Work order dispatching and path planning method based on multi-factor intelligent matching

Abstract

The application discloses a work order dispatching and path planning method based on multi-factor intelligent matching, and particularly relates to the field of work order dispatching and path planning, and comprises grid division, priority sorting, data collection, index calculation and path planning. The application realizes elastic expansion and contraction of geographical units through dynamic grid reconstruction, accurately matches the space-time distribution characteristics of work orders, breaks through the rigid constraints of resource scheduling, innovates a three-dimensional evaluation system integrating spatial topology, dynamic efficiency and geometric complexity, dynamically allocates time period weights to balance global optimization and local adaptability, deeply integrates geometric feature quantization technology, analyzes and predicts path risks through fractal dimension and curvature energy, significantly improves operation safety and equipment endurance, and forms a multi-target collaborative intelligent decision-making closed loop.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of work order dispatching and path planning, and more particularly to a work order dispatching and path planning method based on multi-factor intelligent matching. BACKGROUND

[0002] In the field of intelligent work order scheduling and path planning, traditional methods mainly rely on basic geographic information systems and static road network data for decision-making, and the core goal is to reduce operation and maintenance costs by optimizing mobile paths. The current industry generally adopts a fixed grid division mechanism to divide the service area into equal-sized geographic units, and after manually labeling the work order location, the Dijkstra algorithm is called to calculate the shortest path. The existing technology operation process includes four stages: manually binding geographic labels after receiving work orders, a static path planning engine generates candidate routes, filters schemes based on historical experience values, and finally dispatches execution instructions to devices.

[0003] The existing technology has three key defects: first, static grid division cannot perceive the spatiotemporal distribution characteristics of work order density, and high-density areas are prone to grid resource overload while low-density areas have idle resources; second, the evaluation model only uses straight-line distance or estimated time as a single indicator, ignoring the dynamic fluctuations in traffic conditions, device coordination efficiency, and the coupling relationship between path geometric characteristics; third, the path complexity lacks quantitative standards, and traditional methods cannot identify hidden risk factors such as high-frequency turns and large-angle curves that affect job stability. SUMMARY

[0004] To overcome the above-mentioned defects of the prior art, the present application provides a work order dispatching and path planning method based on multi-factor intelligent matching, which solves the three defects of static grid division rigidity, single-dimensional evaluation ignoring multi-factor interaction, and lack of quantitative standards for path complexity, leading to resource mismatch and execution deviation problems.

[0005] To achieve the above-mentioned purpose, the present application provides the following technical scheme: a work order dispatching and path planning method based on multi-factor intelligent matching, comprising:

[0006] S1: Grid division: based on GIS urban construction coordinates and device positioning data, automatically associate work orders to a preset geographic grid, including geographic level division, dynamic density adjustment, boundary anchor point definition, and device binding;

[0007] S2: Priority sorting: establish a three-level sorting rule to assign weights to work orders, and perform dynamic weight superposition combined with dependency relationship markers, and calculate the top 8 according to the weight total score formula for path planning;

[0008] S3: Data collection: build a multi-source heterogeneous data fusion system, covering spatial topology data, dynamic performance data and road network characteristic data, realize three-dimensional modeling of traffic dynamics, network structure and equipment state through multi-dimensional spatio-temporal data synchronization mechanism;

[0009] S4: Index calculation: establish mathematical model according to the data collected in S3 to calculate spatial topology comprehensive index, dynamic performance index and path complexity index;

[0010] S5: Path planning: establish comprehensive scoring formula according to the three indexes calculated in S4, adjust the comprehensive scoring formula by using dynamic weight, and finally set path planning logic according to the comprehensive score.

[0011] Preferably, the geographical level division takes administrative division as the basic unit, superimposes GIS urban construction coordinate grid in the administrative grid, specifically 500m*500m square grid, each grid corresponds to a unique ID in the database, records the central point longitude and latitude and boundary coordinate set.

[0012] Preferably, the dynamic density adjustment adjusts the grid dynamically according to the work order density threshold, determines the area with >50 work orders / month as high-density area, automatically reduces the grid to 200m*200m, determines the area with <10 work orders / month as low-density area, and combines the grid into 1km*1km large grid, recalculates the grid size based on the work order heat map every month, and automatically updates the boundary through GIS system.

[0013] Preferably, the boundary anchor point definition takes road boundary and natural boundary as the physical boundary of the grid, marks the inflection point through longitude and latitude coordinates, and stores it in the database for real-time matching, wherein the road boundary is bounded by the centerline of the main road, the river in the natural boundary is divided by the centerline of the river, and the mountain is taken as the reference of the contour line. When the work order position is less than 5 meters away from the boundary, the overflow mechanism is triggered, and the work order is preferentially assigned to the adjacent grid with less than 50% of the current number of work orders.

[0014] Preferably, the device binding adopts one-to-one code pre-association rule, each device is bound with coordinates when installed, the system pre-calculates the grid ID, and when the work order is generated, the grid is automatically matched according to the device coordinates, and the device-free work order is distributed through LBS positioning.

[0015] Preferably, in the third-level sorting rule, the first level is the urgency, specifically, 1h single>today single>24h single, indicating that 1 hour processing is required>processing is required on the same day>processing within 24 hours, the weight is 10 when the urgency is 1h single, the weight is 5 when the urgency is today single, and the weight is 1 when the urgency is 24h single; the second level is the work order type, including fault repair and preventive maintenance, the weight is 3 when the work order type is fault repair, and the weight is 1 when the work order type is preventive maintenance; the third level is the dynamic time addition, which is used for 1h single, and the priority is increased when the remaining time is less than 30 minutes, the weight is +2 when the remaining time is less than or equal to 30 minutes, the weight is +5 when the remaining time is less than or equal to 15 minutes, and the addition is added when the stacking time is less than or equal to 30 minutes, the weight is +7; the dependency relationship label is specifically represented as: when the target work order is not completed, it will block other work orders, the target work order is marked as a critical path, and the weight of the critical path work order is multiplied by 1.5.

[0016] Preferably, the weight total score calculation formula is specifically represented as: total score=(urgency weight x 10)+(work order type weight x 5)+(timeout addition x 3), and the screening logic is: first, arrange in descending order of total score, when the scores are the same, arrange in ascending order of receiving time, and finally take the top 8 to enter the path planning.

[0017] Preferably, the space topology data includes grid transition cost, dynamic cross-grid distance, non-Euclidean geometry correction coefficient, and device coordination distance; the dynamic efficiency data includes time window conflict coefficient, processing time length elasticity, traffic flow pulse coefficient, and dependency relationship waiting time; the road network characteristic data includes topology transition index, path fractal dimension, curvature energy integral, and self-similarity fluctuation coefficient.

[0018] Preferably, the grid transition cost reads the grid density label through the space database, calls the map API to obtain the straight line distance d base between the grid center points, and then calculates the penalty value according to C hc =1.2d base or C base =0.8d hc ; C hc represents the grid transition cost, C base =1.2d hc is the high-density grid transition cost, and C base =0.8d s is the low-density grid transition cost; the dynamic cross-grid distance is obtained in real time by the map API according to the actual path distance d s between the starting point (x e , y e ) and the ending point (x api , y t ), and combined with the traffic coefficient f dyn , D api =d t)compute, D dyn denotes dynamic cross-grid distance, f t ∈[0, 1], x s denotes longitude coordinate of starting point, y s denotes latitude coordinate of starting point, x e denotes longitude coordinate of ending point, y e denotes latitude coordinate of ending point, x s , y s , x e and y e Specifically, the decimal system under the WGS84 coordinate system is used; the non-Euclidean geometry correction coefficient is calculated by calculating the Euclidean distance , and comparing the map API road network distance dr to get λ ng = d e / d r , λ ng denotes the non-Euclidean geometry correction coefficient; the device coordination distance is calculated by constructing the minimum spanning tree on the device coordinate set P={(x1, y1),..., (x n , y n )} and calculating the total length D co =∑∥e k ∥, x n denotes the horizontal coordinate of the device in the two-dimensional plane coordinate system, y n denotes the vertical coordinate of the device in the two-dimensional plane coordinate system, D co denotes the device coordination distance, e k is the edge in the minimum spanning tree.

[0019] Preferably, the time window conflict coefficient α t is obtained by calculating the overlap length ratio of the current work order window [s c , e c ] and the historical window set W, and is specifically represented as: α t =∑max(0, min(e c , e i )−max(s c , s i )) / (e c −s c ), e c denotes the end time of the current work order, e i denotes the end time of the scheduled work order, s c denotes the start time of the current work order, s i denotes the start time of the scheduled work order; the processing time length elasticity σ d is obtained by calculating the standard deviation after extracting the same type of work order historical time T={t1,..., t n} by SQL, and is specifically represented as: , t k represent the processing duration of the kth historical work order, μ t represent the sample mean of historical work order processing duration; traffic flow pulse coefficient β f through real-time flow q now and the reference flow q base , β f =∣q now −q base ∣ / q base ; dependency waiting time T wait T wait =max(c j )−t current , c j represent the expected completion time of task j, t current represent the system clock, j∈CP.

[0020] Preferably, the topological migration index I td The number of times the direction changes more than 45 degrees and the total distance ratio are calculated after compressing the GPS trajectory using the Douglas-Peucker algorithm , Δθ represents the change in the direction angle of adjacent line segments, δ is an indicator function, L total represents the total length of the path, N seg represents the total number of trajectory line segments; path fractal dimension D f The slope of logN(r) is calculated by improving the box counting method , r1, r2 represent different grid lengths, N(r) represents the number of grids required to cover the path; curvature energy integral E c After generating the curvature function κ(s) using a third-order spline interpolation, Simpson's integral is used to calculate it κ(s) represents the curvature value of the path at arc length position s, L is the total length of the path, Δs is the integral step size, κ 2i−2 is the starting point curvature of the ith segment, κ 2i−1 is the midpoint curvature, and κ 2i is the end point curvature; self-similar fluctuation coefficient γ s The path coordinate sequence is decomposed into 3 layers by Haar wavelet, and the adjacent scale energy ratio is calculated E l represents the energy of the lth layer of detail coefficients.

[0021] Preferably, the S3 is implemented by using the PostGIS spatial engine to mark the grid density, updating the traffic data every 3 minutes through the map API, collecting the coordinates by the GPS Beidou terminal at a frequency of 1 Hz, performing multi-objective path optimization by the OR-Tools library, processing real-time data streams by Apache Flink, calculating geometric features by the CGAL library, allocating task resources by the Hungarian algorithm, monitoring the abnormality by using a 5-minute sliding window, and calculating the curvature error to be less than 0.1 rad / km.

[0022] Preferably, the spatial topology comprehensive index is specifically represented as: , wherein S represents the spatial topology comprehensive index, and η represents a path efficiency weight coefficient.

[0023] Preferably, the dynamic efficiency index is specifically represented as: , wherein T represents the dynamic efficiency index, and ϕ represents a smoothing constant.

[0024] Preferably, the path complexity index is specifically represented as: , wherein P represents the path complexity index, k1 and k2 are power law adjustment coefficients, and ψ represents a curvature energy smoothing term.

[0025] Preferably, the comprehensive score formula is specifically represented as:

[0026] , wherein Z(t) represents the comprehensive score, w s (t) represents the spatial topology weight, w t (t) represents the dynamic efficiency weight, w p (t) represents the path complexity weight, γ represents an emergency work order gain coefficient, τ represents a nonlinear decay factor, t0 represents a work order generation time, e is a constant, t is time, and the path planning logic is: arranged in descending order of Z(t), when Z(t) is the same, arranged in descending order of S, when S is the same, arranged in ascending order of work order receiving time.

[0027] Preferably, w s (t), w t (t) and w p (t) are dynamic weights, [ws, wt, wp] are weight distribution as [0.3, 0.6, 0.1] in the early peak period 7:00-9:00, weight distribution as [0.4, 0.3, 0.3] in the flat peak period 10:00-16:00, weight distribution as [0.2, 0.5, 0.3] in the late peak period 17:00-19:00, and weight distribution as [0.5, 0.2, 0.3] in the night period 20:00-6:00.

[0028] Technical effects and advantages of the present application:

[0029] The application realizes the elastic expansion of geographic units through a dynamic grid reconstruction technology, automatically adjusts the grid granularity and boundary anchor points based on the time and space distribution characteristics of work orders, and accurately matches resource scheduling with business demand, which breaks through the rigid constraints of fixed grids, maintains service continuity through intelligent overflow mechanism when sudden work orders surge, reduces computing overhead through grid merging during low load period, and realizes the collaborative optimization of infrastructure utilization and response speed;

[0030] The application innovatively constructs a three-dimensional evaluation system of space-time path, integrates spatial topological efficiency, dynamic traffic risk and path geometric complexity into a unified calculation framework, automatically strengthens the time efficiency weight to avoid congestion during peak traffic period, improves the spatial coordination weight to optimize resource allocation in equipment-intensive operation area, and enhances the path stability weight to ensure execution reliability in complex road network section, forming a multi-objective adaptive intelligent decision-making capability;

[0031] The application deeply integrates geometric feature analysis technology, quantifies path detour degree through fractal dimension, and evaluates driving smoothness through curvature energy integral, which first converts road network geometric characteristics into calculable engineering parameters, so that the planning engine can actively avoid high-energy-consumption road sections such as high-frequency turning, steep slope and curve, significantly improve equipment endurance and operation safety, and break through the experience dependence of traditional path planning on geometric risk. BRIEF DESCRIPTION OF DRAWINGS

[0032] Figure 1 It is a schematic diagram of the overall structure of the application.

[0033] Figure 2 It is a schematic diagram of the complete embodiment structure of the application.

[0034] Figure 3 It is a schematic diagram of the medium topology connection of the application.

[0035] Figure 4 It is a schematic diagram of the module structure of the application. DETAILED DESCRIPTION

[0036] The technical solutions in the embodiments of the application will be described clearly and completely below with reference to the drawings in the embodiments of the application. Obviously, the described embodiments are only part of the embodiments of the application, not all the embodiments. Based on the embodiments in the application, all other embodiments obtained by those skilled in the art without creative labor fall within the scope of the application.

[0037] Reference Figures 1-3 The multi-factor intelligent matching-based work order dispatching and path planning method shown includes:

[0038] S1: Grid division: Based on GIS urban construction coordinates and device positioning data, automatically associate work orders to the preset geographical grid, including geographical level division, dynamic density adjustment, boundary anchor point definition, and device binding.

[0039] The geographical level division takes administrative division as the basic unit, superimposes GIS urban construction coordinate grid in the administrative grid, specifically 500m×500m square grid, each grid corresponds to a unique ID in the database, records the center point longitude and latitude and boundary coordinate set;

[0040] The dynamic density adjustment dynamically adjusts the grid according to the work order density threshold, determines the area with >50 work orders / month as high-density area, automatically reduces the grid to 200m×200m, determines the area with <10 work orders / month as low-density area, and merges the grid into 1km×1km large grid. The grid size is recalculated based on the work order heat map every month, and the boundary is automatically updated by the GIS system;

[0041] The boundary anchor point definition takes road boundary and natural boundary as the physical boundary of the grid, marks the inflection point through longitude and latitude coordinates, and stores it in the database for real-time matching. Among them, the road boundary is bounded by the centerline of the main road, the river in the natural boundary is divided by the centerline of the river, and the mountain is taken as the reference of the contour line. When the work order location distance is <5 meters from the boundary, the overflow mechanism is triggered, and the work order is preferentially assigned to the adjacent grid with less than 50% of the current number of work orders;

[0042] The device binding adopts the rule of one thing one code pre-association. Each device is bound with coordinates when installed, and the system pre-calculates the grid ID it belongs to. When the work order is generated, the grid is automatically matched according to the device coordinates. If there is no device work order, it is distributed through LBS positioning.

[0043] S2: Priority sorting: Establish a three-level sorting rule to assign weights to work orders, and dynamically superimpose weights combined with dependency relationship markers. According to the weight total score calculation formula, the top 8 are screened for path planning.

[0044] The third-level sorting rule is: the first level is the emergency degree, specifically 1h single> single on the same day> 24h single, indicating that it needs to be processed within 1 hour> needs to be processed on the same day> needs to be processed within 24 hours, the weight is 10 when the emergency degree is 1h single, the weight is 5 when the emergency degree is single on the same day, and the weight is 1 when the emergency degree is 24h single; The second level is the work order type, including fault repair and preventive maintenance, the weight is 3 when the work order type is fault repair, and the weight is 1 when the work order type is preventive maintenance; The third level is dynamic time addition, which is aimed at 1h single, and the priority is improved when the remaining time is less than 30 minutes, the weight is +2 when the remaining time is less than or equal to 30 minutes, the weight is +5 when the remaining time is less than or equal to 15 minutes, and the weight is +7 when the superposition time is less than or equal to 30 minutes; The dependency relationship label is specifically represented as: when the target work order is not completed, it will block other work orders, and the target work order is marked as a critical path, and the weight of the critical path work order is multiplied by 1.5.

[0045] The weight total score calculation formula is specifically represented as: total score=(emergency degree weight*10)+(work order type weight*5)+(timeout addition*3), and the screening logic is: first arrange in descending order according to the total score, when the scores are the same, arrange in ascending order according to the receiving time, and finally take the top 8 to enter the path planning.

[0046] Example:

[0047] The emergency degree of work order A is 1h single, the type is fault repair, the remaining time is 20 minutes, and the receiving time is 09:00:01. The calculation process is (10*10)+(3*5)+(2*3)=100+15+6, and the total score is 121;

[0048] The emergency degree of work order B is 1h single, the type is preventive maintenance, the remaining time is 50 minutes, and the receiving time is 08:59:59. The calculation process is (10*10)+(1*5)+(0*3)=100+5+0, and the total score is 105;

[0049] The emergency degree of work order C is single on the same day, the type is fault repair, and the receiving time is 09:00:00. The calculation process is (5*10)+(3*5)+(0*3)=50+15+0, and the total score is 65;

[0050] Result: work order A (121 points)> work order B (105 points)> work order C (65 points).

[0051] S3: data acquisition: build a multi-source heterogeneous data fusion system, covering spatial topology data, dynamic efficiency data and road network characteristic data, and realize three-dimensional modeling of traffic dynamics, network structure and equipment state through multi-dimensional space-time data synchronization mechanism.

[0052] The space topology data includes grid transition cost, dynamic cross-grid distance, non-Euclidean geometry correction coefficient and device coordination distance; the dynamic performance data includes time window conflict coefficient, processing time length elasticity, traffic flow pulse coefficient and dependent relationship waiting time length; and the road network characteristic data includes topology detour index, path fractal dimension, curvature energy integral and self-similarity fluctuation coefficient.

[0053] In the collection of the space topology data, the grid transition cost is obtained according to the grid density label by calling the map API to obtain the straight line distance between the grid center points, the transition cost of the high-density grid is 1.2 times the straight line distance, and the transition cost of the low-density grid is 0.8 times the straight line distance; the dynamic cross-grid distance is obtained by calling the map API to obtain the actual path distance between the starting point and the ending point in real time, and the actual path distance is combined with the traffic coefficient to obtain; the non-Euclidean geometry correction coefficient is obtained by comparing the Euclidean distance between the starting point and the ending point with the actual road network distance provided by the map API, and the non-Euclidean geometry correction coefficient is equal to the Euclidean distance divided by the actual road network distance; and the device coordination distance is obtained by constructing a minimum spanning tree for the device coordinate set by the Prim algorithm, and calculating the sum of the lengths of all edges of the tree.

[0054] The grid transition cost is obtained by reading the grid density label from the space database and calling the map API to obtain the straight line distance d base between the grid center points, and then calculating the penalty value C hc =1.2d base or 0.8d base , wherein C hc represents the grid transition cost, C hc =1.2d base is the high-density grid transition cost, and C hc =0.8d base is the low-density grid transition cost; the dynamic cross-grid distance is obtained by calling the map API to obtain the actual path distance d s between the starting point (x s , y e ) and the ending point (x e , y api ), and combining the traffic coefficient f t to calculate D dyn =d api (1+f t ), wherein D dyn represents the dynamic cross-grid distance, f t ∈[0, 1], x s represents the longitude coordinate of the starting point, y s represents the latitude coordinate of the starting point, x e represents the longitude coordinate of the ending point, and y e represents the latitude coordinate of the ending point. s s ​, x e , and y e Specifically, the decimal system under the WGS84 coordinate system is used; the non-Euclidean geometry correction coefficient is calculated by calculating the Euclidean distance , and the map API road network distance dr is compared to obtain λ ng = d e / d r , λ ng represents the non-Euclidean geometry correction coefficient; the device coordination distance is calculated by constructing a minimum spanning tree on the device coordinate set P={(x1, y1),..., (x n , y n )} through the Prim algorithm, and calculating the total length D co =∑∥e k ∥, x n represents the horizontal coordinate of the device in the two-dimensional plane coordinate system, y n represents the vertical coordinate of the device in the two-dimensional plane coordinate system, D co represents the device coordination distance, and e k is an edge in the minimum spanning tree.

[0055] In the collection of the dynamic performance data, the time window conflict coefficient is calculated by calculating the overlap length ratio of the current work order time window and the historical time window set, specifically, the maximum end time of all overlapping periods minus the sum of the minimum start time, divided by the total length of the current work order window; the processing time elasticity is calculated by extracting the historical processing time data of the same type of work order, and calculating the standard deviation of these lengths; the traffic flow pulse coefficient is the absolute difference between the real-time flow and the reference flow divided by the reference flow; the dependence relationship waiting time is calculated by modeling the key path through Petri net, and calculating the difference between the maximum value of the expected completion time of all tasks on the key path and the current system time.

[0056] The time window conflict coefficient α t is calculated by calculating the overlap length ratio of the current work order window [s c , e c ] and the historical window set W, specifically represented as: α t =∑max(0, min(e c , e i )−max(s c , s i )) / (e c −s c ), e c represents the end time of the current work order, e i represents the end time of the scheduled work order, s c represents the start time of the current work order, and s i represents the start time of the scheduled work order; the processing time elasticity σ dThe same kind of work order history duration T = {t1,..., tk} is extracted by SQL n The standard deviation is calculated after the history, which is specifically represented as: , t k The processing duration of the kth historical work order is represented by μ t The sample mean of the historical work order processing duration is represented by β f The traffic flow pulse coefficient is calculated by comparing the real-time flow q now With the reference flow q base , and β f =∣q now −q base ∣ / q base ; The waiting time T wait of the dependency relationship is calculated by modeling the critical path CP of the Petri net, and T wait =max(c j )−t current , c j represents the expected completion time of task j, t current represents the system clock, and j∈CP.

[0057] In the collection of road network characteristic data, the topological tortuosity index is compressed by the Douglas-Peucker algorithm, the number of times the direction angle changes by more than 45 degrees is counted, and divided by the total path length to obtain; the path fractal dimension is determined by using a grid of different side lengths to cover the path, calculating the slope of the logarithm of the required grid number and the grid side length; the curvature energy integral is generated by third-order spline interpolation to generate the curvature function of the path, and Simpson's integral method is used to integrate the curvature square along the path length to obtain; the self-similarity fluctuation coefficient is obtained by three-layer Haar wavelet decomposition of the path coordinate sequence, and the energy ratio of adjacent scale detail coefficients is calculated.

[0058] The topological tortuosity index I td is compressed by the Douglas-Peucker algorithm, and the number of times the direction angle changes by more than 45 degrees is counted and divided by the total mileage , Δθ represents the change in the direction angle of adjacent line segments, δ is an indicator function, L total represents the total length of the path, and N seg represents the total number of trajectory line segments; the path fractal dimension D f is calculated by improving the box counting method to calculate the slope of logN(r) , r1, r2 represent different grid side lengths, and N(r) represents the number of grids required to cover the path; the curvature energy integral E c is calculated by Simpson's integral after generating the curvature function κ(s) by third-order spline interpolation , κ(s) represents the curvature value of the path at arc length position s, L is the total length of the path, and Δs is the integral step size, κ 2i−2The curvature of the starting point of the i-th segment, κ 2i−1 The curvature of the midpoint, κ 2i The curvature of the ending point; the self-similarity fluctuation coefficient γ s The adjacent scale energy ratio is calculated after the path coordinate sequence is decomposed into 3 layers by the Haar wavelet , E l The energy of the l-th layer detail coefficient is represented.

[0059] The S3 is implemented by using the PostGIS spatial engine to mark the grid density, updating the traffic data every 3 minutes through the map API, collecting the coordinates at a frequency of 1 Hz by the GPS Beidou terminal, performing multi-objective path optimization by the OR-Tools library, processing real-time data streams by Apache Flink, calculating geometric features by the CGAL library, allocating task resources by the Hungarian algorithm, monitoring exceptions by using a 5-minute sliding window, and the curvature calculation error is less than 0.1 rad / km.

[0060] S4: Index calculation: according to the data collected in S3, a mathematical model is established to calculate the spatial topology comprehensive index, the dynamic efficiency index, and the path complexity index.

[0061] The spatial topology comprehensive index is the product of the grid transition cost and the dynamic cross-grid distance divided by the non-Euclidean correction coefficient, plus the square root of the device coordination distance multiplied by the path efficiency weight coefficient.

[0062] The spatial topology comprehensive index is specifically represented as: S represents the spatial topology comprehensive index, and η represents the path efficiency weight coefficient.

[0063] The spatial topology comprehensive index realizes progressive analysis through the balance of spatial efficiency and resource coordination. First, the grid transition cost C hc is multiplied by the dynamic distance D dyn , reflecting the actual moving cost affected by the density penalty and the road condition, divided by the non-Euclidean correction coefficient λ ng to eliminate the straight-line distance error, and the device coordination distance D co is taken as an independent term to reflect the cluster operation efficiency, and the contribution proportion of the two is adjusted by the weight coefficient η. The determination of the weight coefficient η uses historical work order data regression analysis: 1000 groups of completed work orders are selected, and the actual fuel consumption / time is taken as the target variable. The weight coefficient η is fitted by the least square method to be 0.68.

[0064] The dynamic efficiency index is the product of the time window conflict coefficient and the processing time elasticity, plus the sum of the traffic flow pulse coefficient and the logarithm of the dependence relationship waiting time, multiplied by the smoothing constant.

[0065] The dynamic efficiency index is specifically represented as: , T represents a dynamic performance index, and φ represents a smoothing constant.

[0066] The dynamic performance index is used to depict the composite risk in the time dimension, the first term α t σ d represents the coupling effect of the time window conflict and the processing time uncertainty, embodies the product relationship of the plan failure rate, and the second term introduces a traffic pulse coefficient β f and the combination of the logarithmic term of the waiting time, wherein ln(T wait + ϕ) non-linearly maps the waiting time to a risk value, ϕ = 0.01 prevents mathematical anomalies from occurring at zero waiting time, the formula structure verification adopts Monte Carlo simulation: 100,000 groups of parameter combinations are generated, the actual delay data are compared, it is determined that the error is reduced by more than 30% in the logarithmic transformation than in the linear form, the dimension of the coefficients is unified through Z-score standardization, and the scale difference of the variables is eliminated.

[0067] The path complexity index is a cubic root of a geometric mean of a topological detour index and a path fractal dimension, plus a product of a curvature energy integral and a self-similarity wave coefficient, multiplied by a power law adjustment coefficient, and added with a curvature energy smoothing term.

[0068] The path complexity index is specifically represented as: , P represents the path complexity index, k1 and k2 are power law adjustment coefficients, and ψ represents the curvature energy smoothing term.

[0069] The path complexity index quantifies the path complexity through multi-scale fusion of geometric features, adopts geometric mean to eliminate the dimension difference, and compresses the value range to a reasonable range through cubic root, wherein the topological detour index I td and the fractal dimension D f The power law coefficients k1 and k2 are introduced, and the weights are determined through factor analysis: six typical paths of urban roads, expressways and the like are selected, principal component analysis is used to obtain k2 = 1.3, which reflects that the contribution degree of the fractal dimension is 1.86 times that of the detour index, and the curvature energy integral E c is increased by ψ = 0.1 to prevent calculation failure of zero curvature road segments, and the value is set according to 2 times the sensor noise level.

[0070] S5: Path planning: a comprehensive score formula is established according to the three indexes calculated in S4, a dynamic weight is used to adjust the comprehensive score formula, and finally a path planning logic is set according to the comprehensive score.

[0071] The comprehensive score formula is obtained by multiplying the space topology weight, the dynamic performance weight and the path complexity weight by the corresponding index respectively, multiplying by the combination of the emergency work order gain coefficient and the index attenuation factor, and finally performing normalization processing.

[0072] The comprehensive score formula is specifically represented as:

[0073] Z(t) = wS S + wT T + wP P, Z(t) represents the comprehensive score, wS s (t) represents the spatial topology weight, wT t (t) represents the dynamic performance weight, wP p (t) represents the path complexity weight, γ represents the emergency work order gain coefficient, τ represents the nonlinear decay factor, t0 represents the work order generation time, e is a constant, t is time, and the path planning logic is: arranged in descending order of Z(t), when Z(t) is the same, arranged in descending order of S, when S is the same, arranged in ascending order of work order receiving time.

[0074] The wS s (t), wT t (t) and wP p (t) are dynamic weights, [ws, wt, wp] are weight distribution [0.3, 0.6, 0.1] in the early peak period 7:00-9:00, weight distribution [0.4, 0.3, 0.3] in the flat peak period 10:00-16:00, weight distribution [0.2, 0.5, 0.3] in the late peak period 17:00-19:00, and weight distribution [0.5, 0.2, 0.3] in the night period 20:00-6:00.

[0075] The comprehensive score formula allocates basic weights according to the characteristics of the period, focuses on time performance in the early peak period and focuses on spatial efficiency at night, the weight value is determined through historical work order completion rate regression analysis, the denominator is normalized to eliminate the influence of weight absolute value, and the total sum of weights in different periods is consistent with the scale of the score, the exponential term 1 / (1+γe −τ(t−t0) ) realizes work order retention penalty, automatically promotes priority over time, is calibrated according to measured data, when the coefficient γ=0.2, the score is improved by 19% when the work order is retained for 2 hours, when Z(t) is the same, the original space index S is sorted to ensure repeatability.

[0076] Example:

[0077] Period: Early peak 8:30 (applicable weight [0.3, 0.6, 0.1])

[0078] Work order A: S=82, T=75, P=68, generation time: 8:00;

[0079] Work order B: S=78, T=88, P=72, generation time: 8:15;

[0080] Work order C: S=95, T=65, P=81, generation time: 8:20;

[0081] Calculate the comprehensive score:

[0082] ;

[0083] ;

[0084] ;

[0085] Result: B (91.0) → A (87.3) → C (80.7).

[0086] Reference Figure 4 , the application is realized through the cooperation of the geographic grid management module, the work order scheduling core module, the data collection adaptation module, the real-time calculation engine module and the resource coordination module, specifically:

[0087] The geographic grid management module: based on the GIS urban construction coordinate system, realize dynamic grid division, maintain grid ID, boundary coordinates and density label through PostGIS spatial engine, trigger dynamic reconstruction according to work order heat map every month;

[0088] The work order scheduling core module: integrate three-level sorting rules and multi-objective path planning algorithm, receive real-time coordinate input of LBS positioning system, generate task sequence containing time window constraint;

[0089] The data collection adaptation module: interface with GPS / Beidou terminal, map API and other multi-source data, complete spatial topology relationship standardization processing;

[0090] The real-time calculation engine module: uses Apache Flink to process data stream, combines OR-Tools and CGAL library to realize dynamic distance calculation and geometric feature analysis;

[0091] Resource coordination module: task allocation through Hungarian algorithm, based on the device coordinate set generated by Prim algorithm to establish coordination network, realize work order-device optimal matching.

[0092] The geographic grid management module provides the basic space reference system for each module, the data collection adaptation module pushes real-time data to the real-time calculation engine for index processing, the work order scheduling core module comprehensively receives geographic grid data and real-time calculation index, generates work order distribution scheme and triggers the resource coordination module to execute.

[0093] The application firstly carries out dynamic geographical grid division, adjusts grid granularity according to real-time work order density and sets boundary anchor points, realizes intelligent matching of equipment and grid through pre-binding rules; then starts a three-level priority sorting mechanism, comprehensively weights scores according to emergency degree, work order type and remaining time, filters high-priority work orders into a path queue; then synchronously collects spatial topology data, real-time traffic efficiency indexes and path geometric characteristics, fuses multi-source data to construct a calculation model; generates spatial efficiency index, time risk index and path complexity index based on preset algorithms respectively; finally combines time period characteristics to dynamically distribute weight coefficients, calculates work order execution priority through a nonlinear scoring model, drives a path planning engine to generate an optimal route, realizes grid resource scheduling and traffic situation self-adaptive intelligent dispatching, and adopts a streaming computing framework to guarantee data real-time performance throughout the whole process, and balances multi-dimensional targets of efficiency, cost and path reliability through a dynamic weight mechanism.

Claims

1. A method for work order dispatching and path planning based on multi-factor intelligent matching, characterized in that, Comprise: S1: mesh division: based on GIS urban construction coordinates and equipment positioning data, automatically associate the work order to the preset geographical grid, including geographical hierarchical division, dynamic density adjustment, boundary anchor point definition and equipment binding; S2: priority sorting: establish three-level sorting rules to assign weights to work orders, and combine dependency relationship markers for dynamic weight superposition, and calculate the top 8 according to the weight total score formula to plan the path; S3: data collection: build a multi-source heterogeneous data fusion system, covering spatial topology data, dynamic efficiency data and road network characteristic data, and realize three-dimensional modeling of traffic dynamics, network structure and equipment state through multi-dimensional spatiotemporal data synchronization mechanism; The spatial topology data includes grid transition cost, dynamic cross-grid distance, non-Euclidean geometry correction coefficient and equipment cooperation distance; dynamic efficiency data includes time window conflict coefficient, processing time elasticity, traffic flow pulse coefficient and dependency relationship waiting time; road network characteristic data includes topological transition index, path fractal dimension, curvature energy integral and self-similarity fluctuation coefficient; In the collection of the spatial topology data, the grid transition cost is obtained by calling the map API to obtain the straight line distance between the grid center points according to the grid density label, the transition cost of high-density grid is 1.2 times the straight line distance, and the transition cost of low-density grid is 0.8 times the straight line distance; The dynamic cross-grid distance is obtained by real-time acquisition of the actual path distance between the starting point and the ending point through the map API, and the actual path distance is combined with the traffic coefficient to calculate; The non-Euclidean geometry correction coefficient is obtained by comparing the Euclidean distance between the starting point and the ending point with the actual road network distance provided by the map API, and the non-Euclidean geometry correction coefficient is equal to the Euclidean distance divided by the actual road network distance; The equipment cooperation distance is obtained by constructing the minimum spanning tree of the equipment coordinate set through Prim algorithm, and calculating the sum of the lengths of all edges of the tree; The grid transition cost reads the grid density label through a spatial database, calls a map API to obtain the straight-line distance d of the grid center point base , then press C hc =1.2d base or 0.8d base , and calculates the penalty value C hc , wherein C hc represents the grid transition cost, C base =1.2d hc is the high-density grid transition cost, and C base =0.8d s is the low-density grid transition cost; the dynamic cross-grid distance is obtained in real time by a map API according to the actual path distance d s of the starting point (x e , y e ) and the ending point (x api , y t ), and is calculated according to D dyn =d api (1+f t ), wherein D dyn represents the dynamic cross-grid distance, f t ∈[0, 1], x s represents the longitude coordinate of the starting point, y s represents the latitude coordinate of the starting point, x e represents the longitude coordinate of the ending point, and y e represents the latitude coordinate of the ending point; x s , y s , x e and y e specifically use the decimal system under the WGS84 coordinate system; the non-Euclidean geometry correction coefficient is obtained by calculating the Euclidean distance between the starting point and the ending point, and is specifically represented as: , and λ ng =d e / d r is obtained by comparing the map API road network distance d r , wherein λ ng represents the non-Euclidean geometry correction coefficient; the device coordination distance is calculated by constructing a minimum spanning tree on a device coordinate set P={(x1, y1),..., (x n , y n )} through a Prim algorithm, and is represented as D co =∑∥e k ∥, wherein x n represents the horizontal coordinate of the device in a two-dimensional plane coordinate system, y n represents the vertical coordinate of the device in the two-dimensional plane coordinate system, D co represents the device coordination distance, and e k is an edge in the minimum spanning tree. In the collection of the dynamic efficiency data, the time window conflict coefficient is obtained by calculating the overlap length ratio of the current work order time window and the historical time window set, which is the sum of the maximum end time of all overlapping periods minus the minimum start time, divided by the total time of the current work order window; The processing time elasticity is obtained by extracting the historical processing time data of similar work orders, and calculating the standard deviation of these time lengths; The traffic flow pulse coefficient is the absolute difference between the real-time flow and the reference flow divided by the reference flow; The dependency relationship waiting time is obtained by modeling the critical path through Petri net, and calculating the difference between the maximum value of the expected completion time of all tasks on the critical path and the current system time; The time window conflict coefficient α t By calculating the current work order window [s] c e c The ratio of the overlap duration between the α and the historical window set W is used to derive the value, specifically expressed as: α t =∑max(0,min(e c e i )-max(s c s i )) / (e c -s c ), e c Indicates the end time of the current work order, e i Indicates the end time of the scheduled work order, s c Indicates the start time of the current work order, s i Indicates the start time of the scheduled work order; processing time flexibility σ d Extract the historical duration T={t1, ...,t} of similar work orders using SQL. n The standard deviation is then calculated and expressed as follows: , t k μ represents the processing time of the k-th historical work order. t The sample mean of historical work order processing time; traffic flow impulse coefficient β f Through real-time traffic q now Compared with the baseline flow rate q base Comparison yields β f =∣q now -q base | / q base Dependency wait time T wait After modeling the critical path CP using Petri nets, calculate T. wait =max(c j )-t current c j Indicates the estimated completion time of task j, t current Represents the system clock, j∈CP; In the collection of the road network characteristic data, the topological detour index is obtained by compressing the GPS trajectory by the Douglas-Peucker algorithm, counting the number of times that the direction angle changes by more than 45 degrees, and dividing by the total path length; the path fractal dimension is determined by covering the path with grids of different lengths, calculating the slope of the logarithm of the number of required grids and the logarithm of the grid length; the curvature energy integral is obtained by generating the curvature function of the path by third-order spline interpolation, and integrating the curvature square along the path length by Simpson's integral method; the self-similarity fluctuation coefficient is obtained by three-level Haar wavelet decomposition of the path coordinate sequence, and calculating the energy ratio of adjacent scale detail coefficients; The topological tortuosity index I td The ratio of the number of direction mutations greater than 45° to the total distance after compressing the GPS trajectory using the Douglas-Peucker algorithm , Δθ represents the change in the direction angle of adjacent line segments, δ is an indicator function, L total represents the total length of the path, N seg represents the total number of trajectory line segments; the path fractal dimension D f The slope of logN(r) is calculated by improving the box counting method , r1, r2 represent different grid lengths, N(r) represents the number of grids required to cover the path; the curvature energy integral E c After generating the curvature function κ(s) using a third-order spline interpolation, Simpson's integral is used to calculate κ(s) represents the curvature value of the path at the arc length position s, L is the total length of the path, Δs is the integral step, κ 2i-2 is the curvature at the starting point of the i-th segment, κ 2i-1 is the midpoint curvature, κ 2i is the curvature at the end point; the self-similarity fluctuation coefficient γ s The adjacent scale energy ratio is calculated by decomposing the path coordinate sequence into 3 layers using the Haar wavelet E l represents the energy of the l-th layer of detail coefficients; S4: index calculation: according to the data collected in S3, a mathematical model is established to calculate the spatial topology comprehensive index, dynamic efficiency index and path complexity index; The spatial topology comprehensive index is obtained by multiplying the grid transition cost and the dynamic cross-grid distance, dividing by the non-Euclidean geometry correction coefficient, and then adding the square root of the device coordination distance multiplied by the path efficiency weight coefficient; The dynamic efficiency index is obtained by multiplying the time window conflict coefficient and the processing time length elasticity, adding the sum of the traffic flow pulse coefficient and the logarithm of the dependence relationship waiting time, and then multiplying by the smoothing constant; The path complexity index is obtained by taking the cubic root of the geometric mean of the topological detour index and the path fractal dimension, adding the product of the curvature energy integral and the self-similarity fluctuation coefficient, multiplying by the power law adjustment coefficient, and adding the curvature energy smoothing term; The path complexity index is specifically represented as: P represents the path complexity index, k1, k2 are power law adjustment coefficients, and ψ represents a curvature energy smoothing term. S5: path planning: according to the three indexes calculated in S4, a comprehensive scoring formula is established, the dynamic weight is adjusted, and finally the path planning logic is set according to the comprehensive score.

2. The multi-factor intelligent matching-based work order distribution and path planning method according to claim 1, characterized in that: The geographical hierarchy division takes administrative division as the basic unit, superimposes a GIS urban construction coordinate grid in the administrative grid, specifically a 500m×500m square grid, each grid corresponds to a unique ID in the database, records the central point longitude and latitude and the boundary coordinate set; The dynamic density adjustment dynamically adjusts the grid according to the work order density threshold, determines the area with >50 work orders / month as a high-density area, automatically reduces the grid to 200m×200m, determines the area with <10 work orders / month as a low-density area, and merges the grid into a 1km×1km large grid, recalculates the grid size based on the work order heat map every month, and automatically updates the boundary through the GIS system; The boundary anchor point definition takes the road boundary and natural boundary as the physical boundary of the grid, marks the inflection points by longitude and latitude coordinates, and stores them in the database for real-time matching, wherein the road boundary is bounded by the center line of the main road, the river is divided by the center line of the river, and the mountain is taken as the reference by the contour line. When the work order position is less than 5 meters from the boundary, the overflow mechanism is triggered, and the work order is preferentially distributed to the adjacent grid with less than 50% of the current number of work orders; The device binding adopts a one-to-one code pre-association rule, each device is bound with coordinates when installed, the system pre-calculates the grid ID, and when a work order is generated, the grid is automatically matched according to the device coordinates. 3.The multi-factor intelligent matching based work order dispatching and path planning method of claim 1, wherein: In the three-level sorting rule, the first level is the emergency degree, specifically 1h single> single on the day> 24h single, indicating that it needs to be processed within 1 hour> needs to be processed on the day> processed within 24 hours, the weight is 10 when the emergency degree is 1h single, the weight is 5 when the emergency degree is single on the day, and the weight is 1 when the emergency degree is 24h single; The second level is the work order type, including fault repair and preventive maintenance, the weight is 3 when the work order type is fault repair, and the weight is 1 when the work order type is preventive maintenance; The third level is dynamic time addition, which is aimed at 1h single, and the priority is improved when the remaining time is less than 30 minutes, the weight is +2 when the remaining time is less than or equal to 30 minutes, the weight is +5 when the remaining time is less than or equal to 15 minutes, and the weight is +7 when the superposition time is less than or equal to 30 minutes; The dependency relationship label is specifically represented as: when the target work order is not completed, it will block other work orders, the target work order is marked as a critical path, and the weight of the critical path work order is multiplied by 1.

5. 4.The multi-factor intelligent matching based work order dispatching and path planning method of claim 1, wherein: The weight total score calculation formula is specifically represented as: total score=(emergency degree weight x 10)+(work order type weight x 5)+(timeout addition x 3), and the screening logic is: first, arrange in descending order according to the total score, when the scores are the same, arrange in ascending order according to the receiving time, and finally take the top 8 to enter the path planning. 5.The multi-factor intelligent matching based work order dispatching and path planning method of claim 1, wherein: The comprehensive score formula is obtained by multiplying the space topology weight, the dynamic efficiency weight and the path complexity weight by the corresponding index, respectively, and then multiplying the combination of the emergency work order gain coefficient and the index attenuation factor, and finally through normalization processing.

Images