An intelligent decision-making system for road maintenance and repair

Abstract

The present invention discloses an intelligent decision-making system for road maintenance and repair, comprising a road state parameter acquisition module, a maintenance and repair reference factor acquisition module, a road maintenance and repair plan generation module, a road maintenance and repair plan analysis module and a road maintenance optimal plan output module; the present invention combines expert decision-making knowledge with the use of a reinforcement learning algorithm, continuously optimizes the maintenance decision-making method, ensures the correctness of maintenance measures, and improves maintenance benefits.

Description

Technical Field

[0001] The invention relates to the field of road maintenance, and in particular to an intelligent decision-making system for road maintenance and repair. Background Art

[0002] In actual work, highway managers have done a lot of inspection work in daily maintenance and special maintenance, and collected a lot of inspection data. However, many valuable data have not been effectively utilized. Only simple statistics have been performed, and no correlation has been formed with other multi-source maintenance data, and in-depth mining and analysis have been carried out. As a result, it is impossible to grasp the causes and temporal and spatial expansion laws of pavement diseases, resulting in weak data support for maintenance and repair plan decisions. On the other hand, when there are multiple maintenance and repair measures to choose from in the face of pavement damage, there is a lack of scientific selection basis and no choice, which makes it impossible to maximize the maintenance and repair benefits of the pavement. Summary of the invention

[0003] The purpose of the present invention is to provide an intelligent decision-making system for road maintenance and repair, including a road state parameter acquisition module, a maintenance and repair reference factor acquisition module, a road maintenance and repair plan generation module, a road maintenance and repair plan analysis module and a road maintenance optimal plan output module.

[0004] The road state parameter acquisition module acquires the state parameters of the road to be maintained and repaired, and sends them to the road maintenance and repair plan generation module.

[0005] The status parameters of the road to be maintained and repaired include road age, road type, pavement structure type, surface composition, traffic conditions, road grade, regional zoning, damage rate DR / pavement condition index PCI, pavement average international roughness index IRI / riding quality index RQI, pavement structure representative deflection DEF / pavement structure strength index PSSI, average maximum rutting depth RD / rutting depth index RDI, pavement skid resistance index SRI, first major damage type, and second major damage type.

[0006] The maintenance and repair reference factor acquisition module acquires the maintenance and repair reference factors and sends them to the road maintenance and repair program analysis module.

[0007] The maintenance and repair reference factors are determined by expert knowledge; cluster factor analysis is used to determine the effective classification fields, and the scoring standards of the maintenance and repair reference factors are determined by expert knowledge and industry standards.

[0008] Preferably, the maintenance reference factors include the materials used to implement the road maintenance plan and the corresponding priority, the grade of the road to be maintained, the cost, the city to which it belongs, and the traffic grade.

[0009] The road maintenance and repair plan generation module stores a road maintenance and repair plan generation model.

[0010] The road maintenance and repair plan generation model includes four tuples<S,A,P,R> Among them, S represents the state parameters of the road to be maintained. A represents several road maintenance reference plans. P represents the probability of the state change of the road surface receiving maintenance. R represents the reward value function.

[0011] The reference plan for road maintenance and repair includes repair plan, daily maintenance plan and preventive maintenance plan.

[0012] Preventive maintenance plan M1 = {Daily inspection M1-1. Daily maintenance M1-2. Daily repair M1-3}. Daily maintenance plan M2 = {Sealing layer M2-1. Functional cover M2-2. Pre-maintenance combination M2-3}. Repair plan M3 = {Milling and covering M3-1. Structural reinforcement M3-2. Surface layer renovation M3-3. Base layer renovation M3-4. Roadbed treatment M3-5}.

[0013] The reward value function is as follows:

[0014]

[0015] In the formula, w i is the weight coefficient of the reward value of the i-th performance indicator. index_i represents the i-th performance index. i=1,2,…,n. R t+1 It is the total reward value after implementing a road maintenance and repair reference plan. is the reward value of the ith performance indicator after executing a certain road maintenance and repair reference plan. The performance indicator is one or more of the state parameters.

[0016] Preferably, the reward value function is as follows:

[0017]

[0018] In the formula, w PCI 、w RQI 、w RDI It is the weight coefficient of the reward value of the road condition index PCI, the driving quality index RQI, and the rutting depth index RDI.

[0019] Road Condition Index PCI Reward Value As shown below:

[0020]

[0021] Where c1, c2, and c3 are the coefficients of road damage quantity, road damage distribution, and main types of road damage. t+1 is the road surface damage. DR_main is the road damage distribution. DR_type It is the main type of road damage.

[0022] Ride Quality Index RQI Reward Value As shown below:

[0023]

[0024] Where d1 and d2 are the road surface roughness condition coefficient and the roughness extreme coefficient. IRI t+1 is the road surface roughness condition. IRI_max is the optimal road surface roughness.

[0025] Rutting Depth Index (RDI) Bonus Value As shown below:

[0026]

[0027] Where, e1 and e2 are the rutting coefficient and the extreme coefficient of rutting depth. t+1 is the rutting depth. RD_max is the maximum rutting depth.

[0028] The road maintenance and repair scheme generation model receives state parameters of the road to be maintained and repaired, and calculates reward values ​​under different road maintenance and repair reference schemes.

[0029] The road maintenance and repair plan generation module writes the road maintenance and repair reference plan with a reward value greater than a preset threshold into the road maintenance and repair plan set, and sends it to the road maintenance and repair plan analysis module.

[0030] The road maintenance and repair scheme analysis module evaluates the road maintenance and repair schemes in the road maintenance and repair scheme set according to the maintenance and repair reference factors, determines the priority of the road maintenance and repair scheme, and sends it to the road maintenance optimal scheme output module.

[0031] The road maintenance and repair plan analysis module evaluates the road maintenance and repair plans in the road maintenance and repair plan set by using the performer-critic model to calculate the reward (Rt+1) of each road maintenance and repair plan, and t+1 Sort the road maintenance and repair plans in descending order to determine their priorities.

[0032] The technical effect of the present invention is unquestionable. The present invention combines expert decision-making knowledge with the use of reinforcement learning algorithms to continuously optimize maintenance decision-making methods, ensure the correctness of maintenance measures, and improve maintenance benefits. BRIEF DESCRIPTION OF THE DRAWINGS

[0033] Figure 1 System decision-making process. DETAILED DESCRIPTION

[0034] The present invention is further described below in conjunction with the embodiments, but it should not be understood that the above subject matter of the present invention is limited to the following embodiments. Without departing from the above technical ideas of the present invention, various substitutions and changes are made according to the common technical knowledge and customary means in the art, which should all be included in the protection scope of the present invention.

[0035] Embodiment 1:

[0036] See also Figure 1 , a road maintenance and repair intelligent decision-making system, including a road state parameter acquisition module, a maintenance and repair reference factor acquisition module, a road maintenance and repair plan generation module, a road maintenance and repair plan analysis module and a road maintenance optimal plan output module and a database.

[0037] The road state parameter acquisition module acquires the state parameters of the road to be maintained and repaired, and sends them to the road maintenance and repair plan generation module.

[0038] The status parameters of the road to be maintained and repaired include road age, road type, pavement structure type, surface composition, traffic conditions, road grade, regional zoning, damage rate DR / pavement condition index PCI, pavement average international roughness index IRI / riding quality index RQI, pavement structure representative deflection DEF / pavement structure strength index PSSI, average maximum rutting depth RD / rutting depth index RDI, pavement skid resistance index SRI, first major damage type, second major damage type (for example, the first major damage type is transverse cracks, and the second major damage type is potholes).

[0039] The maintenance and repair reference factor acquisition module acquires the maintenance and repair reference factors and sends them to the road maintenance and repair program analysis module.

[0040] The maintenance and repair reference factors are determined by expert knowledge; cluster factor analysis is used to determine the effective classification fields, and the scoring standards of the maintenance and repair reference factors are determined by expert knowledge and industry standards.

[0041] In this embodiment, the maintenance reference factors determined by expert knowledge include materials used to implement the road maintenance plan and the corresponding priority, the grade of the road to be maintained, the cost, the city to which it belongs, and the traffic grade.

[0042] The road maintenance and repair plan generation module stores a road maintenance and repair plan generation model.

[0043] The road maintenance and repair plan generation model includes four tuples<S,A,P,R> Among them, S represents the state parameters of the road to be maintained. A represents several road maintenance reference plans. P represents the probability of the state change of the road surface receiving maintenance. R represents the reward value function.

[0044] The reference plan for road maintenance and repair includes repair plan, daily maintenance plan and preventive maintenance plan.

[0045] Preventive maintenance plan M1 = {Daily inspection M1-1. Daily maintenance M1-2. Daily repair M1-3}. Daily maintenance plan M2 = {Sealing layer M2-1. Functional cover M2-2. Pre-maintenance combination M2-3}. Repair plan M3 = {Milling and covering M3-1. Structural reinforcement M3-2. Surface layer renovation M3-3. Base layer renovation M3-4. Roadbed treatment M3-5}.

[0046] The reward value function is as follows:

[0047]

[0048] In the formula, w PCI 、w RQI 、w RDI It is the weight coefficient of the reward value of the road condition index PCI, the driving quality index RQI, and the rutting depth index RDI.

[0049] Road Condition Index PCI Reward Value As shown below:

[0050]

[0051] Where c1, c2, and c3 are the coefficients of road damage quantity, road damage distribution, and main types of road damage. t+1 is the road surface damage. DR_main is the road damage distribution. DR_type It is the main type of road damage.

[0052] Ride Quality Index RQI Reward Value As shown below:

[0053]

[0054] Where d1 and d2 are the road surface roughness condition coefficient and the roughness extreme coefficient. IRI t+1 is the road surface roughness condition. IRI_max is the optimal road surface roughness.

[0055] Rutting Depth Index (RDI) Bonus Value As shown below:

[0056]

[0057] Where, e1 and e2 are the rutting coefficient and the extreme coefficient of rutting depth. t+1 is the rutting depth. RD_max is the maximum rutting depth.

[0058] The road maintenance and repair scheme generation model receives state parameters of the road to be maintained and repaired, and calculates reward values ​​under different road maintenance and repair reference schemes.

[0059] The road maintenance and repair plan generation module writes the road maintenance and repair reference plan with a reward value greater than a preset threshold into the road maintenance and repair plan set, and sends it to the road maintenance and repair plan analysis module.

[0060] The road maintenance and repair scheme analysis module evaluates the road maintenance and repair schemes in the road maintenance and repair scheme set according to the maintenance and repair reference factors, determines the priority of the road maintenance and repair scheme, and sends it to the road maintenance optimal scheme output module.

[0061] The road maintenance and repair plan analysis module evaluates the road maintenance and repair plans in the road maintenance and repair plan set by using the performer-critic model to calculate the reward (Rt+1) of each road maintenance and repair plan, and t+1 Sort the road maintenance and repair plans in descending order and determine the priority of the road maintenance and repair plans. Specifically, let the strategy obtain the state, output an action (At), and then receive a new state (St+1) and reward (Rt+1). The critic model calculates the score q of taking an action in the current state, and the performer model uses this q value to update its own strategy weight.

[0062] The road maintenance optimal solution output module outputs the road maintenance repair solution with the highest priority as the road maintenance optimal solution.

[0063] The database stores data of the road state parameter acquisition module, the maintenance and repair reference factor acquisition module, the road maintenance and repair plan generation module, the road maintenance and repair plan analysis module and the road maintenance optimal plan output module.

[0064] Embodiment 2:

[0065] A road maintenance and repair intelligent decision-making system includes a road state parameter acquisition module, a maintenance and repair reference factor acquisition module, a road maintenance and repair plan generation module, a road maintenance and repair plan analysis module and a road maintenance optimal plan output module.

[0066] The road state parameter acquisition module acquires the state parameters of the road to be maintained and repaired, and sends them to the road maintenance and repair plan generation module.

[0067] The status parameters of the road to be maintained and repaired include road age, road type, pavement structure type, surface composition, traffic conditions, road grade, regional zoning, damage rate DR / pavement condition index PCI, pavement average international roughness index IRI / riding quality index RQI, pavement structure representative deflection DEF / pavement structure strength index PSSI, average maximum rutting depth RD / rutting depth index RDI, pavement skid resistance index SRI, first major damage type, and second major damage type.

[0068] The maintenance and repair reference factor acquisition module acquires the maintenance and repair reference factors and sends them to the road maintenance and repair program analysis module.

[0069] The maintenance reference factors include the materials used to implement the road maintenance plan and the corresponding priority, the grade of the road to be maintained, the cost, the city to which it belongs, and the traffic grade.

[0070] The road maintenance and repair plan generation module stores a road maintenance and repair plan generation model.

[0071] The road maintenance and repair plan generation model includes four tuples<S,A,P,R> Among them, S represents the state parameters of the road to be maintained. A represents several road maintenance reference plans. P represents the probability of the state change of the road surface receiving maintenance. R represents the reward value function.

[0072] The reference plan for road maintenance and repair includes repair plan, daily maintenance plan and preventive maintenance plan.

[0073] Preventive maintenance plan M1 = {Daily inspection M1-1. Daily maintenance M1-2. Daily repair M1-3}. Daily maintenance plan M2 = {Sealing layer M2-1. Functional cover M2-2. Pre-maintenance combination M2-3}. Repair plan M3 = {Milling and covering M3-1. Structural reinforcement M3-2. Surface layer renovation M3-3. Base layer renovation M3-4. Roadbed treatment M3-5}.

[0074] The reward value function is as follows:

[0075]

[0076] In the formula, w i is the weight coefficient of the reward value of the i-th performance indicator. index_i represents the i-th performance index. i=1,2,…,n. R t+1 It is the total reward value after implementing a road maintenance and repair reference plan. is the reward value of the ith performance indicator after executing a certain road maintenance and repair reference plan. The performance indicator is one or more of the state parameters.

[0077] The road maintenance and repair scheme generation model receives state parameters of the road to be maintained and repaired, and calculates reward values ​​under different road maintenance and repair reference schemes.

[0078] The road maintenance and repair plan generation module writes the road maintenance and repair reference plan with a reward value greater than a preset threshold into the road maintenance and repair plan set, and sends it to the road maintenance and repair plan analysis module.

[0079] The road maintenance and repair scheme analysis module evaluates the road maintenance and repair schemes in the road maintenance and repair scheme set according to the maintenance and repair reference factors, determines the priority of the road maintenance and repair scheme, and sends it to the road maintenance optimal scheme output module.

[0080] The step of evaluating the road maintenance and repair scheme in the road maintenance and repair scheme set by the road maintenance and repair scheme analysis module includes:

[0081] For maintenance plan decisions, record historical behaviors (i.e., specific maintenance measures), i.e., A; observe the performance P that gradually decays over time. The initial rewards of all decision plans are 0, and maintenance measures that are judged to be effective obtain corresponding reward values ​​and are superimposed.

[0082] The road maintenance optimal solution output module outputs the road maintenance repair solution with the highest priority as the road maintenance optimal solution.

[0083] Embodiment 3:

[0084] An intelligent decision-making system for pavement maintenance and repair adopts a two-layer decision-making strategy. The specific solution process is shown in the figure. The first layer of decision-making is the scheme-level decision-making, which mainly serves the network-level facility management, such as maintenance planning, local and regional management; the second layer of decision-making is the measure-level decision-making, which mainly serves the project-level facility management, such as the decision-making assistance of a maintenance department or a maintenance design unit.

[0085] Decisions at the plan level are mainly based on the actual attributes and status of the road, with the goal of making scientific decisions to determine the most suitable decision conclusion for the current road section; decisions at the measure level include the selection of measures and materials and cost analysis, mainly considering the regional characteristics of the road, the environment, traffic conditions, funding constraints and maintenance level, and using user habits to characterize potential information such as local, regional and corporate decision-making habits, material and equipment advantages.

[0086] Road maintenance decision makers determine the level of maintenance to be performed based on the combination of macro-indicator limit values. The levels of maintenance include overhaul projects, repair projects, preventive maintenance projects, and routine maintenance.

[0087] The decision-making system has established a four-level set of maintenance measures, and each level corresponds to a set of its sub-levels. The first level is called the engineering level; the second level is the maintenance and repair plans of different degrees, called the plan level; the third level is different maintenance and repair measures, called the measure level; the fourth level includes the optional material combination of different measures, which is called the measure material level. Among them, the set relationship between the engineering level and the plan is as follows:

[0088] Major and medium repair projects (repair projects) M3 = {milling and covering M3-1; structural reinforcement M3-2; surface layer renovation M3-3; base layer renovation M3-4; roadbed treatment M3-5}

[0089] Preventive maintenance project M2 = {sealing layer type M2-1; functional cover M2-2; pre-curing combination M2-3}

[0090] Daily maintenance M1 = {daily inspection M1-1; daily maintenance M1-2; daily repair M1-3}

[0091] Among them, the scheme-level measures are strongly correlated with the current status of the road structure. Therefore, the primary decision-making issue of this project is the scheme decision-making issue, that is, the diagnosis problem of the detailed road structure and damage status.

[0092] 1) MDP model-solution decision

[0093] The MDP model consists of a four-tuple<S,A,P,R> Among them, S represents the state, which represents the existing parameters of the road surface; A represents the action, which represents different maintenance and repair schemes; P represents the state transition probability, which represents the probability of the state change after the road surface is maintained and repaired; R represents the reward value function, which represents the reward or punishment obtained by the agent after executing the action.

[0094] Each indicator is based on the actual storage data, and the primary detection indicator data is used first to prevent possible errors caused by the calculation and conversion of evaluation indicators and the problem of indicators not being universal due to changes in industry evaluation standards. Among them, damage-related indicators are mainly aimed at road damage correlation analysis and dimensionality reduction. Rutting damage is basically unrelated to other types of damage. It is not reasonable to simply classify damage according to deformation, damage, and surface characteristics.

[0095] Therefore, this system determines that the parameters of the MDP model related to damage include the following:

[0096] S = {road age; road type; pavement structure type; surface composition; traffic conditions; road grade; regional zoning; damage rate DR / pavement condition index PCI; pavement average international roughness index IRI / riding quality index RQI; pavement structure representative deflection DEF / pavement structure strength index PSSI; average maximum rutting depth RD / rutting depth index RDI; pavement skid resistance index SRI; first major damage type; second major damage type}, mainly including the characteristic indicators of the above 14 dimensions.

[0097] A={M3-1; M3-2; M3-3; M3-4; M3-5; M2-1; M2-2; M2-3;

[0098] M1-1; M1-2; M1-3}

[0099] According to the decay of the main dynamic indicators after a certain maintenance plan is adopted, penalties or rewards are given, and the target service life and performance decay threshold are set for different maintenance plans. Rewards are given when the pavement performance is improved and the performance decay rate or threshold does not exceed the set range; penalties are given when the performance improvement measures fail quickly and exceed the threshold. The reward value function is defined as follows:

[0100]

[0101] Among them, w i is the weight coefficient of each indicator reward value, index_i is the i-th performance indicator, and there are n indicators in total. The indicator of the reaction reward can be selected according to the actual situation of the actual road tracking indicator. This system selects PCI, RQI, and RDI as long-term tracking indicators, and the reward value function is as follows:

[0102]

[0103] in, It means that the road surface damage has not been seriously decayed, and some damage has occurred, but within a reasonable range. The specific definition is as follows:

[0104]

[0105] Among them, c1, c2, and c3 are the road surface damage coefficient, road surface damage distribution coefficient, and road surface damage main type coefficient.

[0106] It means that the road surface driving quality has not been seriously deteriorated, and the unevenness in some places is within a reasonable range. The specific definition is as follows:

[0107]

[0108] Among them, d1 and d2 are the road surface smoothness condition coefficient and the smoothness extreme coefficient.

[0109] It means that the road surface rutting has not undergone serious decay, and the local positions are uneven, but within a reasonable range. The specific definition is as follows:.

[0110]

[0111] Among them, e1 and e2 are the rutting coefficient and rutting depth extreme coefficient of the road surface flatness condition.

[0112] 2) Dynamic solution of parameters

[0113] In terms of maintenance strategy processing, repair projects and preventive maintenance will have corresponding engineering records, but within the scope of daily maintenance projects, even if large-scale daily maintenance is done, overall records are lacking. Therefore, in data processing, the plan is revised for daily maintenance facilities, and data verification is performed based on the repair status of surface damage. For situations where the amount of surface damage changes greatly, daily maintenance strategies are filled in. High-grade roads without maintenance strategies (highways, first-class highways, second-class highways, urban expressways, and urban trunk roads) default to daily maintenance, and low-grade roads without maintenance strategies (other roads other than high-grade roads) default to daily inspections. In terms of data execution, it should also be noted that in the full life cycle management of actual road operation and maintenance, even if no special maintenance and repair projects are performed, action A needs to be given at certain time intervals based on the maintenance level.

[0114] The state is the 14-dimensional road surface dynamic and static parameters mentioned above, the action is the maintenance plan adopted, and the 12 options in 1 dimension are the maintenance plans actually adopted in history. In terms of reinforcement learning, the fitted-Q learning algorithm is used to learn the optimal strategy. Since the proportion of daily maintenance projects is much higher than other strategies and the maintenance plan data samples are relatively small, in the process of randomly selecting learning samples, grouping is selected for parameter solution and used for all projects.

[0115] During the inverse reinforcement learning process, we set the initial parameters to 0 to eliminate the influence of the initial parameters (prior knowledge). All process reward parameters and weight values ​​are limited to the range of [0,1]. In addition, some indicators use the weight parameters in the existing evaluation criteria as prior knowledge to initialize the weight parameters as a control to speed up the convergence.

[0116] 3) Decision-making based on habits

[0117] Due to the large differences in costs and effectiveness between different measures in the same set of maintenance and repair plans, the maintenance and repair decision-making plan is not completely linearly related to the maintenance and repair cost. The choice of measures cannot be made simply based on the condition of the facility itself, but must be considered comprehensively in terms of economic factors, user habits, etc. In this learning process, in addition to the parameters that can be determined, such as related traffic, environment, and maintenance level, the priority of measures is added to characterize user habits.

[0118] The specific measure material set and fields can also be dynamically adjusted according to user conditions. The sample measure table is shown in the following table. The actions taken include three parameters: recommended measures, material combination, and project unit price. The status parameters include user attributes (funding source restrictions), road grade (high, low), maintenance level (determined by road grade and user attributes, non-independent parameters, implicit parameters, no input required), city, traffic level, and user habits (measure priority assignment, which can be reflected in the selection of material priority type, such as recycled and modified asphalt).

[0119] Table 1 Example measures Table I

[0120]

[0121] Table 2 Sample measures Table II

[0122]

Claims

1. An intelligent decision-making system for road maintenance and repair, characterized in that: It includes a road state parameter acquisition module, a maintenance and repair reference factor acquisition module, a road maintenance and repair plan generation module, a road maintenance and repair plan analysis module and a road maintenance optimal plan output module; The road state parameter acquisition module acquires the state parameters of the road to be maintained and repaired, and sends them to the road maintenance and repair plan generation module; The maintenance and repair reference factor acquisition module acquires the maintenance and repair reference factors and sends them to the road maintenance and repair plan analysis module; The road maintenance and repair plan generation module stores a road maintenance and repair plan generation model; The road maintenance and repair plan generation model includes four tuples<S,A,P,R> ; Where S represents the state parameter of the road to be maintained; A represents several road maintenance reference plans; P represents the probability of the state change of the road surface receiving maintenance; R represents the reward value function; The road maintenance and repair scheme generation model receives the state parameters of the road to be maintained and repaired, and calculates the reward values ​​under different road maintenance and repair reference schemes; The road maintenance and repair plan generation module writes the road maintenance and repair reference plan with a reward value greater than a preset threshold into the road maintenance and repair plan set, and sends it to the road maintenance and repair plan analysis module; The road maintenance and repair scheme analysis module evaluates the road maintenance and repair schemes in the road maintenance and repair scheme set according to the maintenance and repair reference factors, determines the priority of the road maintenance and repair scheme, and sends it to the road maintenance optimal scheme output module; The road maintenance optimal solution output module outputs the road maintenance and repair solution with the highest priority as the road maintenance optimal solution; The reward value function is as follows: In the formula, w i is the weight coefficient of the reward value of the i-th performance indicator; index_i represents the i-th performance index; i = 1, 2, ..., n; R t+1 The total reward value after implementing a certain road maintenance and repair reference plan; is the reward value of the i-th performance indicator after executing a certain road maintenance and repair reference plan; the performance indicator is one or more of the state parameters; The reward value function is as follows: In the formula, w PCI 、w RQI 、w RDI The weight coefficients of the road condition index PCI, the ride quality index RQI, and the rutting depth index RDI reward values; Road Condition Index PCI Reward Value As shown below: Where c1, c2, c3 are the coefficients of road damage quantity, road damage distribution and main types of road damage; DR t+1 is the road surface damage; S DR_main is the road damage distribution; F DR_type It is the main type of road damage; Ride Quality Index RQI Reward Value As shown below: Where, d1 and d2 are the road surface roughness condition coefficient and the roughness extreme coefficient; IRI t+1 is the road surface smoothness condition; IRI_max is the optimal road surface smoothness; Rutting Depth Index (RDI) Bonus Value As shown below: Where, e1 and e2 are the rutting coefficient and the extreme coefficient of rutting depth; RD t+1 is the rutting depth; RD_max is the maximum rutting depth.

2. The intelligent decision-making system for road maintenance and repair according to claim 1 is characterized by: The status parameters of the road to be maintained and repaired include road age, road type, pavement structure type, surface composition, traffic conditions, road grade, regional zoning, damage rate DR / pavement condition index PCI, pavement average international roughness index IRI / riding quality index RQI, pavement structure representative deflection DEF / pavement structure strength index PSSI, average maximum rutting depth RD / rutting depth index RDI, pavement skid resistance index SRI, first major damage type, and second major damage type.

3. The intelligent decision-making system for road maintenance and repair according to claim 1 is characterized by: The reference plan for road maintenance and repair includes repair plan, daily maintenance plan and preventive maintenance plan.

4. The intelligent decision-making system for road maintenance and repair according to claim 1 is characterized by: Preventive maintenance plan M1 = {daily inspection M1-1; daily maintenance M1-2; daily repair M1-3}; daily maintenance plan M2 = {sealing layer type M2-1; functional cover M2-2; pre-curing combination M2-3}; repair plan M3 = {milling and covering M3-1; structural reinforcement M3-2; surface layer renovation M3-3; base layer renovation M3-4; roadbed treatment M3-5}.

5. The intelligent decision-making system for road maintenance and repair according to claim 1 is characterized by: The maintenance and repair reference factors are determined by expert knowledge; cluster factor analysis is used to determine the effective classification fields, and the scoring standards of the maintenance and repair reference factors are determined by expert knowledge and industry standards.

6. The intelligent decision-making system for road maintenance and repair according to claim 5 is characterized by: The maintenance reference factors include the materials used to implement the road maintenance plan and the corresponding priority, the grade of the road to be maintained, the cost, the city to which it belongs, and the traffic grade.

7. The intelligent decision-making system for road maintenance and repair according to claim 1 is characterized by: The road maintenance and repair plan analysis module evaluates the road maintenance and repair plans in the road maintenance and repair plan set by using the performer-critic model to calculate the reward value R of each road maintenance and repair plan. t+1 , according to the reward value R t+1 Sort the road maintenance and repair plans in descending order to determine their priorities.

Images