Special vehicle double-space-scale route planning method suitable for gridding monitoring
Patent Information
- Application Number
- CN202411480505.2
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-10-23
- Publication Date
- 2025-02-14
AI Technical Summary
Within a limited time or distance budget, how to plan a dedicated fleet route to maximize perceived benefits at the grid level, especially in grid monitoring scenarios, coordinate path planning at the grid level and road network level.
Using the dual-space scale route planning method, the route network-level path is generated according to the grid-level path through step 1, and in the path planning of a single special vehicle in step 2, the grid with the largest marginal gain is greedy to choose as the starting grid. Use the improved A-star algorithm to calculate the paths at the grid level to maximize perceived benefits. For multiple vehicles, the method of serializing the routes of each vehicle first and then iteratively adjusting them to avoid overlapping routes and maximize the overall perceived effect.
Under a given time or mileage constraint, multiple vehicles can achieve differentiated coverage of different monitoring objects, maximize monitoring effectiveness, and solve the problems of strong information bias and low monitoring resource utilization in large-scale inspection tasks.
Smart Images
Figure BDA0005097242940000024 
Figure BDA0005097242940000046 
Figure FDA0005097242930000026
Abstract
Description
Technical Field
[0001] The present invention belongs to vehicle route planning technology, and in particular relates to a dual-space-scale route planning method for a special vehicle suitable for grid monitoring. Background Art
[0002] With the development of smart cities, the widespread use of mobile sensing platforms has provided innovative solutions for urban environmental monitoring. Unlike traditional third-party vehicles (such as taxis and buses), special-purpose vehicles have higher flexibility and controllability, and can flexibly navigate in the road network to perform specific data collection tasks. Therefore, special-purpose vehicles are widely used in performing targeted perception tasks or supplementing the capabilities of other perception vehicles.
[0003] Compared with the traditional vehicle routing problem, the route planning of special vehicles for grid monitoring has a unique dual spatial scale characteristic. Specifically, the perceived utility is not quantified by simply visiting nodes or edges, but measured by covering a specific spatial grid. At the same time, the perceived utility can be accumulated according to the utility function, which brings a new challenge: how to coordinate the path planning at the grid level and the road network level. Although these two processes are carried out at different spatial scales, they are interdependent and must be solved in a coordinated manner. Summary of the invention
[0004] In order to solve the problem of how to plan the route of a dedicated vehicle fleet within a limited time or distance budget to maximize the perception benefit at the grid level, the present invention provides a dual-space-scale route planning method for a dedicated vehicle suitable for grid monitoring.
[0005] A dual-space-scale route planning method for a special vehicle suitable for grid monitoring of the present invention comprises the following steps:
[0006] Step 1: Generate a road network level path based on the grid level path.
[0007] make Represents the node set of the road network, Represents a set of spatial grids in the monitoring area. If a node In space grid In, it is represented by v∈g; with r N Represents the path of the vehicle at the road network level, expressed as a sequence of road network nodes, that is, use Represents the path of the vehicle at the grid level, expressed as a sequence of grids, i.e. The paths at the dedicated fleet network level and the path sets at the grid level are expressed as and
[0008] Quantifying perceived benefits: For a grid Its coverage number is N g , the perception score ξ of the grid is calculated as follows:
[0009] ξ=(N g ) β (1)
[0010] In the formula, β is the parameter of the utility function. The smaller β is, the smaller the marginal utility of multiple sampling is.
[0011] Marginal gain: For a grid Its perceptual weight is w g , when the grid has been covered by N g The marginal gain η brought by the next visit g The calculation is as follows:
[0012] η g =w g (ξ(N g +1)-ξ(N g )) (2)
[0013] Generate network-level routes from grid-level routes:
[0014] Given a grid-level route r G =(g1,…,g n ), and convert it into a route at the road network level according to the following rules:
[0015] S1.1: Randomly select a road network node v1 in the starting grid g1 as the starting point of the road network level route.
[0016] S1.2: For any i=1,…,n-1: In the grid g i Find the distance v i The nearest network node is used as the next destination v i+1 , from v i to v i+1 The actual route at the road network level is obtained according to the shortest path algorithm
[0017] S1.3: Connect the routes in step S1.2 to obtain the final route r at the road network level N =(v1,…,v m ).
[0018] Step 2: Dual-space-scale route planning for a single dedicated vehicle.
[0019] The path planning of a single dedicated vehicle is given a fixed time window [T1, T2]; first, a grid with the largest marginal gain is greedily selected as the starting grid o; the overall strategy of route planning is to gradually perform local searches on the target grid while planning the path at the grid level to collect sensor utility in an efficient manner; this process will be repeated while updating the marginal utility of the traversed grids until the time budget T2-T1 is reached.
[0020] Given a starting grid The idea is to find a destination grid and a grid-level route that maximizes the perceived benefit ψ(·):
[0021]
[0022] In the formula, r G It is a line from g o to g d The path at the grid level with the least obstruction, which is calculated by a modified A-star algorithm, where the impedance of the grid g is the inverse of its marginal utility 1 / η g ;This improvement of the A-star algorithm encourages the dedicated vehicle to preferentially scan the grids with high marginal utility; Indicates the route along the road network level The total time of operation; perceived utility ψ(g d ) is the path along the road network level The ratio of the weighted perceived utility obtained by running to the path travel time; by maximizing ψ(g d ), the path planning strategy aims to obtain more perception benefits in a shorter time.
[0023] Since the objective function ψ(·) is embedded with the A-star search algorithm and dual-space-scale path planning components, its complexity is relatively high. Therefore, we enumerate the starting grid g o Start by searching for the destination in all grids within a certain radius R. The specific steps are as follows:
[0024] S2.1: Find the starting grid g o The set of all grids within the radius R is defined as
[0025] S2.2: For Apply the improved A-star algorithm to find the o to g d The grid-level path;
[0026] According to formula (3), ψ(g d ).
[0027] S2.3: Find
[0028] Finally, a single-vehicle dedicated route is found by repeatedly performing the above search process until the time budget limit is reached.
[0029] Step 3: Dual-space-scale route planning for multiple special vehicles.
[0030] For multiple special vehicles, the method of first serializing and generating the route of each vehicle and then iteratively adjusting it is adopted to avoid overlapping of vehicle routes and maximize the overall perceived utility. The specific steps are as follows:
[0031] S3.1: Generate the route of each vehicle in turn according to the algorithm in step 2.
[0032] S3.2: Adjust the route of each vehicle, keep the routes of other vehicles unchanged, remove the route of one of the vehicles, update the marginal gain of each grid, and re-plan the route of this vehicle.
[0033] S3.3: Repeat step S3.2 until the routes of all vehicles do not change.
[0034] The beneficial technical effects of the present invention compared with the prior art are:
[0035] The method of the present invention can generate path planning solutions for objects with different monitoring weights, achieve differentiated coverage of different monitoring objects by multiple vehicles (times) under given time or mileage constraints, and maximize monitoring utility. The invention solves the problem that large-scale inspection tasks are highly dependent on manual experience, have strong information bias, and have low monitoring resource utilization, and can be widely used in various special monitoring scenarios. BRIEF DESCRIPTION OF THE DRAWINGS
[0036] Figure 1 The present invention is a flow chart of a dual-space-scale route planning method for a special vehicle suitable for grid monitoring. DETAILED DESCRIPTION
[0037] The present invention is further described in detail below with reference to the accompanying drawings and specific implementation methods.
[0038] This paper proposes a dual-space-scale route planning method for special vehicles suitable for grid monitoring, aiming to solve how to plan the routes of special fleets within a limited time or distance budget to maximize the perception benefits at the grid level. This problem is significantly different from the traditional vehicle routing problem in two aspects:
[0039] (1) Perception utility is obtained through the coverage of spatial grids, while vehicle route planning is restricted by the road network, forming a dual spatial scale structure;
[0040] (2) Multiple visits to the same grid can bring additional utility gains, which poses new challenges to traditional vehicle route planning.
[0041] In this regard, the present invention provides a dual-space-scale route planning method for a special vehicle suitable for grid monitoring, comprising the following steps:
[0042] Step 1: Generate a road network level path based on the grid level path.
[0043] make Represents the node set of the road network, Represents a set of spatial grids in the monitoring area. If a node In space grid In, it is represented by v∈g; with r N Represents the path of the vehicle at the road network level, expressed as a sequence of road network nodes, that is, use Represents the path of the vehicle at the grid level, expressed as a sequence of grids, i.e. The paths at the dedicated fleet network level and the path sets at the grid level are expressed as and
[0044] Quantifying perceived benefits: For a grid Its coverage number is N g , the perception score ξ of the grid is calculated as follows:
[0045] ξ=(N g ) β (1)
[0046] In the formula, β is the parameter of the utility function. The smaller β is, the smaller the marginal utility of multiple sampling is.
[0047] Marginal gain: For a grid Its perceptual weight is w g , when the grid has been covered by N g The marginal gain η brought by the next visit g The calculation is as follows:
[0048] η g =w g (ξ(N g +1)-ξ(N g )) (2)
[0049] Generate network-level routes from grid-level routes:
[0050] Given a grid-level route r G =(g1,…,g n), and convert it into a route at the road network level according to the following rules:
[0051] S1.1: Randomly select a road network node v1 in the starting grid g1 as the starting point of the road network level route.
[0052] S1.2: For any i=1,…,n-1: In the grid g i Find the distance v i The nearest network node is used as the next destination v i+1 , from v i to v i+1 The actual route at the road network level is obtained according to the shortest path algorithm
[0053] S1.3: Connect the routes in step S1.2 to obtain the final route r at the road network level N =(v1,…,v m ).
[0054] Step 2: Dual-space-scale route planning for a single dedicated vehicle.
[0055] The path planning of a single dedicated vehicle is given a fixed time window [T1, T2]. First, a grid with the largest marginal gain is greedily selected as the starting grid o; the overall strategy of route planning is to gradually search the target grid locally while planning the path at the grid level to collect sensor utility in an efficient way; this process will be repeated while updating the marginal utility of the traversed grids until the time budget T2-T1 is reached.
[0056] Given a starting grid The idea is to find a destination grid and a grid-level route that maximizes the perceived benefit ψ(·):
[0057]
[0058] In the formula, r G It is a line from g o to g d The path at the grid level with the least obstruction, which is calculated by a modified A-star algorithm, where the impedance of the grid g is the inverse of its marginal utility 1 / η g ;This improvement of the A-star algorithm encourages the dedicated vehicle to preferentially scan the grids with high marginal utility; Indicates the route along the road network level The total time of operation; perceived utility ψ(g d ) is the path along the road network level The ratio of the weighted perceived utility obtained by running to the path travel time; by maximizing ψ(gd ), the path planning strategy aims to obtain more perception benefits in a shorter time.
[0059] Since the objective function ψ(·) is embedded with the A-star search algorithm and dual-space-scale path planning components, its complexity is relatively high. Therefore, we enumerate the starting grid g o Departure, all grids within a certain radius R (for example, 5 kilometers) are searched for the destination. The specific steps are as follows:
[0060] S2.1: Find the starting grid g o The set of all grids within the radius R is defined as
[0061] S2.2: For Apply the improved A-star algorithm to find the o to g d The grid-level path;
[0062] According to formula (3), ψ(g d ).
[0063] S2.3: Find
[0064] Finally, a single-vehicle dedicated route is found by repeatedly performing the above search process until the time budget limit is reached.
[0065] Step 3: Dual-space-scale route planning for multiple special vehicles.
[0066] For multiple special vehicles, the method of first serializing and generating the route of each vehicle and then iteratively adjusting it is adopted to avoid overlapping of vehicle routes and maximize the overall perceived utility. The specific steps are as follows:
[0067] S3.1: Generate the route of each vehicle in turn according to the algorithm in step 2.
[0068] S3.2: Adjust the route of each vehicle, keep the routes of other vehicles unchanged, remove the route of one of the vehicles, update the marginal gain of each grid, and re-plan the route of this vehicle.
[0069] S3.3: Repeat step S3.2 until the routes of all vehicles do not change.
[0070] Taking Longquanyi in Chengdu as an example, when the grid of the monitoring area is divided into a spatial grid of 1km×1km, the special vehicle route planning method can make up for the monitoring blind spots of other perception methods (taxis, buses) and further improve the perception utility.
Claims
1. A dual-space-scale route planning method for a special vehicle suitable for grid monitoring, comprising the following steps: Step 1: Generate a road network level path based on the grid level path; make Represents the node set of the road network, Represents a set of spatial grids in the monitoring area. If a node In space grid In, it is represented by v∈g; with r N Represents the path of the vehicle at the road network level, expressed as a sequence of road network nodes, that is, use Represents the path of the vehicle at the grid level, expressed as a sequence of grids, i.e. The paths at the dedicated fleet network level and the path sets at the grid level are expressed as and Quantifying perceived benefits: For a grid Its coverage number is N g , the perception score ξ of the grid is calculated as follows: ξ=(N g ) β (1) In the formula, β is the parameter of the utility function. The smaller β is, the smaller the marginal utility of multiple sampling is. Marginal gain: For a grid Its perceptual weight is w g , when the grid has been covered by N g The marginal gain η brought by the next visit g The calculation is as follows: or g =w g (ξ(N g +1)-ξ(N g )) (2) Generate network-level routes from grid-level routes: Given a grid-level route r G =(g1,…,g n ), and convert it into a route at the road network level according to the following rules: S1.1: Randomly select a road network node v1 in the starting grid g1 as the starting point of the road network level route; S1.2: For any i=1,…,n-1: In the grid g i Find the distance v i The nearest network node is used as the next destination v i+1 , from v i to v i+1 The actual route at the road network level is obtained according to the shortest path algorithm S1.3: Connect the routes in step S1.2 to obtain the final route r at the road network level N =(v1,…,v m ); Step 2: Dual-space-scale route planning for a single dedicated vehicle; The path planning of a single dedicated vehicle is given a fixed time window [T1, T2]; first, a grid with the largest marginal gain is greedily selected as the starting grid o; The overall strategy of route planning is to gradually conduct local searches on the target grid while planning the path at the grid level to collect the sensing utility in an efficient way; This process is repeated while updating the marginal utility of the traversed grid until the time budget T2-T1 is reached; Given a starting grid The idea is to find a destination grid and a grid-level route that maximizes the perceived benefit ψ(·): In the formula, r G It is a line from g o to g d The path at the grid level with the least obstruction, which is calculated by a modified A-star algorithm, where the impedance of the grid g is the inverse of its marginal utility 1 / η g ;This improvement of the A-star algorithm encourages the dedicated vehicle to preferentially scan the grids with high marginal utility; Indicates the route along the road network level Total running time; perceived utility ψ(g d ) is the path along the road network level The ratio of the weighted perceived utility obtained by running to the path travel time; by maximizing ψ(g d ), the path planning strategy aims to obtain more perception benefits in a shorter time; Since the objective function ψ(·) is embedded with the A-star search algorithm and dual-space-scale path planning components, its complexity is relatively high. Therefore, we enumerate the starting grid g o Start by searching for the destination in all grids within a certain radius R. The specific steps are as follows: S2.1: Find the starting grid g o The set of all grids within the radius R is defined as S2.2: For Apply the improved A-star algorithm to find the o to g d The grid-level path; According to formula (3), ψ(g d ); S2.3: Find Finally, a single-vehicle dedicated route is found by repeatedly performing the above search process until the time budget limit is reached; Step 3: Dual spatial scale route planning for multiple special vehicles; For multiple special vehicles, the method of first serializing and generating the route of each vehicle and then iteratively adjusting it is adopted to avoid overlapping of vehicle routes and maximize the overall perceived utility. The specific steps are as follows: S3.1: Generate the route of each vehicle in turn according to the algorithm in step 2; S3.2: Adjust the route of each vehicle, keep the routes of other vehicles unchanged, remove the route of one of the vehicles, update the marginal gain of each grid, and re-plan the route of this vehicle; S3.3: Repeat step S3.2 until the routes of all vehicles do not change.