offspring" solution data after crossover / mutation, and verify it against all hard constraint rule bases.

[0063] ② Key data violation detection points: *Distributed storage capacity violation: Check the loading code data for violations in the same storage location. (v_k, c_j) Does the total allocation exceed its capacity data? .

[0064] *Fuel compatibility violations: Query the "Compatibility Matrix" in the vehicle data and check the fuel types assigned to adjacent compartments in the loading code data. t_i "Whether it is prohibited or not."

[0065] *Driver qualification violation: Compare driver ID in route coding data. d_m "and its personnel data" Capital Quality grade "Vehicle data" Model "and in the path data" Road risk level ".

[0066] ③ Data repair actions (examples): If a warehouse is found to be overloaded, the remaining capacity will be searched for in other warehouse data within the same vehicle, and the loading code will be adjusted to migrate the excess order data.

[0067] If a mismatch in qualifications is detected, select a qualified driver ID from the available driver data pool and replace the non-compliant driver ID in the original path code.

[0068] The repair process is local and limited in steps, designed to quickly restore the "feasibility" of the current solution data, rather than a global re-optimization.

[0069] (5) Fitness assessment: ① Data computation: For the repaired solution data, derive the complete scheduling scheme based on the encoding and decoding, and substitute it into the objective function. + + Perform the calculation.

[0070] ② Data output: Generate the fitness score of the solution data, which will be used for subsequent selection operations.

[0071] Step 4: Output the optimal solution data (1) Data convergence: The algorithm terminates when the preset number of iterations is reached or the quality of the solution data no longer improves.

[0072] (2) Data Decoding and Formatting: The final optimal path encoding and load encoding data are decoded and formatted into executable scheduling instruction data: ①" Personnel scheduling data " ②" Vehicle loading diagram data " ③" Delivery route map data " ④" Global Gantt chart data " II. Summary: Data Flow Characteristics of the Algorithm The "data-driven" nature of this algorithm is reflected in: 1. Input is data: All real-world constraints (physical, regulatory) are abstracted into structured data.

[0073] 2. Solution as Data: The scheduling scheme is encoded as a data structure that can be directly manipulated and evolved by the computer.

[0074] 3. Optimization, i.e., data processing: generating new data through operations such as selection, crossover, and mutation; cleaning and correcting data through repair operators; and evaluating data through evaluation functions.

[0075] 4. Compliance is embedded in data validation: The essence of the repair operator is an automated, rule-based data validation and correction procedure that ensures that the "solution data" flowing during the search process always evolves toward the "feasible domain".

[0076] The solution algorithm has the following characteristics: ① Two-layer coding design: The first layer is path coding, which represents the customer access sequence; the second layer is loading coding, which precisely describes which vehicle and which specific warehouse position each order is assigned to.

[0077] ② Fitness evaluation: using the above objective function Evaluate the merits and demerits of each scheduling scheme.

[0078] ③ Compliance Repair Operator: After the crossover and mutation operations in the genetic algorithm (or particle swarm optimization, ant colony optimization), a dedicated repair operator is introduced. This operator actively scans newly generated solution schemes. If it finds violations of hard constraints such as compartment capacity, fuel compatibility, or driver qualifications, it attempts to make adjustments and repairs within a limited number of steps (such as redistributing excess fuel to other empty compartments in the same vehicle, or re-matching drivers), rather than directly eliminating the solution. This greatly improves the efficiency and stability of the algorithm in searching the feasible solution space.

[0079] I. Core Functions and Positioning of Operators This operator is a local search and correction procedure specifically designed for constraint satisfaction. It does not function as an independent optimizer, but rather as a "solution quality filter and fine-tuner" for the main optimization algorithm (such as a genetic algorithm). Its input consists of "new solution data" that may contain violations after crossover and mutation operations, and its output is "feasible solution data" that satisfies all hard constraints after local correction.

[0080] II. Internal working steps and data processing flow of the operator The operator scans, diagnoses, and repairs the solution data in the following sequence: Step 1: Violation Scanning and Diagnosis (Data Detection Phase) 1. Input: The decoded data of the current individual (chromosome), i.e., the path encoding sequence and the loaded encoding mapping table.

[0081] 2. Action: The operator sequentially calls the following three "data scanners," each targeting a specific type of core hard constraint: (1) Warehouse capacity scanner: ① Data processing: Traverse the load encoding mapping table for each vehicle. v_k Each position c_j Aggregate the demand of all orders allocated to this warehouse. sum(q_i ).

[0082] ② Diagnostic rule: Compare with the rated capacity data of this storage unit. C_{kj} .like sum(q_i)>C_{kj} Then mark the (v_k, c_j) Designate it as an "overcapacity violation point" and record the overcapacity. overflow = sum(q_i) - C_{kj} .

[0083] (2) Oil compatibility scanner: ① Data processing: For each vehicle v_k Read the list of physically adjacent warehouse pairs and the oil compatibility matrix from its digital warehouse model. M (in M[t_a, t_b] = 1 Indicates oil products a and b Adjacent storage is allowed; =0 indicates that storage is prohibited.

[0084] ② Diagnostic rules: Check the loading code table to see which adjacent bays are loaded with the type of oil. (t_p, t_ q) Query M[t_p, t_q] If the result is 0, then the adjacent position pair is marked as a "compatibility violation point".

[0085] (3) Driver qualification matching scanner: ① Data processing: Traverse the path encoding and process each task segment. (d_m, v_k, route_segment) Extract driver d_m Qualification level data, vehicles v_k The minimum required qualification data, and path segments. route_segment The highest road risk level data.

[0086] ② Diagnostic Rules: Verification is performed based on a predefined "Qualification-Vehicle Type-Risk" matching rule table. For example, a rule might stipulate: "Transporting flammable liquids (such as gasoline) through tunnel areas (high-risk sections) must be carried out by a special vehicle driven by a driver with a Class A hazardous materials transport qualification." If the current data combination does not meet any rule, the task segment is marked as a "qualification violation point."

[0087] Step 2: Prioritization and Repair (Data Correction Phase) 1. Input: A list of all violations output in step A.

[0088] 2. Action: The operator is not repaired randomly, but is processed according to the principle of "severity first, coupling second".

[0089] (1) Sorting: Usually sorted by priority of "qualification violation > sub-account over-capacity > compatibility violation". Because qualification issues involve legal red lines, they must be resolved first.

[0090] (2) Specific repair strategies (examples): ① Regarding "qualification violations": Data manipulation: from the currently available driver data pool D_available In the middle, filter out all those that match the current... (v_ k, route_segment) The required set of drivers for the mission D_match .

[0091] Choose a strategy: From D_match Select Current cumulative driving time H_d The smallest driver (considering fatigue levels), replace the original route code. d_m .

[0092] Formula / Logic: New Driver d_new = argmin_{d in D_match} (H_d) Simultaneously update d_new The cumulative driving time data (plus the estimated time for this mission).

[0093] ② Regarding "overcapacity in sub-accounts": Data manipulation: Locating overcapacity positions (v_k, c_j) And the portion of orders exceeding the limit. In the same vehicle v_k Among the other positions, find a list of positions that meet the criteria of "remaining capacity" and "same oil type". Candidates .

[0094] Selection and migration strategy: If Candidates If not empty, then the excess order quantity will be... overflow Migrate to the candidate warehouse with the largest remaining capacity (or distribute evenly among multiple candidate warehouses) and update the loading code table.

[0095] Formula / Logic: Candidate Warehouse Filtering Criteria: C_{kl} - current_load_{kl}>= partial_ overflow and Oil type == t_i If no suitable candidate warehouse is found, "order reallocation" is triggered, removing the non-compliant order from the current vehicle loading table and placing it in the "pending order pool" to await reallocation by the algorithm in subsequent iterations.

[0096] ③ Regarding "violations on oil compatibility": Data manipulation: Locate incompatible adjacent position pairs and their respective oil products. (c_a: t_p, c_b: t_q) .

[0097] Exchange strategy: First try to find someone in the same vehicle who is compatible with yours. t_p Compatible storage space for placement t_q Or vice versa. This involves checking the current oil type and compatibility matrix in other compartments within the vehicle. M .

[0098] Formula / Logic: This is a local search process. Let the type of oil to be replaced be... t_q Target position is c_x The conditions are: 1) c_x Currently empty, or current oil quality. t_y and t_q Compatible ( M[t_y, t_q]=1); 2 ) c_x Not related to any storage and t_ q Incompatible oil products are stored in adjacent warehouses. If not found, it may be necessary to... t_q The entire order is then migrated to other vehicles and enters the "order reassignment" process.

[0099] Step 3: Repair Verification and Backtracking (Data Confirmation Phase) Action: After completing a set of repair operations, immediately rerun the scanner of step A on the local data involved in the repair (such as all the compartments in a vehicle) to ensure that the repair did not introduce new violations (for example, migrating orders to resolve overcapacity may lead to new compatibility issues).

[0100] Backtracking mechanism: If a repair action triggers a more complex chain of violations that cannot be resolved within a finite number of steps, the operator will abandon the repair attempt and record the solution as "unrepairable," allowing it to be naturally eliminated with lower fitness in subsequent algorithm selection stages. This ensures the computational efficiency of the operator and avoids getting stuck in meaningless local correction loops.

[0101] III. Summary 1. Data-driven: Automated diagnosis and correction based entirely on input coded data, attribute data (capacity, aptitude), and rule data (matrix, rule table).

[0102] 2. Locality: The repair scope is usually limited to a single vehicle, a single driver, or a few orders, resulting in low computational overhead.

[0103] 3. Heuristic: Repair strategies (such as selecting the warehouse with the largest remaining capacity) are designed based on domain knowledge, aiming to find feasible solutions quickly, rather than the global optimum.

[0104] 4. Embeddedness: It is called as a standard step in the algorithm's iterative loop to ensure the feasibility of new individuals in the population, thereby guiding the entire search process to focus on the feasible solution space. This is the key to improving the efficiency and stability of the algorithm.

[0105] ④ Iterative optimization: Through a cycle of selection, crossover, mutation, repair, and evaluation, the population is driven to evolve toward a better solution until the termination condition is met.

[0106] I. Detailed Steps and Data Processing Flow of Iterative Optimization Loop The entire iterative process is a data evolution system. Assume that in the... t There is a generation that includes N Population data of individual individuals Population(t) = [Ind_1, Ind_2, ..., Ind_N] Each individual Ind_i That is, a complete scheduling scheme consists of path-encoded data and load-encoded data.

[0107] Step 1: Select – Data Filtering (1) Input data: previous generation population Population(t-1) Fitness score data for all individuals (a list of values ​​calculated by the evaluation process; the lower the score, the better the solution).

[0108] (2) Data processing actions: Based on the fitness score, use roulette wheel selection or tournament selection methods to process the data. Population(t-1) Selecting by probability N Individual.

[0109] ① Roulette Wheel: The probability of each individual being selected is inversely proportional to its fitness. The algorithm generates a random number and "spins the roulette wheel" based on the probability distribution to determine which individual data is selected.

[0110] ② Tournament: Randomly selected from the population k The data from each individual is used to directly select the one with the best fitness.

[0111] (3) Output data: A new intermediate population consisting of superior parent individuals. Parents(t) This step does not create new data, but rather replicates and recombines existing data to ensure that superior genes are preserved.

[0112] Step 2: Cross-referencing – Data Reassembly (1) Input data: paired parent individual data (Ind_a, Ind_b) , from Parents(t) .

[0113] (2) Data processing actions (taking sequential cross OX as an example): ① Randomly select two intersection points on the path-encoded data (customer access sequence) of the two parents.

[0114] ② Swap the subsequences between the two intersections.

[0115] ③ Conflict handling: For duplicate customer IDs that appear after the exchange, they are corrected according to the sequence order of the other parent to ensure that each customer is accessed only once.

[0116] ④ Synchronous Update of Loading Code Data: After a route change, the original loading assignment may become invalid. In this case, the algorithm will execute a loading re-matching procedure: for each order on the new route, based on its fuel type and demand, a new warehouse space that meets capacity and compatibility constraints is found on the designated vehicle for binding, and the loading code table is updated. This process may initially result in violations.

[0117] (3) Output data: Data of two new offspring individuals Child_1, Child_2 They inherit the characteristics of their parents, but the encoded data has changed and may carry new violations.

[0118] Step 3: Mutation – Data Perturbation (1) Input data: Offspring individual data Child_i .

[0119] (2) Data processing actions: Randomly fine-tuning the encoded data of an individual with a small probability. Common operations include: ① Path variation: In path-coded data, the positions of two customers are randomly swapped; or a customer is randomly inserted into another position in the sequence.

[0120] ② Allocation variation: In the loading coded data, randomly select an order, remove it from the currently assigned vehicle slot, and then randomly or heuristically reassign it to a slot of another vehicle.

[0121] (3) Output data: new offspring data after perturbation Mutated_Child_i This introduces new search directions, but it also likely introduces new data violations (such as exceeding capacity limits or mismatched qualifications).

[0122] Step 4: Repair – Data Cleaning and Correction (1) Input data: After crossover and mutation, it may contain violations. Mutated_Child_i Complete encoded data.

[0123] (2) Data processing action: Invoking the compliance repair operator (whose internal steps have been detailed in the previous answer), which acts as a "data doctor": ① Scan Diagnosis: Check hard constraints such as compartment capacity, fuel compatibility, and driver qualifications one by one, and locate the violations in individual data.

[0124] ② Perform repair: Based on the preset repair strategy (such as migrating excess fuel, changing drivers, or exchanging fuel in the storage compartment), directly modify the loading code or path code data of the individual.

[0125] (3) Output data: Repaired feasible offspring data that satisfy all hard constraints Repaired_Child_ i This step ensures that the data flowing to the next stage in the population is "clean," firmly confining the search process to the feasible solution space.

[0126] Step 5: Evaluation – Data Scoring (1) Input data: repaired feasible individuals Repaired_Child_i Complete coded data, as well as all basic data (cost parameters, road network, time, etc.).

[0127] (2) Data processing actions: ① Decoding: Restore the encoded data into a computable scheduling scheme, including the specific route of each vehicle, the arrival / departure time of each station, the loading details of each compartment, driver and employee hours, etc.

[0128] ② Calculate the objective function: *calculate Cost_total The total cost is calculated based on data such as route distance, fuel consumption model, road and bridge tolls, vehicle and labor costs.

[0129] *calculate Time_penalty Iterate through each customer and compare their actual service time with their expected service window. [ET_i, LT_i] If you arrive early or late, follow the formula (e.g.) penalty = α * max(ET_i - arrival, 0) + β * max (arrival - LT_i, 0) Calculate and accumulate the penalty value.

[0130] *calculate Risk_score Combining route risk level data and driver fatigue data H_d and safety rating data S_d Through risk assessment models (such as risk = Σ(Road segment risk coefficient * Distance) * (1 + Fatigue coefficient) number) The safety risk value is then obtained.

[0131] ③ Weighted summation: Substitute into the formula F = w1*Cost_total + w2*Time_penalty + w3*Risk_ score .

[0132] (3) Output data: the fitness score of the individual. F_i (A scalar value). This score is the sole basis for the subsequent "selection" steps.

[0133] Step 6: Population Update and Termination Judgment – ​​Data Iteration (1) Input data: All repaired and evaluated offspring individuals of the new generation [Repaired_Child_1, ..., Repaired_Child_N] and its fitness score.

[0134] (2) Data processing action: The old parent population is completely or partially replaced by a new offspring population to form a new generation population. Population(t) .

[0135] (3) Termination condition judgment: ① Maximum number of iterations: number of iterations t Reaching the preset limit T_max (e.g., generation 1000). (Process: counter judgment) ② Convergence criterion: The optimal fitness score of a population over multiple generations (e.g., 50 generations). min(F_i) No further improvement is allowed (changes are less than the threshold ε). (Process: Compare with historical best fitness data) ③ Time limit: The algorithm runs for longer than the preset time.

[0136] ④ Satisfactory solution: An individual with a fitness score lower than the predetermined target value was found.

[0137] Step 7: Loop If no termination condition is met, then let t = t + 1 ,Will Population(t) As input, jump to step 1 (selection) and begin the next iteration.

[0138] II. Summary This iterative process is a "generate-test-fix-evolve"* data processing pipeline. It works through: 1. Crossover / mutation is used to exploratoryly reconstruct the solution data (encoding).

[0139] 2. The repair operator performs immediate compliance rework on "scrap" data generated during the reconstruction.

[0140] 3. Evaluate and score the qualified product data (feasible solutions) through quality inspection.

[0141] 4. Select which product data qualifies as a "template" for the next generation of production based on quality scores.

[0142] 5. Ultimately, this cycle drives the entire population data to evolve generation after generation towards regions with lower fitness function values ​​(i.e., lower overall cost, penalty, and risk) until optimal or near-optimal scheduling scheme data that meets the requirements is produced.

[0143] 5. Output scheduling scheme The algorithm outputs the final solution, including: ① Staffing schedule: Vehicles, task sequence, and estimated working hours for each driver; ② Vehicle loading diagram: Types, weights, and delivery order of oil in each compartment of each vehicle; ③ Delivery route map: Detailed driving route for each vehicle and estimated arrival / departure time at each stop; ④ Global Gantt Chart: Visualizes the collaborative work of all resources over time.

[0144] The system outputs a feasible solution that balances cost, efficiency, and safety. For example, the system might select a route that is not the shortest but has lower risk and moderate cost, and match it with a qualified but not the most expensive (experienced) driver, while simultaneously improving vehicle utilization by optimizing loading. This is precisely the result of achieving a balance among multiple objectives by adjusting weights and algorithmic search.

[0145] More specific details are as follows: The algorithm's output module takes as input the complete encoded data of the optimal individual (or a set of Pareto optimal individuals) obtained after the iterative optimization terminates, i.e.: Optimal route coding: a sequence that defines the order in which each vehicle (and its associated driver) visits the customer.

[0146] Optimal loading code: a precise mapping that defines which vehicle and which bay will carry each order.

[0147] The core task of the output module is to "translate" this coded data into various forms and charts that business personnel can directly understand and execute. The specific process is as follows: I. Output ①: Generation of personnel shift schedule 1. Data processing source: Optimal path encoding.

[0148] 2. Generation steps: ① Decoding path: Traverse the path encoding and group by vehicle / driver. For example, the parsing result is "Driver D001 drives vehicle V001 and visits stations S005, S002, and S008 in sequence".

[0149] ② Extract task sequence: For each driver, generate a specific task list based on their route and order data: [Start from the oil depot -> proceed to S005 to unload 92# gasoline from warehouse A -> proceed to S002 to unload 0# diesel from warehouse B -> unload warehouse C from S008] 92# gasoline -> Return to oil depot] .

[0150] 3. Calculate the estimated working hours: ① Travel time: Call the route planning engine (such as Dijkstra's or A* algorithm based on road network data) to calculate the travel distance and estimated time for each segment of the journey in the route coding.

[0151] ② Operation time: Based on data such as the unloading volume and unloading port efficiency of each station, the loading and unloading operation time is estimated.

[0152] ③ Summary: Estimated working hours = Departure preparation time + Σ Travel time for each segment + Σ Operation time for each station + Handover time upon return .

[0153] 4. Formatted output: Fill the above information into the standard form template to generate a schedule for each driver, including the following fields: driver name / ID, assigned vehicle, task sequence (stations and actions), planned departure time, planned return time, estimated total working hours, and key time nodes.

[0154] II. Output ②: Generation of vehicle loading diagram 1. Data processing source: Optimal loading code and vehicle digital compartment model data.

[0155] 2. Generation steps: ① Aggregate loading information by vehicle: Query the loading code by vehicle to obtain a list of orders allocated to all warehouse slots of that vehicle.

[0156] ② Binding the delivery sequence: Based on the vehicle's route code (the order in which stations are visited), determine at which station the oil in each compartment will be unloaded. This forms the core data chain of "compartment-oil-weight-unloading station".

[0157] ③ Graphical rendering: * Call the digital outline template of the vehicle compartment.

[0158] * Fill the warehouse map with different colors or patterns based on the loading data (oil type, weight) for each warehouse.

[0159] *Illustrated next to the position or via legend: Storage location number, oil type, loading capacity (t / L), destination (station name) say) .

[0160] ④ Generate a diagram: Output an intuitive diagram so that the loading operator can clearly see what should be loaded into each compartment of each vehicle, how much should be loaded, and where to unload it.

[0161] III. Output ③: Generation of delivery route map 1. Data processing source: optimal path coding and geographic information system (GIS) road network data.

[0162] 2. Generation steps: ① Detailed route planning: For each vehicle's route code (station sequence), a high-precision route planning algorithm is called to calculate the optimal feasible driving route between two points by combining real-time traffic conditions (such as congestion) and regulations (such as the prohibition of hazardous chemicals), rather than just the station order.

[0163] ② Time node calculation: *Known quantities: planned departure time, travel time for each segment of the route, and operation time for each station.

[0164] *Recursive calculation: Time to arrive at station n = Time to leave station n-1 + Travel time from n-1 to n; Time to leave station n = Time to arrive at station n + Operation time at station n.

[0165] ③ Integration and visualization: * The calculated detailed path (consisting of a series of road node coordinates) is plotted on the map as a line with arrows.

[0166] * Mark time information at key points (stations, turning points) along the route.

[0167] * Generate a separate route map for each vehicle, or distinguish different vehicles using different colored lines on a single master map.

[0168] ④ Output content: Map image file, and an accompanying route details table, including: vehicle ID, sequence, stops along the way, estimated arrival time (ETA), estimated departure time (ETD), route mileage, and critical path indication.

[0169] IV. Output ④: Generation of the global Gantt chart 1. Data processing source: Delivery route time data and personnel shift schedule data for all vehicles.

[0170] 2. Generation steps: ① Unified Timeline: Establish a time coordinate system with absolute time (such as from 0:00 to 24:00 on a certain day) as the horizontal axis.

[0171] ② Define the resource axis: The vertical axis lists all resources that need to be coordinated, which usually includes: all drivers and all vehicles (and sometimes also key site unloading points).

[0172] ③ Draw time blocks: *For drivers: Based on their shift schedule, draw a continuous task bar on their corresponding resource line from "planned departure time" to "planned return time".

[0173] For vehicles: Draw more granular time blocks. Divide the vehicle timeline into segments of different colors: travel segments (from A to B), operation segments (unloading at station C), and waiting segments. This requires integrating detailed route schedules for each vehicle.

[0174] ④ Relationships and Visualization: Gantt charts clearly show: *Which driver was driving which vehicle at what time?

[0175] *The status of each vehicle at any given time (driving, working, idle).

[0176] *Do the operating times of different vehicles at the same station conflict (overlapping longitudinal time blocks)?

[0177] * The compactness and resource utilization of the overall operation.

[0178] For example, in a multi-variety transportation task involving 95-octane gasoline, 0# diesel, and kerosene, the model automatically executes triple dynamic constraints. This not only accurately matches the loading capacity requirements of each oil product but also detects safety risks associated with mixing gasoline and kerosene in adjacent storage compartments and automatically adjusts the compartments accordingly. Simultaneously, it ensures that the unloading sequence at the three unloading points along the route is consistent with the compartment layout. The multi-dimensional compliance dynamic matching engine for "people-vehicle-goods-road" retrieves in real-time the validity period of the driver's hazardous materials transportation qualification, the vehicle's tank inspection status, the flammability and explosiveness hazard level of the oil products, and the bridges along the transportation route. Information such as load capacity and restricted hours directly intercepted a dispatch request from a driver whose license had expired, and avoided the planned high-risk congestion section. Meanwhile, the improved heuristic algorithm with embedded compliance repair operators proactively detected a dual problem in the initial generated dispatch plan: a vehicle had a conflict in the category of goods stored in the warehouse and a mismatch between the driver's qualifications and the grade of the goods. Through automatic repair by the operators, a feasible solution that balances loading efficiency and full-process compliance was output within 30 seconds. Compared with the manual adjustment mode of traditional algorithms, the solution generation efficiency was improved by 60%, and the violation rate was reduced to 0.

[0179] Compared with existing general route planning technologies, this invention achieves a qualitative improvement for refined oil product delivery scenarios: (1) Safety and compliance are prioritized: By embedding hard constraints such as compartment compatibility, driver qualifications, and fatigue driving regulations into the core of the optimization algorithm, the generation of illegal scheduling schemes is eliminated from the source, thus achieving inherent safety; (2) Maximize resource utilization efficiency: Through precise matching and optimization of "warehouse-order", the utilization rate and loading rate of vehicle warehouses are significantly improved, empty driving mileage and empty warehouse waste are reduced, and the overall logistics cost can be reduced by more than 10%; (3) Enhanced operational synergy: Guided by "delivery time optimization", the route, loading, and manpower were coordinated, reducing customer waiting time and site operation conflicts, and improving the overall supply chain synergy efficiency; (4) Intelligent and automated scheduling: It overcomes the shortcomings of manual scheduling, such as long time consumption, reliance on experience, and difficulty in global optimization. It can quickly process large-scale orders and complex constraints, and dynamically respond to emergencies.

[0180] As described above, the present invention can be implemented well.

[0181] In the description of this invention, the terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of indicated technical features. Thus, a feature defined as "first" or "second" may explicitly or implicitly include at least one of that feature. In the description of this invention, "a plurality of" means at least two, such as two, three, etc., unless otherwise explicitly specified.

[0182] In the description of this invention, the terms "one embodiment," "some embodiments," "example," "specific example," or "some examples," etc., refer to specific features, structures, materials, or characteristics described in connection with that embodiment or example, which are included in at least one embodiment or example of the invention. In this specification, the 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. Moreover, without contradiction, those skilled in the art can combine and integrate the different embodiments or examples described in this specification, as well as the features of different embodiments or examples.

[0183] In the description of this invention, although embodiments of the invention have been shown and described herein, it is to be understood that the above embodiments are exemplary and should not be construed as limiting the invention. Those skilled in the art can make changes, modifications, substitutions and variations to the above embodiments within the scope of this invention.

[0184] In the description of this invention, all features disclosed in all embodiments of this specification, or steps in all methods or processes implied in the disclosure, may be combined and / or extended or replaced in any way, except for mutually exclusive features and / or steps.

[0185] The above description is merely a preferred embodiment of the present invention and is not intended to limit the present invention in any way. Based on the technical essence of the present invention, any simple modifications, equivalent substitutions, and improvements made to the above embodiments within the spirit and principles of the present invention shall still fall within the protection scope of the present invention.

Claims

1. A method for route planning of dangerous goods transport vehicles, characterized in that, Includes the following steps: Acquire resource data; the resource data includes: fuel data, vehicle data, and driver data; Construct a route planning model with transportation cost, transportation time, and transportation safety risks as objectives. The formula for the objective route planning model is: + + ; in, This represents the minimization operation. Represents the objective function value. Indicates the first weight. Indicates transportation costs, Indicates the second weight. Indicates the transportation time. Indicates the third weight. Indicates the total security risk; Define the constraints for the path planning model; The constraints include: The amount of fuel allocated to the vehicle's storage compartment is less than the compartment's rated capacity; The fuel types in adjacent compartments of the vehicle are compatible; Driver qualifications must be matched with fuel type; Solve the path planning model, and take the path corresponding to the solution of the path planning function that minimizes the objective function value as the optimal path.

2. The method for route planning of dangerous goods transport vehicles according to claim 1, characterized in that, Solving the path planning model, and taking the path corresponding to the solution of the path planning function that minimizes the objective function value as the optimal path, includes the following steps: Resource data is combined and encoded and input into a multi-objective optimization model to generate an initial solution for the path planning function; The resource data combination is changed and encoded as input into the multi-objective optimization model, and the initial solution of the initial path planning is iterated to generate a new solution; All solutions are scored for fitness, and the path corresponding to the solution with the highest fitness is selected as the optimal path.

3. The method for route planning of dangerous goods transport vehicles according to claim 2, characterized in that, Resource data is combined and encoded and input into a multi-objective optimization model. The initial solution of the initial path planning is iterated to generate a new solution, including the following steps: Randomly combine resource data, encode the combined resource data, and generate path codes and / or load codes; Among them, the route code stores driver data and candidate route data, while the loading code stores the warehouse data of the order being assigned to the vehicle.

4. The method for route planning of dangerous goods transport vehicles according to claim 3, characterized in that, The process involves changing the resource data combination and encoding it into a multi-objective optimization model, iterating through the initial solution of the initial path planning to generate a new solution, including the following steps: Two solutions are randomly selected, and all or part of the resource data corresponding to the two solutions are swapped. The resource data is then combined and encoded and input into the multi-objective optimization model. The initial solution of the initial path planning is iterated to generate a new solution.

5. The method for route planning of dangerous goods transport vehicles according to claim 4, characterized in that, The process involves changing the resource data combination and encoding it into a multi-objective optimization model, iterating through the initial solution of the initial path planning to generate a new solution, including the following steps: The resource data corresponding to a certain solution is randomly modified. The modified resource data is then combined and encoded and input into the multi-objective optimization model. The initial solution of the initial path planning is iterated to generate a new solution.

6. The method for route planning of dangerous goods transport vehicles according to claim 5, characterized in that, The process involves changing the resource data combination and encoding it into a multi-objective optimization model, iterating through the initial solution of the initial path planning to generate a new solution, including the following steps: Check if all new solutions meet the constraints: if yes, accept the new solution; if no, verify the new solution. The methods for verifying the new solution include one or more of the following: If a new solution is detected that does not meet the constraint that the amount of oil in the compartment allocated to the vehicle is less than the compartment's rated capacity, then the remaining capacity is searched in other compartments of the same vehicle, the loading code is updated, and the excess oil data is migrated. If a new solution is detected that does not meet the constraint that the oil type is compatible with the adjacent compartments of the vehicle, the order is prohibited from being assigned to one or two of the adjacent compartments of the vehicle, and the loading code data is updated to prohibit the identification of the oil type assigned to the adjacent compartments. If a new solution is detected that does not meet the constraint of matching driver qualification with fuel type, driver data that meets the driver qualification requirements is selected from the available driver data pool, and the selected driver data replaces the driver data in the path code.

7. The method for route planning of dangerous goods transport vehicles according to claim 6, characterized in that, If a new solution is detected that does not meet the constraint that adjacent compartments of the vehicle have compatible oil types, the vehicle should first look for a compartment with compatible oil types within the same vehicle to store the oil.

8. The method for route planning of dangerous goods transport vehicles according to claim 1, characterized in that, It also includes the following steps: Output one or more of the following based on the optimal path: Personnel Shift Schedule: Each driver's assigned vehicle, task sequence, and estimated working hours; Vehicle loading diagram: Types, weights, and delivery order of oil in each compartment of each vehicle; Delivery route map: the route of each vehicle, the estimated arrival time at each station, and the estimated departure time at each station; Global Gantt Chart: Visualizes the collaborative operation of all resource data over time.

9. A method for route planning of dangerous goods transport vehicles according to any one of claims 1 to 8, characterized in that, The constraints also include one or more of the following: The driver's qualifications must match the type of vehicle assigned and the risk level of the roads traversed. Drivers must not exceed the legal limit for continuous driving time; When planning delivery routes, it is essential to ensure that the order of visiting stations matches the order of unloading oil at each storage location. The vehicle departs from the oil depot and eventually returns; Customers are served within the designated time window; Overloading is not allowed.

10. A route planning system for dangerous goods transport vehicles, characterized in that, The method includes a processor, a memory, and a computer program stored in the memory. When the processor executes the computer program, it implements the steps of the method for route planning of dangerous goods transport vehicles as described in any one of claims 1 to 9.