Construction path planning optimization method combining YOLO v8 and A-Star algorithm
By combining YOLO v8 and A-Star algorithms, this method utilizes drone aerial photography to acquire an image database, distinguishes between targets and obstacles, establishes a cost map, identifies dynamic obstacles in real time, and optimizes construction paths. This solves the accuracy and efficiency problems of path planning in dynamic environments in existing technologies, and achieves efficient planning of construction paths.
Patent Information
- Application Number
- CN202511556674.4
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-10-29
- Publication Date
- 2026-01-27
AI Technical Summary
Existing construction path planning technologies have low accuracy in dynamic environments and are prone to misidentifying obstacles. Furthermore, the A-Star algorithm cannot effectively plan the optimal path in dynamic scenarios, leading to increased construction costs.
By combining YOLO v8 and A-Star algorithms, an image database is acquired through drone aerial photography to distinguish between construction targets and obstacles, a cost map is built, the optimal path is selected using a cost evaluation function, and dynamic obstacles are identified in real time using YOLO v8 to optimize path planning.
It improves the accuracy and adaptability of construction path identification, reduces construction costs, avoids inefficient path planning, and provides comprehensive decision support.
Smart Images
Figure CN121409271A_ABST
Abstract
Description
Technical Field
[0001] This invention belongs to the field of engineering construction technology, specifically relating to a construction path planning optimization method that combines YOLO v8 and A-Star algorithms. Background Technology
[0002] Currently, path planning for construction machinery primarily relies on drones or LiDAR to acquire overall environmental information of the construction site and uses the A-Star algorithm to plan paths in a static environment. However, this approach faces multiple problems in practical applications: Firstly, it largely depends on the operator's experience and skill, making it difficult to cope with complex and ever-changing dynamic construction environments, and lacks a solution for effectively combining and analyzing information from the overall construction environment and specific construction processes. Secondly, the A-Star algorithm itself has significant shortcomings: Firstly, it mainly scans the construction site using LiDAR, which has complex obstacle types, making it difficult to classify specific obstacle types and potentially misidentifying objects that look similar to obstacles as obstacles. Secondly, this algorithm can only plan paths based on fixed static data of "coordinates + distance," resulting in low accuracy in identifying movable obstacles in dynamic scenarios. Data processing is also lagging, making it unsuitable for dynamic environments and potentially leading to inefficient path planning due to excessive detours, indirectly increasing construction costs. Summary of the Invention
[0003] The purpose of this invention is to provide a construction path planning optimization method that combines YOLO v8 and A-Star algorithm. By using YOLO v8 to identify and measure dynamic obstacles in the construction process in real time, and combining it with the A-Star algorithm for path planning, the accuracy and adaptability of path selection are significantly improved.
[0004] To achieve the above objectives, the present invention provides the following technical solution: A construction path planning optimization method combining YOLO v8 and A-Star algorithms includes the following steps: Step 1: Before construction, collect images of the entire construction site, obtain image information of construction targets and obstacles, define primary and secondary targets respectively, and establish an image database based on the initial position of construction machinery; Step 2: Based on the image database established in Step 1, the overall construction site image is converted into a network map using 3D reconstruction and semantic segmentation methods; Step 3: Define different types of obstacles as costs with different weights, and convert the network map into a cost map; Step 4: Use the A-Star algorithm to plan different paths on the cost map, find the "trap paths" in the cost map, manually adjust the cost weights on the "trap paths" according to the obstacle avoidance difficulty, select the best path with the lowest cost, and display the selected best path on the left half of the electronic screen located on one side of the operating area inside the construction machinery. Step 5: During formal construction, real-time on-site image information is obtained through a stereo camera installed at the front of the construction machinery. YOLO v8 technology is used to distinguish between primary and secondary target objects, and the target objects are identified and distances are measured in real time to find the shortest path for obstacle avoidance. The effect is then displayed on the right half of the electronic screen. Step 6: Refer to the static optimal path diagram displayed in the left half of the electronic screen and the real-time dynamic scene obtained based on the YOLO v8 algorithm in the right half of the electronic screen, and comprehensively optimize the construction path dynamically.
[0005] Furthermore, the criteria for distinguishing between primary and secondary targets in step 1 are as follows: the construction target is defined as the primary target, and the areas on the left and right sides of the obstacle are defined as secondary targets.
[0006] Furthermore, when the obstacles are stacked, the group of obstacles is regarded as a whole, and the outermost two ends of the whole obstacle are defined as secondary targets.
[0007] Furthermore, the optimal path with the minimum cost described in step 4 is filtered using a cost evaluation function, which satisfies the following expression: in This represents the total cost of the path. Let $\sum$ be the cost of one of the paths in the given path. From the starting point to the current node The cost, For the current node To the next node The cost, and so on. To reach the next node The path cost is calculated by summing the costs of each node and then adding the costs of each node together to obtain the total cost of the entire path. The higher the cost weight, the lower the priority of entering the path planning scope.
[0008] Furthermore, the distinction between primary and secondary targets in step 5 is achieved using YOLO v8 technology combined with a priority evaluation function, which satisfies the following expression: in, The total score after prioritizing the objects. The priority assessment score for the primary target object. The priority assessment score for secondary targets. This is the indicator function for the primary target object. For secondary target indication functions, , Let be the weighting coefficient, satisfying .
[0009] Compared with the prior art, the present invention has the following beneficial effects: This invention utilizes an image database built upon UAV aerial photography technology. Combined with the initial position of construction machinery, a cost map is generated. By increasing the cost weight of objects in "trap paths" and reducing the priority of these paths in planning, path selection under the A-Star algorithm is achieved, thereby generating an optimal path. The YOLO v8 algorithm enhances path planning capabilities in dynamic scenes, facilitating the identification of the shortest obstacle avoidance path and eliminating the influence of dynamic obstacles. On an electronic screen, a static optimal path map and a real-time dynamic scene from the YOLO v8 algorithm's perspective are displayed in a split-screen format. This, combined with the on-site situation, provides operators with comprehensive decision support, preventing the selection of "trap" paths. While the A-Star algorithm is primarily used for path planning in static scenes, the YOLO v8 algorithm enables real-time identification and ranging of dynamic objects. The YOLO v8 algorithm optimizes the insensitivity to dynamic scenes in path planning based on the A-Star algorithm, improving the ability to identify movable obstacles.
[0010] This invention utilizes UAV aerial photography technology to acquire overall construction image information, simultaneously completing the basic modeling of the construction environment's grid map and the acquisition of target object image information, reducing time costs. Primary and secondary targets are defined separately, enabling the system to accurately distinguish between construction targets and obstacles. Furthermore, by acquiring accurate image information, the A-Star algorithm significantly reduces the likelihood of construction targets with similar appearances to obstacles being mistaken for obstacles, improving image recognition capabilities. When obstacles and construction targets with similar appearances appear simultaneously within the same recognition range, the system combines an image database and the YOLOv8 algorithm, using a priority evaluation function to quickly identify and distinguish between primary and secondary targets, avoiding misidentification of construction targets as obstacles.
[0011] Furthermore, the construction target is defined as a primary target, and the sides of the obstacle are defined as secondary targets. A priority evaluation function is used to distinguish between the target and the obstacle, and their priorities are determined. The stacked obstacle group is treated as a whole to prevent the selection of an inefficient, excessively circuitous path under the A-Star algorithm's path planning. Attached Figure Description
[0012] Figure 1 This is a flowchart of the present invention; Figure 2 This is an aerial view of the entire construction site taken by a drone; Figure 3 This is a diagram of a "trap path". Detailed Implementation
[0013] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, the present invention will be described in detail below with reference to the accompanying drawings and specific embodiments.
[0014] like Figure 1 As shown in this embodiment, a construction path planning optimization method combining YOLO v8 and A-Star algorithms includes the following steps: Step 1: Before construction, collect images of the entire construction site, obtain image information of construction targets and obstacles, and define primary and secondary targets respectively. Combine this with the initial positions of construction machinery to establish an image database. like Figure 2 As shown, during the construction preparation phase, based on the initial positions of the construction machinery, drone aerial photography technology is used to acquire images and collect information about the entire construction site. The main focus is on acquiring images of construction targets and obstacles. Construction targets are defined as primary targets, and the areas on either side of obstacles are defined as secondary targets. When obstacles are stacked, the obstacle group is treated as a whole, and the outermost two ends of the overall obstacle group are defined as secondary targets. Finally, based on all the data acquired by the drone, an image database of the entire construction site is established.
[0015] Step 2: Based on the image database established in Step 1, convert the overall construction site images into a web map using 3D reconstruction and semantic segmentation methods. Step 3: Define different types of obstacles as costs with different weights, and convert the network map into a cost map. For example, the cost of a tree is defined as 1, a patch of grass as 3, and a rock as 10, etc. The cost of going from a tree to a patch of grass is 3, and the cost of going from a patch of grass to a tree is 1.
[0016] Step 4: Use the A-Star algorithm to plan different paths on the cost map, identify the "trap paths" in the cost map, manually adjust the cost weights on the "trap paths" according to the obstacle avoidance difficulty, select the optimal path with the lowest cost, and display the selected optimal path on the left half of the electronic screen located on one side of the operating area inside the construction machinery. Objects with higher cost weights have lower priority when entering the path planning range. The cost weights of objects can be adjusted according to the difficulty of obstacle avoidance.
[0017] The optimal path with the minimum cost is selected using a cost evaluation function, which satisfies the following expression: in This represents the total cost of the path. Let $\sum$ be the cost of one of the paths in the given path. From the starting point to the current node The cost, For the current node To the next node The cost, and so on. To reach the next node The path cost is calculated by summing the costs of each node and then adding the costs of each node together to obtain the total cost of the entire path.
[0018] Path planning methods based on the A-Star algorithm prioritize finding the path with the minimum total cost. By manually increasing the cost weight of objects in "trap paths," the priority of "trap paths" under the A-Star algorithm is reduced, thereby selecting the optimal path.
[0019] like Figure 3 As shown, "trap paths" refer to situations where obstacles obstruct the view, preventing the stereo camera from identifying the construction target in time and thus hindering the planning of the optimal path to the target. For example, when the distances to the obstacle avoidance paths on the left and right sides of an obstacle are unequal, the YOLO v8 algorithm will identify the shortest path. However, due to limited information, the identified shortest path may be a longer path to the construction target after obstacle avoidance. Conversely, when the distances to the obstacle avoidance paths on the left and right sides of an obstacle are equal, the YOLO v8 algorithm can choose either path, but this may not be the optimal path to the construction target. Therefore, to accurately plan the optimal path, it is necessary to consider the "trap path" situation.
[0020] Step 5: During formal construction, real-time on-site image information is acquired through a stereo camera installed at the front of the construction machinery. YOLO v8 technology is used to distinguish between primary and secondary targets, and the targets are identified and distances are measured in real time to find the shortest obstacle avoidance path. The results are displayed on the right half of the electronic screen. Among them, YOLO v8 technology distinguishes between primary and secondary targets based on a priority evaluation function.
[0021] The priority evaluation function satisfies the following expression: in, The total score after prioritizing the objects. The priority assessment score for the primary target object. The priority assessment score for secondary targets. This is the indicator function for the primary target object. For secondary target indication functions, , Let be the weighting coefficient, satisfying .
[0022] Step 6: Referencing the static optimal path map displayed in the left half of the electronic screen and the real-time dynamic scene obtained based on the YOLO v8 algorithm in the right half of the electronic screen, dynamically optimize the construction path. By using the YOLO v8 algorithm to identify and measure the distance to secondary targets, the shortest path for obstacle avoidance can be quickly found. At the same time, the path planning map on the left side of the electronic screen should also be consulted to prevent the vehicle from entering a "trap" path due to obstruction by obstacles.
Claims
1. A construction path planning optimization method combining YOLO v8 and A-Star algorithms, characterized in that, Includes the following steps: Step 1: Before construction, collect images of the entire construction site, obtain image information of construction targets and obstacles, define primary and secondary targets respectively, and establish an image database based on the initial position of construction machinery; Step 2: Based on the image database established in Step 1, the overall construction site image is converted into a network map using 3D reconstruction and semantic segmentation methods; Step 3: Define different types of obstacles as costs with different weights, and convert the network map into a cost map; Step 4: Use the A-Star algorithm to plan different paths on the cost map, find the "trap paths" in the cost map, manually adjust the cost weights on the "trap paths" according to the obstacle avoidance difficulty, select the best path with the lowest cost, and display the selected best path on the left half of the electronic screen located on one side of the operating area inside the construction machinery. Step 5: During formal construction, real-time on-site image information is obtained through a stereo camera installed at the front of the construction machinery. YOLO v8 technology is used to distinguish between primary and secondary target objects, and the target objects are identified and distances are measured in real time to find the shortest path for obstacle avoidance. The effect is then displayed on the right half of the electronic screen. Step 6: Refer to the static optimal path diagram displayed in the left half of the electronic screen and the real-time dynamic scene obtained based on the YOLO v8 algorithm in the right half of the electronic screen, and comprehensively optimize the construction path dynamically.
2. The construction path planning optimization method combining YOLO v8 and A-Star algorithm according to claim 1, characterized in that, The criteria for distinguishing between primary and secondary targets in step 1 are as follows: the construction target is defined as the primary target, and the areas on the left and right sides of the obstacle are defined as secondary targets.
3. The construction path planning optimization method combining YOLO v8 and A-Star algorithm according to claim 2, characterized in that, When the obstacles are stacked, the group of obstacles is regarded as a whole, and the two outermost ends of the whole obstacle are defined as secondary targets.
4. The construction path planning optimization method combining YOLO v8 and A-Star algorithm according to claim 3, characterized in that, The optimal path with the minimum cost described in step 4 is selected using a cost evaluation function, which satisfies the following expression: in This represents the total cost of the path. Let $\sum$ be the cost of one of the paths in the given path. From the starting point to the current node The cost, For the current node To the next node The cost, and so on, To reach the next node The path cost is calculated by summing the costs of each node and then adding the costs of each node together to obtain the total cost of the entire path. The higher the cost weight, the lower the priority of entering the path planning scope.
5. The construction path planning optimization method combining YOLO v8 and A-Star algorithm according to claim 4, characterized in that, The distinction between primary and secondary targets in step 5 is achieved using YOLO v8 technology combined with a priority evaluation function, which satisfies the following expression: in, The total score after prioritizing the objects. The priority assessment score for the primary target object. The priority assessment score for secondary targets. This is the indicator function for the primary target object. For secondary target indication functions, , Let be the weighting coefficient, satisfying .