Joint scheduling method and system for road network repair and material distribution

Abstract

The application provides a joint scheduling method and system for road network repair and material distribution, and relates to the field of joint scheduling. Compared with the prior art, the technical scheme provided by the application is improved by using the Q learning algorithm in reinforcement learning, and a joint scheduling method for road network repair and material distribution based on double-layer interactive Q learning is provided. The joint scheduling decision under the objective function can be fed back to the repair team and the transport team, so that the two teams can make decisions that are beneficial to emergency response at the same time, thereby achieving the purpose of joint scheduling and effectively reducing the time cost of solving.

Description

Technical Field

[0001] This invention relates to the field of joint dispatching technology, specifically to a joint dispatching method and system for road network emergency repair and material distribution. Background Technology

[0002] Following natural disasters such as floods and earthquakes, disaster areas require large quantities of various emergency supplies to sustain rescue operations and support the daily lives of affected residents. Therefore, rapid and effective post-disaster emergency supplies delivery is crucial for saving lives and minimizing economic losses in disaster-stricken areas.

[0003] However, natural disasters such as floods and earthquakes can damage roads leading to disaster areas, hindering vehicle traffic and affecting the supply of materials to disaster areas. Therefore, road network repair work after a disaster is also an important part of emergency rescue activities.

[0004] Current research on the joint scheduling of emergency material distribution and emergency road repair faces significant challenges due to the complexity of the joint scheduling model, resulting in high difficulty in algorithm calculation and solution and a long solution time.

[0005] Therefore, there is an urgent need for a joint scheduling method for road network emergency repair and material distribution to shorten the solution time. Summary of the Invention

[0006] (a) Technical problems to be solved

[0007] To address the shortcomings of existing technologies, this invention provides a joint scheduling method and system for road network emergency repair and material distribution, solving the problem of long solution time for joint scheduling of emergency material distribution and emergency road repair in existing technologies.

[0008] (II) Technical Solution

[0009] To achieve the above objectives, the present invention provides the following technical solution:

[0010] In a first aspect of the invention, a joint scheduling method for road network emergency repair and material distribution is provided, the method comprising:

[0011] S1. Obtain the road network parameters of the road network structure and set t = 0;

[0012] S2. Initialize the state space and action space in the emergency repair decision model and the transportation decision model; wherein, the emergency repair decision model and the transportation decision model are both triplets composed of a state space, an action space and a reward function;

[0013] S3. Solve the emergency repair decision-making model based on the Q-learning algorithm to obtain the emergency repair strategy H of the emergency repair team. After all demand points in the road network structure can be reached by the transport team, send the updated road network information to the transport decision-making model based on the buffer table mechanism, and solve the transport decision-making model based on the Q-learning algorithm to obtain the transport strategy P of the transport team;

[0014] S4. Calculate the objective function value f(P) of the objective function based on the emergency repair strategy H and the transport strategy P; where, the objective function is i represents the nodes in the road network structure, C i ′0 represents the cumulative time cost from the reserve point 0 to the demand point i under the rescue strategy P, I i represents the degree of disaster or the importance degree of node i, V d represents the set of demand points;

[0015] S5. Judge whether the function value f(P) < F holds. If so, let F = f(P), and update the emergency repair strategy H to the action space in the emergency repair decision-making model and update the transport strategy P to the action space in the transport decision-making model based on the buffer table of the lock mechanism; if not, directly go to S6; where, F represents the historical optimal objective function value;

[0016] S6. Let t = t + 1, and judge whether t > T max holds. If not, go to S3; if so, output the objective function value f(P), and use the emergency repair strategy H and the transport strategy P as the final joint scheduling plan.

[0017] Optionally, the emergency repair decision-making model includes a state space S r , an action space A r and a reward function R r ;

[0018] Among them, the state space s r is composed of a triple ; is the list of demand points that the transport team can already reach; is the list of the shortest path lengths from each reachable demand point i to the reserve point 0; ε is the list of repaired road sections;

[0019] The action space A r represents the set of damaged road sections that the emergency repair team can reach at the current location;

[0020] The reward function r r is related to the connectivity state of the demand points. When the emergency repair team is in the first emergency repair state from the action set a rAfter selecting and repairing the damaged section, the next repair phase will begin. hour;

[0021] If no new demand is met, then the reward function R r The calculation formula is: l ij Represents the distance from node i to node j, t ij This refers to road sections that are impassable. ij The repair time cost of the repair team, and v represents the average speed of the repair team on the passable road section;

[0022] If a new demand point is met, the reward function R... r The calculation formula is: Indicates the second emergency repair status. The corresponding set of nodes Indicates the first emergency repair status. The corresponding set of nodes, λ represents the preset weighting weight, and D i0 This represents the shortest distance from demand node i to reserve point 0.

[0023] Optionally, the emergency repair decision model includes a state space S. d Action Space A d and the reward function R d ;

[0024] Wherein, the state space S d By binary constitute; This represents a list of reachable demand points. A list of shortest path lengths from rescued demand point i to reserve point 0;

[0025] Action Space A d This represents the set of nodes that the transport team can reach from its current location;

[0026] When the transport team is in the first transport state From action space A d Select the demand node i for rescue, and reach the second transportation state. At that time, the reward function R d The calculation formula is: Where η and β are both preset control variables, and D mi This represents the shortest distance from the current location of the transport team to the demand node i; r i This represents the overall reward after rescuing node i, and the calculation formula is: r i′ represents the overall reward in the previous period, L represents the preset parameters, and f(P) represents the objective function value.

[0027] Optionally, step S3, solving the emergency repair decision model based on the Q-learning algorithm to obtain the emergency repair strategy H of the emergency repair team, after all demand points in the road network structure can be reached by the transportation team, sending the updated road network information to the transportation decision model based on the buffer table mechanism, and solving the transportation decision model based on the Q-learning algorithm to obtain the transportation strategy P of the transportation team, includes:

[0028] S301, The repair team, based on a greedy strategy, starts from action space A r Select and execute action a1, and input the set of damaged road sections that the repair team has repaired into the buffer table. and cumulative time cost C i0 ;

[0029] S302, the emergency repair team changed the emergency repair status from... Transition to Calculate the corresponding Q r Value, and update the repair team's Q r Value table;

[0030] S303. Determine whether the emergency repair status has reached the final state. If yes, obtain the emergency repair strategy H of the emergency repair team. If no, update the action space of the emergency repair decision model and proceed to step S301.

[0031] S304, the transport team retrieves the set of repaired damaged road sections from the buffer table. and cumulative time cost C i0 And update the road network information, and based on a greedy strategy, start from action space A d Select and execute action a2;

[0032] S305, the transport team will change the transport status from Transition to transport status Calculate the corresponding Q d Value, and update the transport team's Q. d Value table;

[0033] S306. Determine whether the transportation state has reached the final state. If yes, obtain the transportation strategy P of the transportation team. If no, update the action space of the transportation decision model and proceed to step S304.

[0034] In a second aspect of the invention, a joint dispatching system for road network emergency repair and material distribution is provided, the system comprising:

[0035] The first acquisition module is used to execute S1, acquire the road network parameters of the road network structure, and set t=0;

[0036] An initialization module for performing S2 to initialize the state space and action space in the emergency repair decision-making model and the transportation decision-making model; wherein, the types of the emergency repair decision-making model and the transportation decision-making model are both triples composed of a state space, an action space, and a reward function.

[0037] A second acquisition module for performing S3 to solve the emergency repair decision-making model based on the Q-learning algorithm to obtain the emergency repair strategy H of the emergency repair team. After all demand points in the road network structure can be reached by the transportation team, the updated road network information is sent to the transportation decision-making model based on the buffer table mechanism, and the transportation decision-making model is solved based on the Q-learning algorithm to obtain the transportation strategy P of the transportation team.

[0038] A first calculation module for performing S4 to calculate the objective function value f(P) of the objective function based on the emergency repair strategy H and the transportation strategy P; wherein, the objective function is i represents a node in the road network structure, C i ′0 represents the cumulative time cost from the reserve point 0 to the demand point i under the rescue strategy P, I i represents the disaster-affected degree or the importance degree of node i, V d represents the set of demand points.

[0039] A first judgment module for performing S5 to judge whether the function value f(P) < F holds. If so, let F = f(P), and update the emergency repair strategy H to the action space in the emergency repair decision-making model and update the transportation strategy P to the action space in the transportation decision-making model based on the buffer table with a lock mechanism; if not, directly go to S6; where F represents the historical optimal objective function value.

[0040] A second judgment module for performing S6 to let t = t + 1 and judge whether t > T max holds. If not, go to S3; if so, output the objective function value f(P), and use the emergency repair strategy H and the transportation strategy P as the final joint scheduling plan.

[0041] Optionally, the emergency repair decision-making model includes a state space S r , an action space A r and a reward function R r ;

[0042] wherein, the state space s r is composed of a triple ; is a list of demand points that the transportation team can already reach; is a list of the shortest path lengths from each reachable demand point i to the reserve point 0; ε is a list of repaired road sections;

[0043] Action Space A r This indicates the location of the damaged road sections that the repair team can reach from their current position.

[0044] Reward function R r Related to the connectivity status of the demand point, when the repair team is in the first repair state From action set a r After selecting and repairing the damaged section, the next repair phase will begin. hour;

[0045] If no new demand is met, then the reward function R r The calculation formula is: l ij Represents the distance from node i to node j, t ij This refers to road sections that are impassable. ij The repair time cost of the repair team, and v represents the average speed of the repair team on the passable road section;

[0046] If a new demand point is met, the reward function r r The calculation formula is: Indicates the second emergency repair status. The corresponding set of nodes Indicates the first emergency repair status. The corresponding set of nodes, λ represents the preset weighting weight, and D i0 This represents the shortest distance from demand node i to reserve point 0.

[0047] Optionally, the emergency repair decision model includes a state space S. d Action Space A d and the reward function R d ;

[0048] Wherein, the state space S d By binary constitute; This represents a list of reachable demand points. A list of shortest path lengths from rescued demand point i to reserve point 0;

[0049] Action Space A d This represents the set of nodes that the transport team can reach from its current location;

[0050] When the transport team is in the first transport state From action space A d Select the demand node i for rescue, and reach the second transportation state. At that time, the reward function R d The calculation formula is: Where η and β are both preset control variables, and D mi This represents the shortest distance from the current location of the transport team to the demand node i; r i This represents the overall reward after rescuing node i, and the calculation formula is: r i ′ represents the overall reward in the previous period, L represents the preset parameters, and f(P) represents the objective function value.

[0051] Optionally, the second acquisition module includes:

[0052] The first input unit is used to execute S301, and the repair team uses a greedy strategy to input data from the action space A. r Select and execute action a1, and input the set of damaged road sections that the repair team has repaired into the buffer table. and cumulative time cost C i0 ;

[0053] The first update unit is used to execute S302, whereby the repair team changes the repair status from... Transition to Calculate the corresponding Q r Value, and update the repair team's Q r Value table;

[0054] The first judgment unit is used to execute S303 and determine whether the emergency repair status has reached the final state. If yes, the emergency repair strategy H of the emergency repair team is obtained; if no, the action space of the emergency repair decision model is updated and the process is transferred to step S301.

[0055] The first acquisition unit is used to execute S304, whereby the transport team retrieves the set of repaired damaged road sections from the buffer table. and cumulative time cost C i0 And update the road network information, and based on a greedy strategy, start from action space A d Select and execute action a2;

[0056] The second update unit is used to execute S305, which changes the transport status of the transport team from... Transition to transport status Calculate the corresponding Q d Value, and update the transport team's Q. d Value table;

[0057] The second judgment unit is used to execute S306 to determine whether the transportation state has reached the final state. If yes, the transportation strategy P of the transportation team is obtained; if no, the action space of the transportation decision model is updated and the process proceeds to step S304.

[0058] In a third aspect of the present invention, there is provided an electronic device, including a processor, a communication interface, a memory, and a communication bus. Among them, the processor, the communication interface, and the memory complete communication with each other through the communication bus;

[0059] The memory is used to store a computer program;

[0060] The processor is configured to implement the steps of any joint scheduling method for road network emergency repair and material distribution when executing the program stored on the memory.

[0061] In a fourth aspect of the present invention, there is provided a computer-readable storage medium, in which a computer program is stored, and when the computer program is executed by a processor, the steps of any joint scheduling method for road network emergency repair and material distribution are implemented.

[0062] (III) Beneficial effects

[0063] The present invention provides a joint scheduling method for road network emergency repair and material distribution. Compared with the prior art, it has the following beneficial effects:

[0064] The method includes: S1. Obtain the road network parameters of the road network structure and set t = 0; S2. Initialize the state space and action space in the emergency repair decision model and the transportation decision model; S3. Solve the emergency repair decision model based on the Q-learning algorithm to obtain the emergency repair strategy H of the emergency repair team. After all demand points in the road network structure can be reached by the transportation team, send the updated road network information to the transportation decision model based on the buffer table mechanism, and solve the transportation decision model based on the Q-learning algorithm to obtain the transportation strategy P of the transportation team; S4. Calculate the objective function value f(P) of the objective function based on the emergency repair strategy H and the transportation strategy P; S5. Judge whether the function value f(P) < F holds. If so, set F = f(P), and update the emergency repair strategy H to the action space in the emergency repair decision model and update the transportation strategy P to the action space in the transportation decision model based on the buffer table with a lock mechanism; if not, directly go to S6; where F represents the historical optimal objective function value; S6. Set t = t + 1, and judge whether t > T max holds. If not, go to S3; if so, output the objective function value f(P), and use the emergency repair strategy H and the transportation strategy P as the final joint scheduling plan.

[0065] Based on the above processing, compared with the existing technology, this invention improves upon the Q-learning algorithm in reinforcement learning and proposes a joint scheduling method for road network emergency repair and material distribution based on two-layer interactive Q-learning. This method can feed back the joint scheduling decision under the objective function to the emergency repair team and the transportation team, enabling both to make decisions that are conducive to emergency response at the same time, thereby achieving the purpose of joint scheduling and effectively reducing the time spent on solving the problem. Attached Figure Description

[0066] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0067] Figure 1 A schematic diagram of a damaged road network provided for an embodiment of the present invention;

[0068] Figure 2 A flowchart illustrating a joint scheduling method for road network emergency repair and material distribution, provided as an embodiment of the present invention;

[0069] Figure 3 This is a schematic diagram of an interaction based on a buffer table, provided as an embodiment of the present invention.

[0070] Figure 4 An experimental result diagram provided for an embodiment of the present invention;

[0071] Figure 5 Another experimental result diagram provided for an embodiment of the present invention;

[0072] Figure 6 Another experimental result diagram provided for an embodiment of the present invention;

[0073] Figure 7 This is a structural diagram of an electronic device provided in an embodiment of the present invention. Detailed Implementation

[0074] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention are described clearly and completely. Obviously, the described embodiments are only some embodiments of the present invention, not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0075] This application provides a joint scheduling method for road network repair and material distribution, solving the problem of long solution times in existing technologies for joint scheduling of emergency material distribution and emergency road repair. It implements a joint scheduling method for road network repair and material distribution based on two-layer interactive Q-learning, which can feed back the joint scheduling decisions under the objective function to the repair team and the transportation team, enabling both to make decisions conducive to emergency response simultaneously, thereby achieving the purpose of joint scheduling and effectively reducing the solution time.

[0076] To better understand the above technical solutions, the following will provide a detailed explanation of the technical solutions in conjunction with the accompanying drawings and specific implementation methods.

[0077] First, basic information about the technical solution provided by this invention will be explained. For example... Figure 1 As shown, Figure 1 This is a schematic diagram of the damaged road network.

[0078] The damaged road network after the disaster can be represented by an undirected graph G = (V, E). Here, V represents the set of nodes in the road network, including the reserve point "0" and the demand point set V. d The set of non-demand points V z E represents the set of road segments in the road network, including the set of passable road segments. z and the set of damaged road sections E d .

[0079] I i This indicates the degree of damage suffered by node i.

[0080] When node i∈V d At that time, I i >0 and I i The larger the value, the more severe the disaster at node i, and the higher the urgency of rescue.

[0081] For demand point i∈V d μ i ∈{0,1}, representing the emergency state of the demand point; if μ i If μ is 0, it means that demand point i does not have an emergency state. i A value of 1 indicates that demand point i is in an emergency state; road segment e ij ∈E, through parameter e ij ∈{0,1} represents road segment e ij The passage status; if e ij If it is 0, it means that road segment e ij It is not yet passable, if e ij If the value is 1, it indicates that the road segment is e. ij Passage is permitted.

[0082] In addition, the repair team was on a road with a length of lij The average speed on a passable road segment is v; for an impassable road segment e ij The repair team's repair time cost is t ij The time cost for the transport team to rescue each demand node i is d. i ,in, This represents the collection of rescue time expenditures; the longer the rescue time, the greater the demand for supplies at the point of need.

[0083] Using parameter C respectively i0 and C′ i0 Let represent the cumulative time cost for the repair team and transport team from the reserve point 0 to node i; set the maximum acceptable distance D from node i to the demand point 0. i After the road network is repaired, the shortest distance D from the demand node i to the reserve point 0 is... i0 It must not exceed the maximum acceptable distance D i This indicates that demand point i is reachable, i.e., D. i0 ≤D i ; through parameter θ i Let ∈{0,1} represent the reachability state of demand point i, i.e., θ i If θ is 0, it means that the demand point i is not yet reachable; i If the value is 1, it means that the demand point i is reachable.

[0084] In the technical solution provided by this invention, the repair team and the transport team depart from the same reserve point, carrying sufficient emergency relief supplies. The travel time of the road segment, the repair time of the damaged road segment, and the demand for supplies from the points of need are all fixed. Specifically, the technical solution provided by this invention only retains reserve nodes in the actual road network, nodes in the disaster-stricken area, and nodes in some potentially affected areas. Therefore, the scale of the road network to be processed is reduced, and the road network information is only partially observable. The repair team and the transport team, departing from the reserve point, can only obtain partial information about the road network environment (i.e., information about damaged road segments). The locational relationships between passable road segments and points of need in the damaged road network need to be gradually obtained through exploration.

[0085] The repair team's restoration plan and the transport team's rescue plan are as follows:

[0086] Each time, the repair team randomly selects one damaged road segment from the set of damaged road segments for repair, until all demand points are met. The damaged road segment that makes all demand points reachable is called a repair scheme, denoted by vector H.

[0087]

[0088] in, arrive This indicates the repair sequence of the emergency repair team. A complete repair plan H must satisfy the reachability of all required points. Figure 1 In the middle, H = <e 01 ,e 15 ,e 54 ,e 45 ,e 58 This is a repair plan for the emergency repair team. Before a complete repair plan is reached, [we will use...] Record the damaged road sections that have been repaired.

[0089] The transport team randomly selects one demand point from the demand point set for each rescue operation, until all demand points have been rescued. The total number of rescued demand points is called a rescue plan, represented by a vector P.

[0090]

[0091] Among them, i1 to i k The rescue order of the transport team is represented by P. A complete rescue plan P must satisfy all the needs to be rescued. The rescue plan of the transport team depends to some extent on the repair strategy of the emergency repair team. The relationship between the two is represented by P = C(H).

[0092] If the route leading to the demand point is segment e ij In a damaged state, i.e., ε ij =0, which makes node j unreachable, i.e., θ j =0, the transport team cannot pass and can only choose other road sections or wait in place for road section e. ij Repaired. When there are no other points of need for rescue, the rescue team's waiting time at node i is ω. i ,exist Figure 1 A rescue plan for the Chinese transport team is P = <5,7,3,9>.

[0093] A complete emergency relief operation is carried out collaboratively by repair teams and transportation teams. Repair teams are responsible for repairing road sections to ensure accessibility to points of need; transportation teams are responsible for delivering relief supplies to these points. For damaged road sections, transportation teams can only pass after the repair teams have completed repairs, so the repair efficiency of the repair teams affects the rescue efficiency of the transportation teams. The purpose of emergency relief operations is to ensure timely rescue to points of need and to reduce casualties and property damage; therefore, the repair teams indirectly determine the efficiency of emergency relief operations, while the transportation teams directly determine their efficiency.

[0094] The joint scheduling of road network repair and material distribution can be described as the following two-level programming problem:

[0095]

[0096] in, I represents the cumulative time cost from reserve point 0 to demand point i under rescue strategy P. i This indicates the importance of node i. Previous studies have typically used the cumulative time cost as the objective function of joint scheduling to formulate different repair and rescue strategies, ignoring the reality that different regions have different degrees of disaster.

[0097] The technical solution provided by this invention takes into account the rescue efficiency of the demand points: for demand points that are severely damaged, their time urgency is high, and they need to be connected to the reserve point 0 in the shortest possible time to open up the lifeline and receive timely rescue.

[0098] Markov decision models are used to describe the process by which an agent makes decisions in an environment and continuously optimizes them iteratively. As the agent performs actions and receives rewards, it continuously adjusts its strategy in order to maximize the total reward (in this paper, minimizing the objective function).

[0099] As can be clearly seen from the road network model, in the repair of damaged road networks and the distribution of supplies, decisions are made sequentially in chronological order, which are all sequential decision-making processes. They also have dynamic characteristics of being partly random and partly controlled by the decision-maker, which is very consistent with the Markov sequential decision-making process. Therefore, the decision-making process of the repair team and the transportation team is described based on the Markov decision model.

[0100] Furthermore, Q-learning, as a reinforcement learning method, finds the optimal policy by iteratively updating the Q-value function on a known model. Q-learning is suitable for solving path planning problems in discrete action and state spaces, maximizing long-term rewards, and in uncertain environments. Initially, the agent can only obtain partial information about the current road network, and the actions to be performed are independent of previous states, depending only on the current state. Future states depend on the current state and the chosen actions. Q-learning, however, does not require knowledge of the overall environment; it only needs to know which actions can be chosen in the current state. Therefore, Q-learning can be used to solve the joint scheduling strategy between repair teams and transportation teams.

[0101] Based on the above, this invention provides a joint scheduling method for road network emergency repair and material distribution, comprising the following steps:

[0102] S1. Obtain the road network parameters of the road network structure and set t=0.

[0103] S2. Initialize the state space and action space in the emergency repair decision model and the transportation decision model; wherein, the emergency repair decision model and the transportation decision model are both triplets composed of state space, action space and reward function.

[0104] S3. Solve the emergency repair decision-making model based on the Q-learning algorithm to obtain the emergency repair strategy H of the emergency repair team. After all demand points in the road network structure can be reached by the transport team, send the updated road network information to the transport decision-making model based on the buffer table mechanism, and solve the transport decision-making model based on the Q-learning algorithm to obtain the transport strategy P of the transport team.

[0105] S4. Calculate the objective function value f(P) of the objective function based on the emergency repair strategy H and the transport strategy P; where, the objective function is i represents the nodes in the road network structure, C i ′0 represents the cumulative time cost from the reserve point 0 to the demand point i under the rescue strategy P, I i represents the degree of disaster or the importance of node i, V d represents the set of demand points.

[0106] S5. Judge whether the function value f(P) < F holds. If so, let F = f(P), and update the emergency repair strategy H to the action space in the emergency repair decision-making model and update the transport strategy P to the action space in the transport decision-making model based on the buffer table of the lock mechanism; if not, directly go to S6; where, F represents the historical optimal objective function value.

[0107] S6. Let t = t + 1, and judge whether t > T max holds. If not, go to S3; if so, output the objective function value f(P), and use the emergency repair strategy H and the transport strategy P as the final joint scheduling plan.

[0108] Based on the above processing, compared with the prior art, the present invention is improved by using the Q-learning algorithm in reinforcement learning, and proposes a joint scheduling method for road network emergency repair and material distribution based on double-layer interactive Q-learning, which can feedback the joint scheduling decision under the objective function to the emergency repair team and the transport team, enabling the two to make decisions beneficial to emergency response simultaneously, so as to achieve the purpose of joint scheduling and effectively reduce the time consumed for solving.

[0109] For step S1, the road network parameters include the nodes in the road network, the degree of disaster of the nodes, the distances between different nodes, the reserve points and other contents of the foregoing basic information.

[0110] For step S2, the emergency repair decision-making model includes a state space S r 、an action space A r and a reward function R r .

[0111] Among them, the state space s r is composed of a triple ; A list of demand points that the transport team can already reach; ε is a list of the shortest path lengths from each reachable demand point i to reserve point 0; ε is a list of repaired road segments.

[0112] Action Space A r This indicates the location of the damaged road sections that the repair team can reach from their current position.

[0113] Reward function R r Related to the connectivity status of the demand point, when the repair team is in the first repair state From action set a r After selecting and repairing the damaged section, the next repair phase will begin. hour;

[0114] If no new demand is met, then the reward function R r The calculation formula is: l ij Represents the distance from node i to node j, t ij This refers to road sections that are impassable. ij The repair time cost of the repair team, and v represents the average speed of the repair team on the passable road section;

[0115] If a new demand point is met, the reward function R... r The calculation formula is: Indicates the second emergency repair status. The corresponding set of nodes Indicates the first emergency repair status. The corresponding set of nodes, λ represents the preset weighting weight, and D i0 This represents the shortest distance from demand node i to reserve point 0.

[0116] The emergency repair decision model represents the decision-making model of the emergency repair team. In actual work, the decision-making model of the emergency repair team can be described as a triple consisting of a state space, an action space, and a reward function: <S r A r ,R r >.

[0117] Among them, the status of the emergency repair team is s r By triplet constitute:

[0118]

[0119] in, A list of reachable demand points; ε is a list of the shortest path lengths from each reachable demand point i to reserve point 0; ε is a list of repaired road segments.

[0120] The action set (i.e., action space) A of the emergency repair team r With the current state Relevant. The repair team decides which damaged road section to repair based on its current location. To execute repairs more efficiently, the repair team's action set A... r Defined as the set of damaged road segments reachable from the current location:

[0121] A r ={e ij ∣ε ij =0∧(θ) i =1∨θ j =1)}

[0122] The reward function R of the emergency repair team r Depending on the connectivity of the demand points, the first thing to explore in the process of repairing a damaged road network is the damaged road section that is closest to the current location.

[0123] When the repair team is in operation From action set a r Select damaged road section e ij After emergency repairs, it will reach the next stage. There are usually two situations:

[0124] No new demand points have been identified, that is At that time, although no new demand points were connected after the damaged road section was repaired, it enabled... The shorter D′ has been updated. i0 This reduces travel time. In this situation, the repair team performs action e. ij The immediate reward value afterward

[0125]

[0126] New demand points have been identified, namely At this time, after the damaged road section is repaired, the new demand point is connected to the reserve point 0. In order to open up all demand points as soon as possible, the repair team will be given a higher reward, and the higher the degree of damage to the demand point being opened, the higher the reward. At this time, the repair team performs action e. ij The reward afterward is

[0127]

[0128] Where λ∈(0,1) is the weighting weight, representing the reward function's preference for the repair team's efficiency in terms of repair efficiency and traffic efficiency.

[0129] The shorter the shortest path length, the less the cumulative time cost, and the greater the importance of the newly opened demand node by the repair team, the greater the immediate reward for the repair team. The repair team indirectly affects the objective function. The higher the repair efficiency of the repair team, the sooner the road segment can be opened, thereby reducing the travel time of the transport team to a certain extent and making the objective function more optimal.

[0130] In some embodiments, the emergency repair decision model includes a state space S. d Action Space A d and the reward function R d .

[0131] Wherein, the state space S d By binary constitute; This represents a list of reachable demand points. A list of shortest path lengths from rescued demand point i to reserve point 0;

[0132] Action Space A d This represents the set of nodes that the transport team can reach from its current location;

[0133] When the transport team is in the first transport state From action space A d Select the demand node i for rescue, and reach the second transportation state. At that time, the reward function R d The calculation formula is:

[0134] Where η and β are both preset control variables, and D mi This represents the shortest distance from the current location of the transport team to the demand node i; r i This represents the overall reward after rescuing node i, and the calculation formula is: r′ i Let P represent the total reward in the previous period, L represent the preset parameters, and f(P) represent the objective function value.

[0135] The transportation decision model represents the decision-making model of the transportation fleet, which consists of a triple consisting of a state space, an action space, and a reward function: d A d ,R d > In actual work, the state S of the transportation team d By binary constitute:

[0136]

[0137] in, A list of reachable demand points; ​This is a list of the shortest path lengths from rescued point i to reserve point 0; for example... Figure 1 As shown, the transport team departs from reserve point 0.

[0138] The action set a of the transport team d Not only with the current state It's also related to the repair team's repair strategy (H). Different repair strategies will connect different demand points, and the transport team's operational space will also differ. For higher rescue efficiency, transport demand points...

[0139] a d ={i∣θ i =1∧μ i =0, i∈V d}

[0140] The goal is for the transport fleet to reach the demand point quickly, and for a strategy P with a better joint scheduling objective function, each action within it should be given an overall reward based on the objective function. The update strategy for this overall reward r is as follows:

[0141]

[0142] Where L is a given parameter, and f(P) is the objective function of joint scheduling.

[0143] When the transport team is in status From action space a d After selecting the demand point i for rescue, the next state is reached. Reward function R d Designed as:

[0144]

[0145] Where η and β are control variables, r i For the overall return of the transport team performing action i, D mi It is the shortest distance from the position before the action is performed to the required point i.

[0146] The transport team hopes to open up all demand points as soon as possible. The closer to a newly opened demand point, the greater the immediate reward value. It is also related to the objective function. The better the objective function, the greater the overall reward for performing the current action, and the greater the immediate reward value.

[0147] In some embodiments, step S3 includes the following:

[0148] S301, The repair team, based on a greedy strategy, starts from action space A r Select and execute action a1, and input the set of damaged road sections that the repair team has repaired into the buffer table. and cumulative time cost C i0 ;

[0149] S302, the emergency repair team changed the emergency repair status from... Transition to Calculate the corresponding Q r Value, and update the repair team's Q r Value table;

[0150] S303. Determine whether the emergency repair status has reached the final state. If yes, obtain the emergency repair strategy H of the emergency repair team. If no, update the action space of the emergency repair decision model and proceed to step S301.

[0151] S304, the transport team retrieves the set of repaired damaged road sections from the buffer table. and cumulative time cost C i0 And update the road network information, and based on a greedy strategy, start from action space A d Select and execute action a2;

[0152] S305, the transport team will change the transport status from Transition to transport status Calculate the corresponding Q d Value, and update the transport team's Q. d Value table;

[0153] S306. Determine whether the transportation state has reached the final state. If yes, obtain the transportation strategy P of the transportation team. If no, update the action space of the transportation decision model and proceed to step S304.

[0154] In actual work process, such as Figure 2 As shown, Figure 2 This is a flowchart illustrating a joint scheduling method for road network emergency repair and material distribution, provided as an embodiment of the present invention. Within each learning cycle (i.e., t in this invention), the repair team and the transport team can respectively... r A r ,R r >and d A d ,R d >, thus calculating the corresponding Q value, i.e., Q r and Q d .

[0155] When all demand points are reachable, the repair team exchanges road network information with the transport team through a buffer table. Furthermore, an optimal action set update strategy is used to feed back better decisions to both the repair and transport teams, thereby achieving joint scheduling.

[0156] Inspired by the producer-consumer problem in operating systems, this invention simulates repair teams and transportation teams as producers and consumers, respectively, who exchange information through a lock-based buffer table. See also​​ Figure 3 ,like Figure 3 As shown, the interaction process between the repair team and the transport team in each cycle is as follows:

[0157] After each damaged road section is repaired, the repair team enters the set of damaged road sections repaired and the cumulative time cost into the buffer table in sequence. Once the repair team has met all the reachability requirements, writing data to the buffer table stops. Before each action selection, the transport team reads the repair team's cumulative time cost C from the buffer table. i0 and the cumulative time cost C of the transport team i Compare '0', if C i '0' is greater than the j-th C in the buffer table i0 Less than the (j+1)th C i0 Then the j-th element in the buffer table will be... Retrieve and update the road network status of the transport team until all demand points are met and rescued.

[0158] Then, after each training cycle, the joint scheduling objective function f(P) based on the repair strategy H and the transportation strategy P is calculated and compared with the historical best objective function value. If it is better (i.e. less than the historical best objective function value), it means that the joint scheduling strategy in this training cycle is better, and the state-action sets corresponding to strategy H and strategy P are updated to their respective optimal action sets.

[0159] In actual operation, the repair team or transportation team selects actions based on a greedy strategy. With a high probability, they randomly select an action from the optimal action set of their decision model, while with a low probability, they abandon historical learning experience and randomly select an action from the action set. In this way, the repair team and transportation team have a certain probability of selecting actions that are beneficial to the joint scheduling objective function, rather than selecting actions that are only beneficial to their own objective functions, thereby achieving the purpose of joint scheduling and improving learning efficiency.

[0160] Furthermore, to avoid Q-learning from getting stuck in local optima, it is necessary to ensure that Q-learning has a certain degree of exploratory nature in the early stages and a certain degree of convergence in the later stages.

[0161] Extensive experiments have demonstrated that the best results are achieved when the process explores during the first two-thirds of the cycle and converges during the last third. Therefore, this invention defines the first two-thirds of the cycle as an exploratory process, where Q-learning randomly selects an action from the optimal action set in the decision model with a low probability, and randomly selects an action from the action set with a high probability. During the last third of the cycle, Q-learning is in a convergent process, where Q-learning randomly selects an action from the optimal action set in the decision model with a high probability, and randomly selects an action from the action set with a low probability.

[0162] In practical applications, the road network repair and material distribution joint scheduling method based on two-layer interactive Q-learning provided by this invention is used to solve the problem of joint scheduling of road network repair and material distribution under disaster-stricken road networks. The specific steps are as follows:

[0163] Step a: Initialize the relevant parameters of the road network model, decision model, and Q-learning.

[0164] Step b: The repair team and the transport team depart from reserve point 0 at the same time and initialize their respective states S. r S d And action sets A1 and A2.

[0165] Step c: The repair team selects and executes action a1 according to the greedy strategy, and inputs the set of repaired damaged road sections and the cumulative time cost into the buffer table.

[0166] Step d: The repair team starts from the status... Transition to state Calculate the corresponding Q r Value and update the repair team's Q r Value table:

[0167]

[0168] Here, α and γ both represent preset parameter values.

[0169] Step e: If the final state has been reached, end the learning of the repair team in this round; otherwise, update the action set A1 of the repair team and go to step 3 to continue learning until the complete strategy H is obtained.

[0170] Step f: The transport team reads data from the buffer table and updates the road network status, and selects and executes action a2 according to the greedy strategy;

[0171] Step g, the transport team from the status Transition to state Calculate the corresponding Q d Value and update the transport team's Q. d Value table:

[0172]

[0173] Step h: If the final state has been reached, end the current round of transport team learning; otherwise, update the action set A2 of the repair team and proceed to step 6 to continue learning until the complete strategy P is obtained.

[0174] Step i: Calculate the joint scheduling objective function value f(P) based on policy H and policy P. If it is better than the historical best objective function value, update the state-action sets of policy H and policy P to their respective optimal action sets; otherwise, do not update the optimal action set and proceed to the next step.

[0175] Step j: If the maximum number of training cycles has been reached, end the training and output the current best objective function value and the corresponding optimal joint scheduling strategy (strategy H and strategy P) for the repair team and the transportation team; otherwise, return to step b to continue training.

[0176] To verify the effectiveness of the technical solution provided by this invention, the technical solution was experimentally compared with traditional ant colony optimization (ACO) and dynamic programming (DP), such as... Figure 4-6 As shown, IQLJS represents the joint scheduling method for road network emergency repair and material distribution based on two-layer interactive Q-learning proposed in this invention.

[0177] based on Figure 4-5 As can be seen from the content, the IQLJS method proposed in this invention achieves a better mean objective function and faster execution time than the ACO and DP algorithms. Specifically, the IQLJS algorithm reduces its running time through an optimal action set update strategy, thus obtaining a superior objective function. Furthermore, the interaction between the repair team and the transport team through the buffer table enables them to quickly make reasonable joint scheduling decisions. Therefore, compared to the ACO and DP algorithms, the IQLJS method can obtain a better objective function within a reasonable timeframe.

[0178] like Figure 6 As shown, from left to right, they represent the DP algorithm, ACO algorithm, and IQLJS method. According to... Figure 6 As can be seen from the content, the repair rate of the technical solution provided by this invention is also higher than that of the ACO algorithm and the DP algorithm.

[0179] This invention also provides a joint dispatching system for road network emergency repair and material distribution, the system comprising:

[0180] The first acquisition module is used to execute S1, acquire the road network parameters of the road network structure, and set t=0;

[0181] The initialization module is used to execute S2 and initialize the state space and action space in the emergency repair decision model and the transportation decision model; wherein, the emergency repair decision model and the transportation decision model are both triplets composed of state space, action space and reward function;

[0182] A second acquisition module, configured to execute S3, solve the emergency repair decision-making model based on the Q-learning algorithm to obtain the emergency repair strategy H of the emergency repair team. After all demand points in the road network structure can be reached by the transportation team, send the updated road network information to the transportation decision-making model based on the buffer table mechanism, and solve the transportation decision-making model based on the Q-learning algorithm to obtain the transportation strategy P of the transportation team;

[0183] A first calculation module, configured to execute S4, calculate the objective function value f(P) of the objective function based on the emergency repair strategy H and the transportation strategy P; where the objective function is i represents a node in the road network structure, C i ′0 represents the cumulative time cost from the reserve point 0 to the demand point i under the rescue strategy P, I i represents the disaster-affected degree or importance degree of node i, V d represents the set of demand points;

[0184] A first judgment module, configured to execute S5, judge whether the function value f(P) < F holds. If so, let F = f(P), and update the emergency repair strategy H to the action space in the emergency repair decision-making model and update the transportation strategy P to the action space in the transportation decision-making model based on the buffer table of the lock mechanism; if not, directly go to S6; where F represents the historical optimal objective function value;

[0185] A second judgment module, configured to execute S6, let t = t + 1, and judge whether t > T max holds. If not, go to S3; if so, output the objective function value f(P), and use the emergency repair strategy H and the transportation strategy P as the final joint scheduling plan.

[0186] It can be understood that the joint scheduling system for road network emergency repair and material distribution provided in the embodiments of the present invention corresponds to the above-mentioned joint scheduling method for road network emergency repair and material distribution. The explanations, examples, beneficial effects, etc. of the relevant content can refer to the corresponding content in the joint scheduling method for road network emergency repair and material distribution, and will not be elaborated here.

[0187] The embodiments of the present invention also provide an electronic device, as Figure 7 shown, including a processor 701, a communication interface 702, a memory 703, and a communication bus 704. Among them, the processor 701, the communication interface 702, and the memory 703 complete mutual communication through the communication bus 704,

[0188] The memory 703 is used to store a computer program;

[0189] When the processor 701 executes the program stored in the memory 703, it implements any of the above-mentioned joint scheduling methods for road network emergency repair and material distribution.

[0190] The communication bus mentioned in the above electronic devices can be a Peripheral Component Interconnect (PCI) bus or an Extended Industry Standard Architecture (EISA) bus, etc. This communication bus can be divided into address bus, data bus, control bus, etc. For ease of illustration, only one thick line is used to represent it in the diagram, but this does not mean that there is only one bus or one type of bus.

[0191] The communication interface is used for communication between the aforementioned electronic devices and other devices.

[0192] The memory may include random access memory (RAM) or non-volatile memory (NVM), such as at least one disk storage device. Optionally, the memory may also be at least one storage device located remotely from the aforementioned processor.

[0193] The processors mentioned above can be general-purpose processors, including central processing units (CPUs), network processors (NPs), etc.; they can also be digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, or discrete hardware components.

[0194] In another embodiment of the present invention, a computer-readable storage medium is also provided, which stores a computer program that, when executed by a processor, implements the steps of any of the above-described joint scheduling methods for road network emergency repair and material distribution.

[0195] In another embodiment of the present invention, a computer program product containing instructions is also provided, which, when run on a computer, causes the computer to execute any of the joint scheduling methods for road network emergency repair and material distribution described in the above embodiments.

[0196] In the above embodiments, implementation can be achieved, in whole or in part, through software, hardware, firmware, or any combination thereof. When implemented in software, it can be implemented, in whole or in part, as a computer program product. The computer program product includes one or more computer instructions. When the computer program instructions are loaded and executed on a computer, all or part of the processes or functions described in the embodiments of the present invention are generated.

[0197] The computer may be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions may be stored in a computer-readable storage medium or transmitted from one computer-readable storage medium to another. For example, the computer instructions may be transmitted from one website, computer, server, or data center to another via wired (e.g., coaxial cable, fiber optic, digital subscriber line (DSL)) or wireless (e.g., infrared, wireless, microwave, etc.) means. The computer-readable storage medium may be any available medium that a computer can access, or a data storage device such as a server or data center that integrates one or more available media. The available medium may be a magnetic medium (e.g., floppy disk, hard disk, magnetic tape), an optical medium (e.g., DVD), or a semiconductor medium (e.g., a solid-state drive (SSD)).

[0198] In summary, compared with the prior art, the technical solution provided by the present invention has the following beneficial effects:

[0199] 1. Compared with existing technologies, the technical solution provided by this invention innovates in solving this type of problem. Unlike the traditional Ant Colony Algorithm (ACO) and Dynamic Programming (DP), it improves upon the Q-learning algorithm in reinforcement learning and proposes a joint scheduling method for road network repair and material distribution based on two-layer interactive Q-learning (IQLJS). Experimental results show that the average objective function obtained by the IQLJS method proposed in this invention is superior to that of the ACO and DP algorithms.

[0200] 2. Most existing research separates road network repair from material scheduling, which does not reflect real-world scenarios. Alternatively, their joint scheduling models are too complex, making solutions extremely difficult, and the proposed algorithms cannot handle situations with large-scale road networks and high damage rates. This invention optimizes the joint scheduling model and is applicable to scenarios with large-scale road networks and high damage rates.

[0201] 3. The optimal action set update strategy provided by this invention not only solves the problem of slow table updates, but also feeds back the joint scheduling decision under the optimal objective function to the repair team and the transportation team, so that both can make decisions that are conducive to emergency response, thereby achieving the purpose of joint scheduling.

[0202] 4. The technical solution provided by this invention establishes a buffer table interaction method based on a lock mechanism, thereby enabling efficient communication between the repair team and the transportation team, avoiding data confusion and inconsistency, and ensuring the accuracy of experimental results.

[0203] It should be noted that, in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitations, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes said element.

[0204] The above embodiments are only used to illustrate the technical solutions of the present invention, and are not intended to limit it. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A joint scheduling method for road network emergency repair and material distribution, characterized in that, The method includes: S1. Obtain the road network parameters of the road network structure, and let ; S2. Initialize the state space and action space in the emergency repair decision model and the transportation decision model; wherein, the emergency repair decision model and the transportation decision model are both triplets composed of state space, action space and reward function; S3. Solve the emergency repair decision model based on the Q-learning algorithm to obtain the emergency repair strategy of the repair team. Once all demand points in the road network structure can be reached by the transport fleet, the updated road network information is sent to the transport decision model based on the buffer table mechanism. The transport decision model is then solved using the Q-learning algorithm to obtain the transport fleet's transport strategy. ; S4. Based on the aforementioned emergency repair strategy and the aforementioned transportation strategy Calculate the objective function value Wherein, the objective function is ; Represents nodes in the road network structure. This indicates that under rescue strategy P, the distance from reserve point 0 to demand point... Cumulative time cost, Represents a node The extent of the disaster or the key points Importance Represents a set of demand points; S5. Determine the function value Is it true? If so, then let And based on the buffer table of the locking mechanism, the emergency repair strategy is... Update the action space in the emergency repair decision model and the transportation strategy Update the action space of the transportation decision model; otherwise, proceed directly to S6; where, This represents the historical optimal objective function value; S6, Order ,judge Check if the condition is true. If not, go to S3; if yes, output the objective function value. And the emergency repair strategy H and the transportation strategy P are used as the final joint scheduling scheme; Step S3 involves solving the emergency repair decision model using the Q-learning algorithm to obtain the emergency repair strategy for the repair team. Once all demand points in the road network structure can be reached by the transport fleet, the updated road network information is sent to the transport decision model based on the buffer table mechanism. The transport decision model is then solved using the Q-learning algorithm to obtain the transport fleet's transport strategy. ,include: S301, the repair team uses a greedy strategy to move from the action space Select and execute actions And input the set of damaged road sections that the repair team has repaired into the buffer table. and cumulative time expenditure ; S302, the emergency repair team changed the emergency repair status from... Transition to Calculate the corresponding Value, and update the emergency repair team. Value table; S303. Determine whether the emergency repair status has reached the final state. If so, obtain the emergency repair team's repair strategy. If not, update the action space of the emergency repair decision model and proceed to step S301; S304, the transport team retrieves the set of repaired damaged road sections from the buffer table. and cumulative time expenditure And update the road network information, and based on a greedy strategy from the action space Select and execute actions ; S305, the transport team will change the transport status from Transition to transport status , calculate the corresponding Value, and update the transport team Value table; S306. Determine if the transportation status has reached the final state. If so, obtain the transportation strategy of the transportation team. If not, update the action space of the transportation decision model and proceed to step S304.

2. The method according to claim 1, characterized in that, The emergency repair decision model includes a state space. Action space and return function ; Wherein, the state space By triplet constitute; List of demand points that the transport team can already reach; For each reachable demand point List of shortest path lengths to reserve point 0; List of repaired road sections; Action space This indicates the location of the damaged road sections that the repair team can reach from their current position. Reward function Related to the connectivity status of the demand point, when the repair team is in the first repair state From Action Set After selecting the damaged section for emergency repair, the next repair phase will begin. hour; If no new demand points are addressed, then the reward function... The calculation formula is: ; Indicates that by node To the node distance, This refers to sections of road that are impassable. The time and expenses incurred by the repair team in carrying out emergency repairs. This indicates the average speed of the repair team on the passable road section; If a new demand point is met, then the reward function... The calculation formula is: ; Indicates the second emergency repair status The corresponding set of nodes Indicates the first emergency repair status. The corresponding set of nodes Indicates the preset weighting weights, Indicates demand node The shortest distance to reserve point 0.

3. The method according to claim 1, characterized in that, The emergency repair decision model includes a state space. Action space and return function ; Among them, the state space By binary constitute; This represents a list of reachable demand points. For the points of need for rescue List of shortest path lengths to reserve point 0; Action space This represents the set of nodes that the transport team can reach from its current location; When the transport team is in the first transport state From the action space Select the demand node Conduct rescue operations and reach the second transportation status. At that time, the reward function The calculation formula is: ;in, and All are preset control variables. This indicates the distance of the transport fleet from its current location to the demand node. The shortest distance; This indicates that the requirement node has been completed. The overall reward after the rescue, calculated using the following formula: ; Indicates the total rewards within the previous period. Indicates the preset parameters, This represents the objective function value.

4. A joint dispatching system for road network emergency repair and material distribution, characterized in that, The system includes: The first acquisition module is used to execute S1, acquire the road network parameters of the road network structure, and set... ; The initialization module is used to execute S2 and initialize the state space and action space in the emergency repair decision model and the transportation decision model; wherein, the emergency repair decision model and the transportation decision model are both triplets composed of state space, action space and reward function; The second acquisition module is used to execute S3 and solve the emergency repair decision model based on the Q-learning algorithm to obtain the emergency repair strategy of the emergency repair team. Once all demand points in the road network structure can be reached by the transport fleet, the updated road network information is sent to the transport decision model based on the buffer table mechanism. The transport decision model is then solved using the Q-learning algorithm to obtain the transport fleet's transport strategy. ; The first calculation module is used to execute S4 based on the emergency repair strategy. and the aforementioned transportation strategy Calculate the objective function value Wherein, the objective function is ; Represents nodes in the road network structure. This represents the cumulative time cost from reserve point 0 to demand point i under rescue strategy P. Represents a node The extent of the disaster or the key points Importance Represents a set of demand points; The first judgment module is used to execute S5 and judge the function value. Is it true? If so, then let And based on the buffer table of the locking mechanism, the emergency repair strategy is... Update the action space in the emergency repair decision model and the transportation strategy Update the action space of the transportation decision model; otherwise, proceed directly to S6; where, This represents the historical optimal objective function value; The second judgment module is used to execute S6 and command. ,judge Check if the condition is true. If not, go to S3; if yes, output the objective function value. And the emergency repair strategy H and the transportation strategy P are used as the final joint scheduling scheme; The second acquisition module includes: The first input unit is used to execute S301, and the repair team uses a greedy strategy to access the action space. Select and execute actions And input the set of damaged road sections that the repair team has repaired into the buffer table. and cumulative time expenditure ; The first update unit is used to execute S302, whereby the repair team changes the repair status from... Transition to Calculate the corresponding Value, and update the emergency repair team. Value table; The first judgment unit is used to execute S303 to determine whether the emergency repair status has reached the final state. If so, it obtains the emergency repair strategy of the emergency repair team. If not, update the action space of the emergency repair decision model and proceed to step S301; The first acquisition unit is used to execute S304, whereby the transport team retrieves the set of repaired damaged road sections from the buffer table. and cumulative time expenditure And update the road network information, and based on a greedy strategy from the action space Select and execute actions ; The second update unit is used to execute S305, which changes the transport status of the transport team from... Transition to transport status , calculate the corresponding Value, and update the transport team Value table; The second judgment unit is used to execute S306, determine whether the transportation status has reached the final state, and if so, obtain the transportation strategy of the transportation team. If not, update the action space of the transportation decision model and proceed to step S304.

5. The system according to claim 4, characterized in that, The emergency repair decision model includes a state space. Action space and return function ; Wherein, the state space By triplet constitute; List of demand points that the transport team can already reach; For each reachable demand point List of shortest path lengths to reserve point 0; List of repaired road sections; Action space This indicates the location of the damaged road sections that the repair team can reach from their current position. Reward function Related to the connectivity status of the demand point, when the repair team is in the first repair state From Action Set After selecting the damaged section for emergency repair, the next repair phase will begin. hour; If no new demand points are addressed, then the reward function... The calculation formula is: ; Indicates that by node To the node distance, This refers to sections of road that are impassable. The time and expenses incurred by the repair team in carrying out emergency repairs. This indicates the average speed of the repair team on the passable road section; If a new demand point is met, then the reward function... The calculation formula is: ; Indicates the second emergency repair status The corresponding set of nodes Indicates the first emergency repair status. The corresponding set of nodes Indicates the preset weighting weights, Indicates demand node The shortest distance to reserve point 0.

6. The system according to claim 4, characterized in that, The emergency repair decision model includes a state space. Action space and return function ; Among them, the state space By binary constitute; This represents a list of reachable demand points. For the points of need for rescue List of shortest path lengths to reserve point 0; Action space This represents the set of nodes that the transport team can reach from its current location; When the transport team is in the first transport state From the action space Select the demand node Conduct rescue operations and reach the second transportation status. At that time, the reward function The calculation formula is: ;in, and All are preset control variables. This indicates the distance of the transport fleet from its current location to the demand node. The shortest distance; This indicates that the requirement node has been completed. The overall reward after the rescue, calculated using the following formula: ; Indicates the total rewards within the previous period. Indicates the preset parameters, This represents the objective function value.

7. An electronic device, characterized in that, It includes a processor, a communication interface, a memory, and a communication bus, wherein the processor, the communication interface, and the memory communicate with each other through the communication bus; Memory, used to store computer programs; A processor, when executing a program stored in memory, implements the steps of the method described in any one of claims 1-3.

8. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a computer program that, when executed by a processor, implements the steps of the method described in any one of claims 1-3.

Images