Priority-based multi-task scheduling method and system for shared travel

Abstract

The invention discloses a shared trip-oriented priority-based multi-task scheduling method and system, and the method comprises the steps: constructing a node set containing supply nodes and demand nodes and a task set of multiple types of tasks, configuring truck parameters, calculating the driving time between the nodes, defining the net demand quantity of the nodes, and judging the types of the nodes. The method comprises the following steps of: calculating the net demand quantity, distributing task priorities, setting a unit reward value, constructing a mixed integer programming model integrating multiple constraints, solving and outputting a scheduling scheme by adopting a customized operator algorithm, and refining sub-steps of calculating the net demand quantity, distributing the priorities and constructing the model so as to optimize a process; and a system matching method function module supports ordered operation of each link. According to the method and the system, accurate adaptation of multiple tasks and priorities is realized, the task response efficiency and the scheme feasibility are improved, and the resource allocation rationality and the overall efficiency of shared travel operation are further improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of shared travel multi-task scheduling, and in particular to a priority-based multi-task scheduling method and system for shared travel. BACKGROUND

[0002] With the rapid development of the shared travel industry, the deployment scale of shared bicycles, shared electric vehicles and other travel tools continues to expand, and multiple types of tasks emerge in parallel during operation, including normal vehicle rebalancing, faulty vehicle repair and low battery vehicle battery replacement. Such tasks need to rely on trucks and other transportation tools to allocate resources between different nodes, and there are differences in supply and demand at each node, and the urgency of tasks is different. If there is a lack of scientific scheduling mechanism, it is easy to lead to unbalanced distribution of resources, delayed response to tasks, and affect user travel experience and operational efficiency. At the same time, the increase in the number of shared travel system nodes and the complexity of task types put higher requirements on the accuracy and dynamic adaptability of the scheduling scheme, and it is urgent to build a scheduling method and system that takes into account multiple task types and priorities to optimize resource allocation and improve overall operational efficiency.

[0003] The prior art has two significant shortcomings in the aspect of shared travel multi-task scheduling: on the one hand, most scheduling methods do not fully integrate multi-type tasks and priority mechanisms, and only design scheduling strategies for a single task type, which cannot reasonably allocate resources according to the urgency of tasks, leading to difficulties in timely processing of critical tasks such as urgent fault repair, high-demand node vehicle replenishment, and reducing system response efficiency; on the other hand, the existing scheduling model has deficiencies in constraint condition consideration and solution optimization, and some models do not fully consider actual operational constraints such as truck capacity, node processing capacity and time window, or the solution algorithm lacks customization optimization, making it difficult to generate a scheduling scheme that takes into account feasibility and optimality, and the node type is not dynamically determined and the resource net demand is not updated in real time, which cannot adapt to the dynamic characteristics of the shared travel system. SUMMARY

[0004] In order to overcome the shortcomings and deficiencies of the prior art, the present application provides a priority-based multi-task scheduling method and system for shared travel.

[0005] The technical scheme adopted by the present application is a priority-based multi-task scheduling method for shared travel, comprising the following steps: S1. Constructing a node set and a task set of a shared travel system, the node set comprising a supply node, a normal demand node for vehicle rebalancing operation, and a node capable of performing multi-functional tasks, and the task set comprising normal shared bicycle rebalancing, normal shared electric vehicle rebalancing, fault shared bicycle maintenance, fault shared electric vehicle maintenance, and low-battery shared electric vehicle battery replacement tasks; S2. Configuring parameters of a truck set, the parameters comprising the capacity and driving speed of each truck, determining the driving distance of each truck between nodes, and calculating the driving time of each truck between any two nodes; S3. Defining the net demand of each node for different types of tasks, judging whether the node is a demand node or a supply node according to the positive or negative value of the net demand, and recording the initial net demand of different types of tasks; S4. Assigning priority levels to different types of tasks, the priority level corresponding to the urgency of the task, and setting a unit reward value for each type of task, the unit reward value being positively correlated with the priority level; S5. Constructing a mixed integer programming model, integrating multi-commodity flow, multi-visit vehicle rebalancing, fault maintenance, battery replacement, and priority mechanism into the model, the model comprising truck capacity constraints, node capacity constraints, and time window constraints; S6. Solving the mixed integer programming model by using an adaptive large neighborhood search heuristic algorithm, designing a customized operator to optimize the solving process, and outputting a scheduling scheme for different types of tasks, the scheduling scheme comprising truck driving paths, task execution sequences, and resource transportation amounts.

[0006] Further, in the step S5 of constructing the mixed integer programming model, a task priority weight coefficient is introduced to calculate the objective function, and the expression of the objective function is: , is the priority weight coefficient of task r, and increases with the increase of the priority level, is a 0-1 variable, represents that truck k passes through arc transports task , and vice versa, is the reward coefficient of task r, is the completion identifier of task r, represents that task r is completed, , and vice versa; meanwhile, a resource balance constraint formula is added to the model: , is the resource amount of truck k transporting task r from node i to node j, is the initial net demand of node i for task r.

[0007] Further, the iteration update process of the adaptive large neighborhood search heuristic algorithm in the step S6 adopts the following formula: , wherein, is the current solution of the tth iteration, is the solution of the tth iteration, is the solution of the tth iteration, is the new solution generated by the neighborhood search, is the objective function, is the temperature parameter of the tth iteration, is the random number within the interval; the selection probability formula of the customized operator in the algorithm is: , is the selection probability of the operator o, is the initial weight of the operator o, is the historical performance coefficient of the operator o, is the operator set.

[0008] Further, the dynamic update of the node net demand in S3 adopts the formula: , is the net demand of node i for task r at time t, is the net demand of node i for task r at time , is the resource amount transported by truck k from node i to node j for task r at time t, is the resource loss coefficient of node i to node j for task r at time t, ; at the same time, the node type judgment threshold formula is: , is the type judgment threshold of node i for task r, is the Kronecker function, when when .

[0009] Further, the calculation of the task priority level in S4 adopts the formula: , is the weight of the mth influence factor, and is the standardized value of task r under the mth influence factor, is the total number of influence factors; the determination formula of the unit reward value is: , is the basic reward value, is the reward coefficient, .

[0010] Further, the time window constraint of the model in S5 adopts the formula: , is the earliest start time of node i to execute task r, is the latest end time of node i to execute task r, Process task for node i Time, 0-1 variable, denotes the processing of task r for node i before the processing of task r for node i, then the opposite; the truck capacity constraint formula is: , is the load factor of truck k on arc , , is the capacity of truck k.

[0011] Further, the S3 comprises the following steps: S31. Collecting historical data of different types of tasks of each node in a preset time period, the historical data including the number of task generation, the number of task completion and the number of resource remaining, screening the collected historical data, and eliminating abnormal value data, the abnormal value data being determined by a standard of deviating from three times of standard deviation of the average value; S32. Calculating resource inflow and resource outflow of each node for different types of tasks according to the screened historical data, the resource inflow being the total amount of resources related to different types of tasks received by the node, and the resource outflow being the total amount of resources related to different types of tasks output by the node; S33. Subtracting the resource inflow from the calculated resource outflow to obtain the initial net demand of each node for different types of tasks, and recording the numerical value of the initial net demand and the corresponding task type and node identification; S34. Comparing the initial net demand of each node for different types of tasks with the preset node resource threshold, and preliminarily judging the possible type of the node according to the comparison result, to provide node attribute basis for scheduling scheme.

[0012] Further, the S4 comprises the following steps: S41. Analyzing factors affecting the urgency of tasks of the shared travel system, the factors including task waiting time, node demand gap size, and task influence range on user travel, setting evaluation grades for each influencing factor, and each evaluation grade corresponding to specific quantitative standards; S42. Organizing professional personnel to score the performance of different types of tasks under each influencing factor according to the set quantitative standards, taking the average value of the scores of different professional personnel to obtain the final score of different types of tasks under different influencing factors; S43. Assigning corresponding weight values to different influencing factors according to their importance, determining the weight values by the analytic hierarchy process, multiplying the final scores of different types of tasks under different influencing factors by the corresponding weight values, and summing to obtain the priority score of different types of tasks; S44. Dividing different priority levels according to the distribution range of the priority score, mapping the priority score with the priority level, completing the allocation of the priority level of different types of tasks, and determining the corresponding unit reward value according to the priority level.

[0013] Further, the S5 comprises the following sub-steps: S51. Explicitly defining the decision variables of the mixed integer programming model, the decision variables including truck driving path variables, task execution order variables and resource transportation amount variables, defining the value range and physical meaning of each decision variable, and ensuring that the decision variables can accurately describe the calibration elements in the scheduling process; S52. Analyzing the constraint conditions in the operation process of the shared travel system, the constraint conditions including truck driving time constraints, node processing capacity constraints and resource supply and demand balance constraints, and converting each constraint condition into a mathematical expression, the mathematical expression including corresponding variables and parameters; S53. Constructing the objective function of the model, the objective function being oriented to minimize the system operation cost and maximize the task priority benefit, integrating the cost items and benefit items into a unified mathematical formula, the cost items including transportation costs and loading and unloading costs, and the benefit items including task completion rewards; S54. Testing the constructed mathematical model, the testing contents including the integrity of variable definition, the rationality of constraint conditions and the logic of objective function, correcting the problems found in the testing, and ensuring that the model can accurately reflect the actual scheduling demand.

[0014] The application discloses a priority-based multi-task scheduling system for shared travel, which is applied to a priority-based multi-task scheduling method for shared travel and comprises a task and node information collection and processing unit, a truck parameter configuration and path calculation unit, a net demand calculation and node type judgment unit and a task priority allocation and reward setting unit.

[0015] Beneficial effects: The application provides a priority-based multi-task scheduling method and system for shared travel, which constructs a multi-task set including normal vehicle rebalancing, fault maintenance and battery replacement, combines node net demand calculation and priority allocation mechanism, realizes deep integration of multi-type tasks and priorities, effectively overcomes the shortcomings of the prior art that only designs strategies for single tasks and cannot allocate resources according to the degree of urgency, accurately judges the supply and demand properties of each node, determines the degree of task urgency, and then allocates resources to key tasks such as emergency fault maintenance and high-demand node resource replenishment through priority correlation unit reward values, greatly improves the response speed and task processing accuracy of the system; at the same time, the mixed integer programming model is constructed by integrating actual operation constraints such as multi-commodity flow, vehicle capacity and time window, and solved by the adaptive large neighborhood search algorithm with customized operators, and relying on the node type dynamic judgment, resource net demand real-time update and scheduling scheme dynamic adjustment mechanism, the shortcomings of the prior art that the model constraints are not comprehensive, the solving algorithm lacks optimization and is difficult to adapt to the dynamic changes of the system are overcome, the comprehensive constraint conditions ensure the feasibility of the scheduling scheme, the optimized solving algorithm improves the optimality of the scheme, the dynamic adjustment mechanism enables the system to adapt to the changes of node supply and demand and task demand in real time, and finally realizes reasonable allocation of resources, improves the task execution efficiency, and balances the operation cost control and user travel experience improvement. BRIEF DESCRIPTION OF DRAWINGS

[0016] Figure 1 The method steps flowchart of the application; Figure 2 The system unit composition diagram of the application. DETAILED DESCRIPTION

[0017] It should be noted that the embodiments in the present application and the features in the embodiments can be combined with each other without conflict, and the present application will be further described in detail below in combination with the drawings and specific embodiments.

[0018] As shown in Figure 1 A priority-based multi-task scheduling method for shared travel, comprising the following steps: S1. Constructing a node set and a task set of a shared travel system, the node set including a supply node, a normal demand node for vehicle rebalancing operation and a node capable of performing multi-functional tasks, and the task set including normal shared bicycle rebalancing, normal shared electric vehicle rebalancing, fault shared bicycle maintenance, fault shared electric vehicle maintenance and low-battery shared electric vehicle battery replacement tasks; Specifically, S1 constructs the node set and task set of the shared travel system, providing a basic data framework for the subsequent scheduling process, which is implemented in two directions according to node type and task type. When constructing the node set, first define the parameter standards of the three types of nodes: the supply node is located in the warehouse area or transportation hub on the edge of the city, the single node occupies an area of 200 to 500 square meters, can accommodate 50 to 200 shared bicycles (single vehicle occupies an area of 0.8 to 1.2 square meters) or 30 to 150 shared electric vehicles (single vehicle occupies an area of 1.5 to 2 square meters), and is equipped with 20 to 50 sets of backup batteries (each set of battery weighs 15 to 25 kilograms); the normal demand node is set at the entrance of the residential area or the pedestrian street in the commercial area, the single node occupies an area of 50 to 150 square meters, has no resource storage function, can receive 10 to 30 shared bicycles or 5 to 20 shared electric vehicles at a time, and needs to reserve 1.2 times the vehicle parking space during receiving; the multifunctional task node is located on the side of the urban secondary road, the single node occupies an area of 150 to 300 square meters, in addition to accommodating 20 to 80 shared bicycles, 15 to 60 shared electric vehicles and 10 to 30 sets of backup batteries, it also needs to divide a 20 to 50 square meter maintenance area and is equipped with 1 to 3 sets of fault detection equipment (each set of equipment has a power of 500 to 1000 watts) and 1 to 3 persons holding maintenance qualification certificates. The execution parameters of the five types of tasks in the task set are as follows: the normal shared bicycle rebalancing task has a single dispatch quantity of 5 to 20 vehicles, and the vehicle fixing device spacing is 0.5 meters during dispatch; the normal shared electric vehicle rebalancing task has a single dispatch quantity of 3 to 15 vehicles, and the vehicle needs to use a special fixing support (each support has a bearing capacity of 30 to 50 kilograms); the fault shared bicycle maintenance task targets the brake system, tires and other components, and the single vehicle maintenance time is 15 to 40 minutes, tools such as wrenches and air pumps are needed; the fault shared electric vehicle maintenance task targets the battery, motor and other components, and the single vehicle maintenance time is 20 to 50 minutes, devices such as multimeters and charging piles are needed; the low battery shared electric vehicle battery replacement task targets vehicles with less than 20% of the remaining battery, and the single vehicle replacement time is 3 to 8 minutes, and insulated gloves are needed during replacement. By clearly defining the above parameters, the node and task data can be directly used for subsequent scheduling calculation.

[0019] S2. Configure the parameters of the truck set, including the capacity and driving speed of each truck, determine the driving distance between each node, and calculate the driving time of each truck between any two nodes; Specifically, step S2 performs truck set parameter configuration and inter-node travel time calculation, which is the key to ensure that the scheduling scheme meets the actual transportation capacity. The implementation process needs to be carried out in two parts: parameter setting and time calculation. Truck parameter configuration needs to determine the core indicators according to the type of transportation task: for shared bicycle transportation, the truck interior dimensions are 6 to 9 meters long, 2.3 to 2.5 meters wide, and 2 to 2.5 meters high, and can load 30 to 80 vehicles at a time (based on the area occupied by a single shared bicycle of 0.8 to 1.2 square meters, with a 10% gap), and is equipped with 3 to 5 layers of adjustable shelves (each layer can bear 500 to 800 kg); for shared electric vehicle transportation, the truck interior dimensions are the same as those for shared bicycle transportation, but a special fixed slot needs to be installed (each slot is 1.8 to 2.2 meters apart), and can load 20 to 50 vehicles at a time, while a 10 to 20 square meter battery storage area is set up at the back of the truck, equipped with 20 to 40 groups of battery fixing racks (each group can bear 20 to 30 kg); the driving speed of all trucks needs to be set according to the road type, at 40 to 60 km / h on urban trunk roads (speed limit 60 km / h), and at 20 to 30 km / h in congested areas such as residential areas and commercial areas (speed limit 40 km / h), and each truck needs to be equipped with a GPS positioning device (positioning accuracy 1 to 5 meters) and a real-time road condition receiving module. The inter-node travel distance calculation needs to rely on the geographic information system, first collect the latitude and longitude coordinates of each node (accuracy to 6 decimal places), then calculate the straight-line distance based on the coordinates, and then make corrections based on the actual road network: if there is a direct trunk road between the two nodes, the correction coefficient is 1.05 to 1.1; if it needs to be detoured through secondary roads or residential area roads, the correction coefficient is 1.2 to 1.5. The driving time calculation is based on the corrected driving distance and the set speed, using the method of basic time plus redundancy time: basic time = corrected driving distance / driving speed, and redundancy time is 10% to 20% of the basic time (adjusted according to real-time road conditions, 20% during congestion and 10% during smooth traffic). For example, the corrected driving distance is 6 km, the driving speed on the trunk road is 40 km / h, the basic time is 9 minutes, and the redundancy time is 1.8 minutes, so the final driving time is determined to be 10.8 minutes. Through accurate calculation, it is ensured that each transportation time can be directly used for time window planning in the scheduling scheme.

[0020] S3. Define the net demand amount of each node for different types of tasks, judge whether the node is a demand node or a supply node according to the positive and negative values of the net demand amount, and record the initial net demand amount of different types of tasks; Specifically, S3 defines the node net demand and judges the node type, clarifies the system resource supply and demand distribution, implements the promotion according to the data collection, calculation and judgment three links, and each link clarifies the specific numerical standard. The data collection link needs to determine the statistical time period and data type: the statistical time period is set to 1 hour (from the whole point, such as 10:00-11:00), and the resource inflow and outflow of each node in the time period are collected for each type of task - the inflow includes the number of shared travel tools transported to the node by trucks, the number of backup batteries, the outflow includes the number of shared travel tools ridden away from the node by users, the number of shared travel tools taken away by trucks, the number of fault vehicles leaving the node after repair, and the number of old batteries returned after replacement, data collection needs to be recorded in real time through the node intelligent terminal (2-4 terminals are equipped for each node, with a sampling frequency of 1 time / minute), and the error needs to be controlled within ±2 vehicles (or groups). The net demand calculation link needs to be calculated according to the type of task: for normal shared bicycle rebalancing tasks, the net demand = the outflow of shared bicycles in the node in the time period - the inflow; the calculation logic of normal shared electric vehicle rebalancing tasks, fault shared bicycle maintenance tasks, fault shared electric vehicle maintenance tasks, and low-capacity shared electric vehicle battery replacement tasks is the same, only the corresponding resource type (such as shared electric vehicle, backup battery) is replaced, and the calculation result needs to be kept to the integer place, if a decimal appears, round up (for example, if the calculation result is 8.2, the net demand is 9). The node type judgment link needs to be based on the net demand value: when the net demand is positive, it is determined as a demand node, indicating that the resource demand of this node for this type of task is greater than the supply, and resources need to be allocated from the outside; when the net demand is negative, it is determined as a supply node, indicating that the resource supply of this node for this type of task is greater than the demand, and resources can be output to the outside; At the same time, the specific value of the net demand, the corresponding task type and the statistical time period need to be recorded, and the recorded data needs to be synchronized to the system database (storage frequency 1 time / hour), which provides the basis for subsequent resource scheduling allocation. Through the standardized calculation and judgment process, the node supply and demand status is clear and traceable.

[0021] S4. Assign priority levels to different types of tasks, with priority levels corresponding to task urgency, and set a unit reward value for each type of task, with the unit reward value positively correlated with the priority level; Specifically, S4 establishes a task importance ranking mechanism by assigning task priority levels and setting unit reward values, implements two processes of priority assessment and reward setting, and each process has clear numerical standards. Priority level assignment needs to determine the assessment indicators and weights: the assessment indicators include task waiting time, node demand gap size, and user travel impact range, and the weights of the three indicators are all set to 1 / 3 (total weight is 1). Task waiting time is counted from the task generation time, accurate to minutes, and if the waiting time exceeds 2 hours (120 minutes), the indicator score is 1, otherwise the score is 0; the node demand gap size = node net demand of this type of task / node maximum receiving capacity of this type of task (normal demand node shared bicycle maximum receiving capacity 30 vehicles, shared electric vehicle 20 vehicles, multifunctional node shared bicycle 80 vehicles, shared electric vehicle 60 vehicles), and if the gap ratio exceeds 50%, the indicator score is 1, otherwise the score is 0; the user travel impact range is measured by the number of residential areas within 1 km of the node (based on 2000 to 5000 people per residential area) and the number of commercial areas (based on 5000 to 10000 square meters per commercial area), and if it covers more than 5 residential areas or more than 2 commercial areas, the indicator score is 1, otherwise the score is 0. Priority level = sum of three indicator scores x (5 / 3), result is an integer (e.g. if the score sum is 2.4, the priority level is 4), a total of 1 to 5 levels (1 is the lowest, 5 is the highest), where the score sum 0 to 0.6 corresponds to level 1, 0.7 to 1.2 corresponds to level 2, 1.3 to 1.8 corresponds to level 3, 1.9 to 2.4 corresponds to level 4, and 2.5 to 3 corresponds to level 5. The unit reward value is linked to the priority level, and the basic unit reward value is set to 10 yuan (for level 1 task), and the unit reward value increases by 5 yuan for each level of priority, i.e. 15 yuan for level 2 task, 20 yuan for level 3 task, 25 yuan for level 4 task, and 30 yuan for level 5 task. The reward value is calculated accurately to the yuan without decimal places. In the implementation process, the system needs to automatically collect three indicator data and calculate the priority level and reward value every 30 minutes (e.g. 10:00, 10:30, 11:00), and the updated data needs to be synchronized to the dispatch task list to ensure that the task importance ranking matches the operation demand in real time and provides a basis for resource priority allocation.

[0022] S5. Construct a mixed integer programming model, integrate multi-commodity flow, multi-visit vehicle rebalancing, fault maintenance, battery replacement, and priority mechanism into the model, which includes truck capacity constraints, node capacity constraints, and time window constraints; Specifically, S5 integrates the constraints and objectives of multi-task scheduling by constructing a mixed integer programming model, and implements two parts of promotion according to the elements integration and constraint construction, and each part has specific numerical standards. The element integration part converts four types of core elements into model parameters: the multi-commodity flow element distinguishes the transportation parameters of shared bicycles, shared electric vehicles, and spare batteries. The shared bicycle transportation has a unit volume of 0.8 to 1.2 cubic meters per vehicle and a unit weight of 15 to 20 kilograms per vehicle. The shared electric vehicle transportation has a unit volume of 1.5 to 2 cubic meters per vehicle and a unit weight of 30 to 40 kilograms per vehicle. The spare battery transportation has a unit volume of 0.2 to 0.3 cubic meters per group and a unit weight of 15 to 25 kilograms per group. The model needs to allocate transportation space according to different volumes and weights. The vehicle rebalancing element of multiple visits needs to set the maximum number of daily visit nodes for a single truck to be 8 to 12, and the time interval between two consecutive visits to the same node should not be less than 2 hours to avoid repeated transportation of resources. The fault maintenance and battery replacement element includes maintenance time and replacement time in the time constraint. The maintenance time for a faulty shared bicycle is 15 to 40 minutes per vehicle, the maintenance time for a faulty shared electric vehicle is 20 to 50 minutes per vehicle, and the replacement time for a battery is 3 to 8 minutes per vehicle. The model needs to reserve the corresponding time in the node stay time. The priority mechanism element converts the priority level into a weight factor. The weight factors for tasks of levels 1 to 5 are set to 1, 1.5, 2, 2.5, and 3, respectively. The model objective function needs to calculate the task completion benefit according to the weight factor. The constraint condition construction clearly defines the numerical range of three types of core constraints: the truck capacity constraint is set according to the transportation type. The maximum load of a shared bicycle transportation truck is 30 to 80 vehicles per trip, the maximum load of a shared electric vehicle transportation truck is 20 to 50 vehicles per trip, and the maximum load of a spare battery transportation truck is 20 to 40 groups per trip. The total load of the truck should not exceed 10 tons (including the weight of the vehicle). The node capacity constraint is set according to the node type. The maximum number of shared bicycles received at a normal demand node is 10 to 30 per trip, and the maximum number of shared electric vehicles received is 5 to 20 per trip. The maximum number of shared bicycles received at a multifunctional node is 20 to 80, the maximum number of shared electric vehicles received is 15 to 60, and the maximum number of spare batteries received is 10 to 30. The received amount should not exceed the upper limit of the node storage space. The time window constraint needs to set the execution time period for each type of task for each node. The single task execution time window length is 1 to 2 hours (e.g., 14:00-15:00, 15:30-17:30). The task start execution time should not be earlier than the time window start time, and the end execution time should not be later than the time window end time. By clearly defining the above parameters and constraints, the model can accurately map the actual scheduling scenario and provide a reliable mathematical framework for subsequent solution.

[0023] S6. Adopting an adaptive large neighborhood search heuristic algorithm to solve the mixed integer programming model, designing a customized operator to optimize the solution process, and outputting scheduling schemes for different types of tasks. The scheduling scheme includes truck driving path, task execution order, and resource transportation amount.

[0024] Specifically, S6 uses an adaptive large neighborhood search heuristic algorithm to solve the mixed integer programming model, outputs an executable scheduling scheme, and implements the three links of operator design, iterative solving, and scheme output. Each link has specific numerical values and process standards. In the operator design link, to construct a scientific and efficient destruction and repair operator parameter system, three types of differentiated operation modes are designed for the destruction operator: the random path removal operator focuses on adjusting the overall path structure, randomly selects 1 to 3 complete driving paths for each truck to remove, and strictly controls the removal number to be between 10% and 15% of the total path number, to stimulate new combination possibilities by breaking the solidified path allocation; the execution order adjustment operator focuses on optimizing the task arrangement within the path, locally adjusts the task execution order of 2 to 5 nodes each time, with an adjustment scale of 8% to 12% of the total number of nodes, to alleviate the problem of poor task connection within the path; the responsible truck replacement operator focuses on task and capacity re-matching, selects 2 to 4 tasks to replace their responsible trucks each time, with a replacement number accounting for 5% to 10% of the total number of tasks, aiming to balance the load pressure of different trucks. At the same time, based on statistical analysis and effect verification of historical solving data, differentiated application probabilities are set for the three types of operators — 30% for the random path removal operator, 40% for the execution order adjustment operator, and 30% for the responsible truck replacement operator, ensuring the relevance of operator application and achieving dynamic balance between destruction intensity and solution space exploration efficiency through probability allocation, laying a foundation for the effective operation of subsequent repair operators. And the probability is recalibrated every 20 iterations (error ± 5%); the repair operator corresponds to three types, the path re-planning operator plans paths based on node supply-demand matching degree (matching degree ≥ 80%), the re-ordering operator sorts tasks according to task priority level (from high to low), and the re-allocation operator allocates tasks according to truck remaining capacity (remaining capacity ≥ 1.1 times the required capacity of the task), and the verification scheme after repair needs to meet all constraints, with a verification pass rate of ≥ 95%. The iterative solving link needs to set the core parameters: the initial number of iterations is 1000, the iteration step is 1 second, the temperature parameter (controls the probability of accepting worse solutions) is initially 100, and the temperature decreases by 10 every 100 iterations (cooling coefficient 0.9); in each iteration, a new candidate scheme is generated through the destruction operator, then optimized through the repair operator, and the objective function value (operation cost + task revenue, cost calculation includes fuel cost and labor cost, revenue calculation is based on unit reward value × completed task number) of the candidate scheme is calculated. If the candidate scheme value is better than the current optimal scheme (difference ≥ 5%), the optimal scheme is directly updated; if the difference is < 5%, the acceptance probability is calculated according to the current temperature (temperature 100, acceptance probability 20%; temperature 10, acceptance probability 2%), a random value between 0 and 1 is generated, and if the value is less than the acceptance probability, the candidate scheme is accepted.The scheme output link clearly outputs the numerical accuracy of the content: the truck driving path needs to mark the distance of each section of the road (accurate to 0.1 kilometers) and the estimated driving time (accurate to 0.1 minutes), the task execution order needs to mark the estimated start and end time of each task (accurate to minutes), and the resource transportation quantity needs to mark the loading and unloading quantity of each type of resource (accurate to vehicle or group). Through the standardized algorithm implementation process, it is ensured that the output scheme has optimality and executability.

[0025] Preferably, in the construction of the mixed integer programming model in S5, a task priority weight coefficient is introduced to calculate the objective function, and the expression of the objective function is: , is the priority weight coefficient of task r, and increases with the increase of the priority level, is a 0-1 variable, represents that truck k passes through arc transport task , then the opposite, is the reward coefficient of task r, is the completion identifier of task r, represents that task r is completed, then the opposite; At the same time, the resource balance constraint formula is added to the model: , is the resource quantity of truck k transporting task r from node i to node j, is the initial net demand quantity of node i for task r.

[0026] Specifically, S5 constructs a mixed integer programming model to introduce a task priority weight coefficient to optimize the objective function, and supplement resource balance constraints to improve model accuracy. In the design of the objective function, the task priority weight coefficient is set according to the task urgency, with a value range of 1 to 5, where the coefficient of the highest priority fault maintenance task is set to 5, the coefficient of the lowest priority normal shared bicycle rebalancing task is set to 1, and the coefficients of the remaining tasks are set to 2, 3, 4 in order of urgency; the 0-1 variable is used to identify the truck transportation state, and only 0 or 1 values exist to ensure accurate transportation path records; the reward coefficient value range is 0.5 to 2, which is positively related to the task priority, with a high priority task reward coefficient close to 2 and a low priority task reward coefficient close to 0.5; the completion identifier is also a 0-1 variable, which directly reflects whether the task is executed. In the design of the resource balance constraint, the resource transportation amount is determined according to the truck capacity and node demand, and the single transportation amount does not exceed the maximum loading capacity of the truck corresponding to the task, such as the shared bicycle transportation truck single transportation amount does not exceed 80 vehicles; the initial net demand is the node task net demand value calculated in step S3, and the positive number represents demand and the negative number represents supply. In implementation, first, the net demand data of step S3 and the priority data of step S4 are called through the system, and the objective function is calculated to calculate the total operating income and cost, and then the resource balance constraint is verified to verify whether the transportation amount meets the node supply and demand, ensuring that the model output result not only meets the priority orientation, but also realizes the balance of the total amount of resources, avoiding resource surplus or shortage.

[0027] Preferably, the iterative update process of the adaptive large neighborhood search heuristic algorithm in S6 adopts the following formula: wherein, is the current solution of the tth iteration, is the solution of the tth iteration, is the new solution generated by neighborhood search, is the objective function, is the temperature parameter of the tth iteration, is the is a random number in the interval [0, 1]; the selection probability formula of the customized operator in the algorithm is: , is the selection probability of operator o, is the initial weight of operator o, is the historical performance coefficient of operator o, is the operator set.

[0028] ​Specifically, the adaptive large neighborhood search heuristic algorithm of S6 is optimized, and an iterative updating process and operator selection probability formula are designed to improve the solving efficiency and optimality of the algorithm. In the iterative updating, the state of the solution only exists in the current solution and the new solution, the new solution is generated by neighborhood search, the objective function is used to evaluate the pros and cons of the solution, and the smaller the numerical value is, the better the scheme is; the initial value of the temperature parameter is set to 100, and it is reduced by 10 every 100 iterations, and the cooling coefficient is fixed at 0.9, which ensures that the algorithm has a wide exploration range in the early stage and stable convergence in the later stage; the random number is generated in the interval of 0 to 1, and whether to accept the poor solution is determined by comparing it with the probability value, so as to avoid the algorithm falling into local optimum. When calculating the operator selection probability, the initial weight is set according to the historical performance of the operator, the initial weight of the path re-planning operator is set to 0.4, the adjustment execution order operator is set to 0.3, and the replacement responsible truck operator is set to 0.3; the historical performance coefficient is calculated according to the number of times that the operator generates the optimal solution in the past 50 iterations, the more the number of times that the optimal solution is generated, the closer the coefficient is to 1, and vice versa, the closer the coefficient is to 0.1; the operator set includes three kinds of customized operators, covering the adjustment directions of path, order and responsibility allocation. When implemented, the algorithm first selects the operator according to the initial weight to generate a new solution, compares the target function value with the current solution after calculation, judges whether to update the solution according to the temperature parameter and the random number, and recalculates the operator selection probability every 20 iterations to ensure that the algorithm always selects the operator with better performance, improving the solving efficiency and scheme quality.

[0029] Preferably, the dynamic updating of the node net demand in S3 adopts the formula: , is the net demand of node i at time t for task r, is the net demand of node i at time for task r, is the resource amount transported by truck k from node i to node j for task r at time t, is the resource loss coefficient of node i to node j for task r at time t, ; at the same time, the node type judgment threshold formula is: , is the type judgment threshold of node i for task r, is the Kronecker function, when when .

[0030] Specifically, the S3 node net demand calculation, design dynamic update formula and node type judgment threshold formula, realize the real-time accurate judgment of node state. In the dynamic update of net demand, the time interval is set to 1 hour, which is consistent with the statistical time period of step S3, to ensure data continuity; the resource loss coefficient value range is 0.95 to 1, which is adjusted according to the transportation distance. The coefficient of short-distance transportation (distance less than 5 kilometers) is set to 0.99, and the coefficient of long-distance transportation (distance greater than 20 kilometers) is set to 0.95. The intermediate value is used for medium-distance transportation, which is used to correct the resource loss in the transportation process, such as the reduction of available quantity caused by slight damage of the vehicle; the net demand value is rounded up to the nearest integer, which ensures the accuracy of node demand calculation. When calculating the node type judgment threshold, the number of influencing factors is determined according to the node attributes, usually including task type, node capacity, geographical location, and the weight is set to 1 / 3; Kronecker delta is a binary function named after German mathematician Leopold Kronecker, usually with two integer variables. Its definition is: when the two independent variables are equal, the function value is 1; when the two independent variables are not equal, the function value is 0. It has been widely used in many fields such as mathematics, physics and engineering. For example, in the node type judgment threshold formula in this paper, it is used to select the net demand item related to task r. Only when the task type is consistent, take 1, otherwise take 0, to ensure that the threshold calculation is only for the current task type. When implemented, the system automatically collects transportation data and resource loss data every hour, calculates the latest net demand by substituting the dynamic update formula, and judges whether the node type changes by combining the threshold formula, such as the original demand node changing to a supply node due to resource replenishment, to ensure that the scheduling scheme can adapt to the changes of node state in real time.

[0031] Preferably, the calculation of the S4 task priority level adopts the formula: , is the weight of the mth influencing factor, and is the normalized value of task r under the mth influencing factor, is the total number of influencing factors; the determination formula of unit reward value is: , is the basic reward value, is the reward coefficient, .

[0032] Specifically, the task priority of S4 is set with unit reward value, the priority level calculation and unit reward value determination formula are designed, and the scientificity of priority allocation and the rationality of reward mechanism are improved. In the priority level calculation, the influencing factors usually include task waiting time, node demand gap, and influence range, and the weights are all set to 1 / 3 to ensure the balanced influence of each factor on the priority; the standardized value is calculated according to the ratio of the actual value of the factor to the maximum value, for example, if the maximum value of the waiting time is set to 4 hours and a task waits for 2 hours, the standardized value is 0.5, and the value range is 0 to 1; the total number of influencing factors is fixed to 3 to avoid priority fluctuation caused by the change of the number of factors. In the unit reward value setting, the basic reward value is set to 10 yuan, which is the reward standard of the lowest priority task; the reward coefficient value range is 0.01 to 0.1, which is adjusted according to the operation cost, and the coefficient is set to 0.01 when the cost is high and to 0.1 when the cost is low to ensure that the total reward is controlled within a reasonable range. In implementation, the actual data of the three influencing factors is collected through the node intelligent terminal, the standardized value is calculated, then multiplied by the weight to obtain the priority score, the priority level (1 to 5 levels) is determined according to the score, and finally the specific reward amount is calculated by substituting the unit reward value formula, and is updated to the dispatching task list at the same time to provide clear reward guidance for resource priority allocation.

[0033] Preferably, the time window constraint of the model in S5 adopts the formula: , is the earliest start time of node i performing task r, is the latest end time of node i performing task r, is the time of node i processing task , is a 0-1 variable, indicates that node i processes task before performing task r; and the truck capacity constraint formula is: , is the load coefficient of truck k on arc , , is the capacity of truck k.

[0034] Specifically, the S5 model constraint condition, the design time window constraint and the truck capacity constraint formula ensure that the model meets the actual operation scene time and capacity limit. In the time window constraint, the earliest start time and the latest end time are set according to the node operation period, such as the earliest start time of the residential area node task is set to 7:00, and the latest end time is set to 22:00, and the business district node is set to 8:00 to 23:00; the processing time is determined according to the task type, so the processing time of a single vehicle for a shared bicycle fault is 15 to 40 minutes, the processing time of a shared electric vehicle fault is 20 to 50 minutes, and the battery replacement time is 3 to 8 minutes; the 0-1 variable is used to identify the task processing order, and only two states of 0 or 1 exist, to ensure the accuracy of the order record. In the truck capacity constraint, the load coefficient is in the range of 0.8 to 1, which is adjusted according to the truck load rate, and the coefficient is set to 0.8 when the load rate is lower than 60%, and the coefficient is set to 1 when the load rate is higher than 80%, to avoid truck overload or empty load; the truck capacity is the maximum loading capacity of each type of truck set in step S2, such as the shared electric vehicle transport truck capacity is set to 20 to 50 vehicles. When implemented, first, the time window parameters are set according to the node type and task type, and then the maximum transportation capacity of each time is determined by combining the truck capacity and the load coefficient. The two constraints are substituted into the mixed integer programming model, and the model is generated by the scheduling scheme in step S5, to ensure that the scheduling scheme does not exceed the node operation period in time and does not exceed the truck carrying capacity in capacity, to improve the executability of the scheme.

[0035] Preferably, the S3 includes the following steps: S31. Collecting historical data of different types of tasks of each node in a preset time period, the historical data including the number of task generation, the number of task completion and the number of resource remaining, screening the collected historical data, and eliminating abnormal value data, the abnormal value data being determined by a standard of deviating from three times the standard deviation of the average value; S32. Calculating the resource inflow and the resource outflow of each node for different types of tasks according to the screened historical data, the resource inflow being the total amount of resources related to different types of tasks received by the node, and the resource outflow being the total amount of resources related to different types of tasks output by the node; S33. Subtracting the resource inflow from the calculated resource outflow to obtain the initial net demand of each node for different types of tasks, and recording the numerical value of the initial net demand and the corresponding task type and node identification; S34. Comparing the initial net demand of each node for different types of tasks with the preset node resource threshold, and preliminarily judging the possible type of the node according to the comparison result, to provide the node attribute basis for the scheduling scheme.

[0036] Specifically, S3 improves the node net demand calculation and node type judgment process through four sub-steps S31-S34 to ensure data accuracy and subsequent scheduling adaptability. When S31 is implemented, the preset time period is fixed at 1 hour, and the collected historical data includes the number of task generation (recorded every 5 minutes, error ≤ 1 vehicle), the number of task completion (recorded in real time, uploaded to the system immediately after completion) and the number of resource remaining (inspected every 10 minutes, the inspection range includes available vehicles and standby batteries); the abnormal value screening adopts the three standard deviation standard, first calculates the average value and standard deviation of the historical data, and marks and removes the data deviating from the average value by more than 3 times the standard deviation, such as the average value of shared bicycles generated in a node within 1 hour is 20 vehicles, and the standard deviation is 5 vehicles, then the data exceeding 35 vehicles or less than 5 vehicles needs to be removed. In S32, the resource inflow quantity is the total number of various resources unloaded by the transport vehicle (double checked according to the vehicle scheduling record and the node receiving record), and the resource outflow quantity is the total number of resources consumed by the user, the vehicle away and the maintenance (calculated comprehensively through the user order, scheduling instruction and maintenance record), both of which are accurate to 1 vehicle (or 1 group). In S33, when calculating the initial net demand, the resource outflow quantity is directly subtracted from the resource inflow quantity, the result is kept as an integer, and the corresponding task type (such as normal shared bicycle rebalancing) and node identifier (such as node 1 in area A) are recorded simultaneously, and the recording frequency is consistent with the preset time period. In S34, the node resource threshold is set according to the node type, the shared bicycle threshold of the supply node is 50 vehicles, and the shared electric vehicle is 30 vehicles, the normal demand node shared bicycle threshold is 10 vehicles, and the shared electric vehicle is 5 vehicles, the initial net demand is compared with the threshold, and the node type is preliminarily judged, which provides clear node attribute basis for subsequent scheduling scheme development.

[0037] Preferably, S4 includes the following sub-steps: S41. Analyze the factors affecting the urgency of shared travel system tasks, including task waiting time, node demand gap size, and task influence range on user travel, set evaluation grades for each influencing factor, and each evaluation grade corresponds to specific quantitative standards; S42. Organize professional personnel to score the performance of different types of tasks under each influencing factor according to the set quantitative standards, the scoring results adopt the percentage system, take the average of the scores of different professional personnel, and obtain the final scores of different types of tasks under different influencing factors; S43. According to the importance of different influencing factors, give corresponding weight values, the weight values are determined by the analytic hierarchy process, multiply the final scores of different types of tasks under different influencing factors by the corresponding weight values, and sum them up to obtain the priority scores of different types of tasks; S44. According to the distribution range of the priority scores, divide different priority levels, map the priority scores to the priority levels, complete the allocation of the priority levels of different types of tasks, and determine the corresponding unit reward value according to the priority level.

[0038] Specifically, S4 improves the objectivity of priority determination and the rationality of reward mechanism through the four links of S41-S44 in the task priority allocation and unit reward value setting process. In S41, the influencing factors of combing are fixed as three categories: task waiting time, node demand gap size, and task influence range on user travel. The task waiting time evaluation level is divided into 5 grades, each grade corresponds to a 2-hour interval (such as 0-2 hours, 2-4 hours, etc.), the node demand gap size is divided into 5 grades according to the proportion of the maximum capacity of the node (such as 0-20%, 20%-40%, etc.), and the influence range is divided into 5 grades according to the number of residential and commercial areas covered (such as 1-2, 3-4, etc.), each grade is set with clear quantitative standards. In S42, 3-5 professionals with more than 5 years of experience in shared travel operation are organized to score, the score is in percentage (accurate to 1 point), the arithmetic mean (rounded to 1 decimal place) of the scores of different personnel for the same task is taken as the final score of the task under the corresponding influencing factor. In S43, the weights of each influencing factor are determined by the analytic hierarchy process, first construct the judgment matrix (value according to the 1-9 scale method), then calculate the weight vector and perform consistency check (consistency ratio CR≤0.1), finally determine the weight of three types of factors is 1 / 3; the final score is multiplied by the corresponding weight and summed to get the priority score of each type of task (accurate to 0.1 points). In S44, the priority score distribution range is divided into 1-5 levels, 1 level corresponds to 10-30 points, 2 level 30-50 points, 3 level 50-70 points, 4 level 70-90 points, 5 level 90-100 points, the completion score is mapped to the priority level; the unit reward value is set according to the priority level, 1 level 10 yuan, 2 level 15 yuan, 3 level 20 yuan, 4 level 25 yuan, 5 level 30 yuan, a fixed priority and reward value mapping table is established to ensure that the reward value can be quickly matched in subsequent scheduling.

[0039] Preferably, the S5 comprises the following steps: S51. explicitly defining the decision variables of the mixed integer programming model, the decision variables including truck travel path variables, task execution order variables and resource transportation amount variables, defining the value range and physical meaning of each decision variable, and ensuring that the decision variables can accurately describe the calibration elements in the scheduling process; S52. analyzing the constraint conditions in the operation process of the shared travel system, the constraint conditions including truck travel time constraints, node processing capacity constraints and resource supply and demand balance constraints, and converting each constraint condition into a mathematical expression, the mathematical expression including corresponding variables and parameters; S53. constructing the objective function of the model, the objective function being oriented to minimize the system operation cost and maximize the task priority benefit, integrating the cost items and benefit items into a unified mathematical formula, the cost items including transportation costs and loading and unloading costs, and the benefit items including task completion rewards; S54. testing the constructed mathematical model, the testing contents including the integrity of variable definition, the rationality of constraint conditions and the logic of objective function, correcting the problems found in the testing, and ensuring that the model can accurately reflect the actual scheduling demand.

[0040] Specifically, S5 improves the mixed integer programming model construction process through four steps S51-S54 to ensure that the model can accurately reflect the actual scheduling requirements. In S51, the decision variables are clearly divided into three categories: truck driving path variables (indicating whether a truck passes through a certain path), task execution sequence variables (indicating the execution order of node tasks), and resource transportation volume variables (indicating the amount of transportation resources). The value range of each variable is set according to the actual scene, with path variables and sequence variables being 0-1 variables, and transportation volume variables being non-negative integers (minimum unit 1 vehicle or 1 group). At the same time, the physical meaning of each variable is clearly defined, such as path variable taking 1 indicating that the truck passes through the path, and taking 0 indicating that it does not pass through. In S52, the constraint conditions include truck driving time constraints (total driving time for single scheduling does not exceed 8 hours), node processing capacity constraints (single node processes no more than 20 vehicles per hour), and resource supply and demand balance constraints (system total supply and demand difference ≤5%). When converting each constraint condition into a mathematical expression, specific parameters need to be substituted, such as driving time constraints combined with node-to-node driving time calculated in step S2 to ensure that the constraints conform to the actual operation capacity. In S53, when constructing the objective function, the cost items include transportation costs (truck fuel costs calculated at 2 yuan per kilometer), loading and unloading costs (1 yuan for each shared bicycle and 2 yuan for each shared electric vehicle), and the revenue item is the task completion reward (calculated according to the unit reward value set in step S4). The cost items and revenue items are integrated into a mathematical formula of "total revenue - total cost" with the goal of maximizing this value. In S54, when testing the model, the integrity test confirms that all decision variables have been defined, the reasonableness test verifies the matching degree of the constraint conditions with the actual scene, and the logical test ensures that the objective function and the constraint conditions have no contradictions. Problems found in the test (such as missing variables and overly strict constraints) are corrected in a timely manner, and finally a complete model that can be used for solving is formed.

[0041] As Figure 2As shown, a priority-based multi-task scheduling system for shared travel, which is applied to a priority-based multi-task scheduling method for shared travel, comprises: a task and node information collection and processing unit, configured to collect task information and node information in a shared travel system, the task information including task type and task generation time, the node information including node location and node resource capacity, and to classify, arrange and format the collected information to provide standardized data for subsequent steps; a truck parameter configuration and path calculation unit, configured to input basic parameters of trucks, including capacity and driving speed, to calculate the driving distance of each truck between any two nodes according to the node location information, to derive the driving time in combination with the driving speed, and to generate a truck driving parameter table; a net demand calculation and node type judgment unit, configured to receive the standardized data output by the task and node information collection and processing unit, to calculate the net demand of each node for different types of tasks, to determine whether the node is a demand node or a supply node according to the net demand value, and to output the node type judgment result; a task priority allocation and reward setting unit, configured to determine the factors and weights affecting the priority of the task, to calculate the priority level of different types of tasks, to set the corresponding unit reward value according to the priority level, and to establish a mapping relationship table of priority and reward value; a mixed integer programming model construction and solving unit, configured to receive the output data, to construct a mixed integer programming model including multiple constraint conditions and an objective function, to call an adaptive large neighborhood search heuristic algorithm to solve the model, and to generate a preliminary scheduling scheme; and a scheduling scheme output and dynamic adjustment unit, configured to receive the preliminary scheduling scheme generated by the mixed integer programming model construction and solving unit, to verify the feasibility of the scheme, to output the verified scheme to the operation terminal of the shared travel system, to simultaneously receive real-time system operation feedback data, and to dynamically adjust the scheduling scheme according to the feedback data.

[0042] A priority-based multi-task scheduling method and system for shared travel, by constructing a multi-task set including normal vehicle rebalancing, fault maintenance and battery replacement, combining node net demand calculation to accurately determine the supply and demand properties of each node, matching the priority allocation mechanism to clearly determine the task urgency, and associating the priority with the unit reward value, the deep integration of multi-type tasks and priority is realized. This design directly overcomes the shortcomings of the prior art that only designs strategies for single tasks and cannot allocate resources according to the degree of urgency, so that critical tasks such as urgent fault maintenance and high-demand node resource replenishment can obtain resource support in priority, greatly improving the response speed of the system to high-urgency tasks, improving the accuracy of task processing, and avoiding the loss of operating efficiency caused by resource mismatch.

[0043] The scheduling method and system are also excellent in model construction and dynamic adaptation. A mixed integer programming model is constructed by integrating actual operation constraints such as multi-commodity flow, vehicle capacity and time window, to ensure that the model can fully reflect the actual operation demand, and an adaptive large neighborhood search algorithm with customized operators is used for solving, to further improve the optimality of the scheduling scheme. Meanwhile, relying on the dynamic judgment of node types, real-time update of resource net demand and dynamic adjustment mechanism of scheduling scheme, the system can adapt to the changes of node supply and demand and task demand in real time. This series of designs overcome the shortcomings of the existing models, such as incomplete consideration of constraints, lack of optimization of solving algorithm and difficulty in adapting to dynamic changes of the system, ensure that the scheduling scheme has practical feasibility, and can be adjusted in time according to the changes of the system, effectively improve the rationality and flexibility of resource allocation, ultimately improve the overall task execution efficiency, and balance the operation cost control and user travel experience improvement.

[0044] In the description of the present application, it should be noted that unless otherwise explicitly specified and limited, the terms "arrangement", "installation", "connection", "link", "fixing" should be understood broadly, for example, can be fixed connection, can also be detachable connection, or integrally connected; can be mechanical connection, can also be electrical connection; can be directly connected, can also be indirectly connected through an intermediate medium, can be the communication inside two elements. For those skilled in the art, the specific meaning of the above terms in the present application can be understood according to the specific circumstances.

[0045] Although the embodiments of the present application have been shown and described, it can be understood by those skilled in the art that various equivalent changes, modifications, replacements and variations of the embodiments can be made without departing from the principles and spirits of the present application, and the scope of the present application is defined by the appended claims and their equivalent scope.

Claims

1. A priority-based multi-task scheduling method for shared mobility, characterized in that, Includes the following steps: S1. Construct the node set and task set of the shared mobility system. The node set includes supply nodes, normal demand nodes for vehicle rebalancing operations, and nodes capable of performing multi-functional tasks. The task set includes normal shared bicycle rebalancing, normal shared electric bicycle rebalancing, faulty shared bicycle repair, faulty shared electric bicycle repair, and low-battery shared electric bicycle battery replacement tasks. S2. Configure the parameters of the truck set, including the capacity and speed of each truck, determine the travel distance of each truck between nodes, and calculate the travel time of each truck between any two nodes. S3. Define the net demand for each node for different types of tasks, and determine whether a node is a demand node or a supply node based on the positive or negative value of the net demand. Record the initial net demand for different types of tasks. S4. Assign priority levels to different types of tasks. The priority level corresponds to the urgency of the task. At the same time, set the unit reward value for each type of task. The unit reward value is positively correlated with the priority level. S5. Construct a mixed-integer programming model that integrates multi-commodity flow, multiple-visit vehicle rebalancing, fault repair, battery replacement, and priority mechanisms into the model. The model includes truck capacity constraints, node capacity constraints, and time window constraints. S6. Use an adaptive large neighborhood search heuristic algorithm to solve the mixed-integer programming model, design customized operators to optimize the solution process, and output scheduling schemes for different types of tasks. The scheduling schemes include truck travel paths, task execution order, and resource transportation volume.

2. The priority-based multi-task scheduling method for shared mobility according to claim 1, characterized in that, When constructing the mixed integer programming model in S5, task priority weight coefficients are introduced to calculate the objective function, and the expression of the objective function is: , The priority weight coefficient for task r is... And it increases with increasing priority level. For 0-1 variables, This indicates that truck k passes through arc Transportation mission Conversely, Let be the reward coefficient for task r. This serves as a completion marker for task r. This indicates that task r is complete. Conversely, a resource balance constraint formula is added to the model: , Let r be the amount of resources that truck k uses to transport task r from node i to node j. Let i be the initial net requirement for task r.

3. The priority-based multi-task scheduling method for shared mobility according to claim 1, characterized in that, The iterative update process of the adaptive large neighborhood search heuristic algorithm in S6 adopts the following formula: ,in, for The current solution in the next iteration For the first The solution of the next iteration A new solution generated for neighborhood search. Let be the objective function. Let be the temperature parameter for the t-th iteration. for Random numbers within the interval; the selection probability formula for the customized operator in the algorithm is: , Let o be the choice probability of operator o. Let o be the initial weight of operator o. The historical performance coefficients of operator o, Let be the set of operators.

4. The priority-based multi-task scheduling method for shared mobility according to claim 1, characterized in that, The dynamic update of the net demand of nodes in S3 is achieved using the following formula: , Let be the net demand of node i for task r at time t. Let node i be at time i For the net requirement of task r, Let t be the amount of resources that truck k uses to transport task r from node i to node j at time t. Let be the resource loss coefficient for the transportation task r from node i to node j at time t. Meanwhile, the threshold formula for determining node type is: , For node i, determine the threshold based on the type of task r. For the Kronecker function, when when .

5. A priority-based multi-task scheduling method for shared mobility according to claim 1, characterized in that, The calculation of task priority level in S4 uses the following formula: , Let m be the weight of the m-th influencing factor. and Let r be the standardized value of task r under the m-th influencing factor. The total number of influencing factors; the formula for determining the unit reward value is: , Based on the base reward value, As the reward coefficient, .

6. A priority-based multi-task scheduling method for shared mobility according to claim 1, characterized in that, The time window constraint for the model in S5 is defined by the following formula: , The earliest start time for executing task r for node i. The latest end time for task r to be executed on node i. Process tasks for node i Time, For 0-1 variables, This indicates that node i processes the task before executing task r. Conversely, the truck capacity constraint formula is: , For truck k in arc The load factor on, , Let k be the capacity of the truck.

7. A priority-based multi-task scheduling method for shared mobility according to claim 1, characterized in that, S3 includes the following steps: S31. Collect historical data of different types of tasks for each node within a preset time period. The historical data includes the number of tasks generated, the number of tasks completed, and the remaining resources. Filter the collected historical data and remove outlier data. Outlier data is determined by deviating from the average value by three times the standard deviation. S32. Based on the filtered historical data, calculate the resource inflow and resource outflow for each node for different types of tasks. The resource inflow is the total amount of resources related to different types of tasks received by the node, and the resource outflow is the total amount of resources related to different types of tasks output by the node. S33. Subtract the resource inflow from the calculated resource outflow to obtain the initial net demand for each node for different types of tasks. Record the value of the initial net demand, the corresponding task type, and the node identifier. S34. Compare the initial net demand for different types of tasks for each node with the preset node resource threshold. Based on the comparison results, preliminarily determine the possible type of the node, providing a basis for node attribute formulation for scheduling schemes.

8. A priority-based multi-task scheduling method for shared mobility according to claim 1, characterized in that, S4 includes the following steps: S41. Identify the factors affecting the urgency of tasks in the shared mobility system, including task waiting time, the size of the demand gap at each node, and the scope of the task's impact on user travel. Assign an evaluation level to each factor, with each evaluation level corresponding to a specific quantitative standard. S42. Organize professionals to score the performance of different types of tasks under each influencing factor according to the established quantitative standards. The scoring results are on a percentage basis. The average of the scores from different professionals is used to obtain the final score for different types of tasks under different influencing factors. S43. Assign corresponding weight values ​​based on the importance of different influencing factors. The weight values ​​are determined using the analytic hierarchy process (AHP). Multiply the final scores of different types of tasks under different influencing factors by their corresponding weight values ​​and sum them to obtain the priority scores for different types of tasks. S44. Based on the distribution range of priority scores, divide tasks into different priority levels. Map the priority scores to the priority levels to complete the allocation of priority levels for different types of tasks. Simultaneously, determine the corresponding unit reward value based on the priority level.

9. A priority-based multi-task scheduling method for shared mobility according to claim 1, characterized in that, S5 includes the following steps: S51. Define the decision variables of the mixed-integer programming model, including truck travel path variables, task execution order variables, and resource transportation volume variables. Define the value range and physical meaning of each decision variable to ensure that the decision variables can accurately describe the calibration elements in the scheduling process; S52. Analyze the constraints in the operation of the shared mobility system, including truck travel time constraints, node processing capacity constraints, and resource supply and demand balance constraints. Transform each constraint into a mathematical expression, which includes the corresponding variables and parameters; S53. Construct the objective function of the model. The objective function is guided by minimizing system operating costs and maximizing task priority benefits. Integrate cost items and benefit items into a unified mathematical formula. Cost items include transportation costs and loading and unloading costs, and benefit items include task completion rewards; S54. Test the constructed mathematical model. The test includes the completeness of variable definitions, the rationality of constraints, and the logic of the objective function. Correct any problems found during the test to ensure that the model can accurately reflect the actual scheduling needs.

10. A priority-based multi-task scheduling system for shared mobility, characterized in that, This system is applied to a priority-based multi-task scheduling method for shared mobility as described in claim 1, comprising: a task and node information acquisition and processing unit, used to acquire task information and node information in the shared mobility system, the task information including task type and task generation time, and the node information including node location and node resource capacity, classifying, organizing, and formatting the acquired information to provide standardized data; a truck parameter configuration and route calculation unit, used to input the basic parameters of the trucks, including capacity and driving speed, calculate the driving distance of each truck between any two nodes based on the node location information, deduce the driving time based on the driving speed, and generate a truck driving parameter table; and a net demand calculation and node type judgment unit, used to receive the standardized data output by the task and node information acquisition and processing unit, calculate the net demand of each node for different types of tasks, and determine the net demand based on the net demand value. The system comprises the following components: a task priority allocation and reward setting unit, which determines the factors and weights influencing task priority, calculates the priority levels for different types of tasks, sets corresponding unit reward values ​​based on priority levels, and establishes a mapping table between priority and reward values; a mixed-integer programming model construction and solution unit, which receives the output data, constructs a mixed-integer programming model including multiple constraints and objective functions, calls an adaptive large neighborhood search heuristic algorithm to solve the model, and generates a preliminary scheduling scheme; and a scheduling scheme output and dynamic adjustment unit, which receives the preliminary scheduling scheme generated by the mixed-integer programming model construction and solution unit, verifies the feasibility of the scheme, outputs the verified scheme to the operation terminal of the shared mobility system, and simultaneously receives real-time system operation feedback data to dynamically adjust the scheduling scheme based on the feedback data.

Images