A global path planning and return method based on satellite map

Abstract

The application relates to the technical field of automatic driving global path planning, and particularly discloses a global path planning and return method based on a satellite map, a global path planning module in an automatic driving system of an unmanned vehicle, and the global path planning module comprises the following steps: constructing a road network according to a satellite map, globally planning according to a task point, re-planning, and returning, the global path planning module is provided with an I / O data interface, can receive vehicle position information, and send a global planning result; the satellite map is manually distinguished to obtain a drivable road section, and manual marking of the road section is completed through relevant software; according to a breadth-first algorithm, an optimal driving route is solved according to the sequence of the task points, and the vehicle successfully arrives at a task end point; after the vehicle arrives at the task end point, the program allows two return modes to be selected, i.e., a return along an original road and a return along the nearest road; the application draws a driving road network based on the satellite map, realizes global path planning and return, and has the advantages of strong adaptability, low operation difficulty and low hardware requirement.

Description

Technical Field

[0001] This invention relates to the field of global path planning technology for autonomous driving, and specifically to a global path planning and return method based on satellite maps. Background Technology

[0002] At present, some listed companies have relatively mature civilian autonomous driving technology, which can provide relatively safe and reliable control solutions for vehicles on urban structured roads.

[0003] In civilian autonomous driving systems, such as Figure 1 As shown, the implementation of global path planning relies on high-precision map information, which is generally collected and marked by real vehicles. This high-precision map records detailed information such as distance, road conditions, and road signs, serving as an important reference factor for local planning. Combined with various positioning and perception modules, it comprehensively acquires real-time vehicle information, completes local planning and decision-making, and then generates control signals to send operation commands to the vehicle chassis, thus forming the general framework of autonomous driving.

[0004] Disadvantages of existing technology:

[0005] Existing technologies are relatively mature, possessing comprehensive technical reserves in positioning, perception, decision-making, and control, but all rely on manually collected high-precision map data. While these solutions are suitable for structured urban roads and can ensure the accuracy and safety of autonomous driving, they are unsuitable for the operational environment of military vehicles. For military unmanned driving applications, the complex and rugged off-road environment of battlefields and the special needs of material transfer, storage, and transportation must be considered. Civilian global planning technologies based on high-precision maps require preprocessing work such as data collection, annotation, and correction of the driving environment, making them unsuitable for global path planning in unknown environments. A more reasonable and feasible technical solution is needed for specific situations.

[0006] Therefore, it is necessary to design a global path planning and return method based on satellite maps to solve the problem that high-precision maps cannot be obtained in existing technologies. Summary of the Invention

[0007] In view of the problems existing in the prior art, the purpose of this invention is to provide a global path planning and return method based on satellite maps.

[0008] The technical solution adopted by this invention to solve its technical problem is: a global path planning and return method based on satellite maps, specifically including the following steps:

[0009] S1. The global path planning module in the autonomous driving system of unmanned vehicles includes building a road network based on satellite maps, global planning based on task points, replanning, and return. The global path planning module has an I / O data interface, which can receive vehicle location information and send global planning results.

[0010] S2. For satellite maps, manually identify drivable road sections and manually mark the road sections using relevant software. This step directly affects the success of subsequent real vehicle tests.

[0011] S3. After the vehicle receives the autonomous driving command, the program reads the road network information and the given task point information. According to the width-first algorithm, the program solves the optimal driving route in the order in which the task points appear. During the vehicle's journey, the program continuously receives the vehicle's position information and updates the vehicle control data in real time until the vehicle successfully reaches the task destination.

[0012] S4. After the vehicle arrives at the mission endpoint, an interactive box pops up on the display interface, allowing the program to select two return modes: return via the original route and return via the shortest route.

[0013] Preferably, the manually marked driving segments in step S2 should be located in the center of the road as much as possible. The marked segments should be densified, smoothed, and corrected using relevant software. The intersecting roads should be disconnected from each other to become independent roads, and the coordinates of the endpoints of the two connected roads should be consistent.

[0014] Preferably, the driving section is a complex off-road surface. The main geographical information, including passable roads, major obstacles that significantly affect driving, and forks, is identified manually. The road network is drawn manually, and the software assists in completing waypoint densification, smooth extraction, and correction of independent waypoints. Given the latitude and longitude coordinates of several points on the satellite map, the software can calculate the accurate latitude and longitude coordinates of all road network points.

[0015] Preferably, if a passable route exists in the optimal driving route solved in step S3, it is marked on the road network display interface. If there are special cases where the given task points cannot be connected to each other, the task points are only marked on the road network display interface. The program allows the mouse to drag and fine-tune the position of the task points until a reasonable planned route is displayed. For the corners at the intersection of two road segments in the route, Bézier curves are used for smoothing to facilitate the vehicle's turning action. At the same time, the program receives the vehicle's current location information. When the closest distance between the vehicle and the road network is no more than 20 meters, the nearest point relative to the vehicle's current position is marked on the road network. Based on the obtained route information, the program sequentially provides the latitude and longitude coordinates of the next two hundred points and sends them to the local planning module as the basis for vehicle driving control.

[0016] Preferably, the local planning module has supporting local planning, vehicle control, and perception modules, which work together to be ultimately installed on the vehicle to complete autonomous driving control. The communication between the local planning, vehicle control, and perception modules is based on ROS.

[0017] Preferably, the task point information in step S3 consists of the sequence number, latitude and longitude coordinates, speed, and task attributes that the vehicle needs to pass through in sequence. The task point information program uses the kd-tree nearest neighbor search algorithm, which searches for the coordinates of the task point with all road network points as the main body. When the nearest distance from the task point to all road network points is not greater than 20 meters, the nearest neighbor point of the task point among all road network points is obtained. This is taken as a must-pass point for the vehicle, and the sequence number, speed, and various attributes of the task point are assigned to this nearest neighbor point.

[0018] Preferably, the breadth-first algorithm in step S3 starts from the task starting point, and each time searches for a passable edge to the next task point. The edge with the minimum cost is used as the planned route for that segment, and so on until the task endpoint.

[0019] Preferably, in step S3, the return route is arranged in reverse order of the task points, and the breadth-first algorithm is used again to solve for the return route. When selecting the shortest route for return, only the original task endpoint is taken as the return start point and the original task start point as the return endpoint. The A* algorithm is used to solve for the shortest return route. If the difference between the return start heading angle and the heading angle when the vehicle arrives at the endpoint is greater than 180°, the program automatically takes four points near the vehicle's location to generate a Bézier curve. The sampled waypoints are used as the basis for the vehicle to turn around autonomously. The two task points of the Bézier curve can be dragged with the mouse to manually correct the turning route. During the vehicle's journey, the program continuously receives the vehicle's position information and updates the vehicle control data in real time until the vehicle successfully returns to the task start point.

[0020] Preferably, the comprehensive cost of the A* algorithm is composed of the actual length of the route already traversed and the Euclidean distance from the current position to the target destination, and the route generated by triggering the shortest path return satisfies the strict shortest requirement.

[0021] Preferably, the application for dragging task points with the mouse is built on Qt and uses signals and keys to complete human-computer interaction.

[0022] The present invention has the following beneficial effects:

[0023] The global path planning and return method based on satellite maps designed in this invention solves the problem that autonomous vehicles cannot drive on unknown off-road environments. It draws a road network based on satellite maps and then realizes global path planning and return. Among the various existing processing methods, it stands out for its advantages such as strong adaptability, low operation difficulty and low hardware requirements.

[0024] The global path planning and return-to-home method based on satellite maps designed in this invention eliminates the need for manual high-precision road network acquisition. Instead, it forms a road network by manually marking drivable areas on satellite maps. Based on this road network, the unmanned platform performs global path planning using a breadth-first algorithm, referencing relevant essential mission points. After reaching the target destination, it can trigger the return-to-home function. The unmanned platform uses the breadth-first algorithm to obtain the original return-to-home route or uses the A* algorithm to solve for the shortest return-to-home route. This provides important reference for the transfer, storage, and transportation of combat supplies in complex and unknown battlefield environments. Attached Figure Description

[0025] Figure 1 This is a general framework diagram of an autonomous driving system.

[0026] Figure 2 This is a logic block diagram of signal exchange between the vehicle control module, global planning module, and local planning module of the present invention.

[0027] Figure 3 This is a flowchart of the basic logic of global path planning based on satellite maps.

[0028] Figure 4 This is a flowchart of the return-to-home process for global path planning and return-to-home methods based on satellite maps. Detailed Implementation

[0029] The technical solutions of the embodiments of the present invention will be described in further detail below with reference to the accompanying drawings. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without creative effort are within the scope of protection of the present invention.

[0030] Key term definitions:

[0031] Road network—A network of interconnected and interwoven roads within a certain area. In this patent, the road network is represented as a topology of drivable trajectories displayed by a series of latitude and longitude coordinate points. Replanning—When unforeseen situations arise during action, action planning requires continuous understanding of the environment and replanning of actions. In this patent, this is specifically manifested in the replanning of routes to obtain new paths when unavoidable obstacles appear in the planned road.

[0032] Example 1.

[0033] like Figures 2-4As shown, a global path planning and return method based on satellite maps is proposed. It is a global path planning module in an autonomous driving system for unmanned vehicles, based on the battlefield off-road environment. The global path planning module includes several main parts: building a road network based on satellite maps, global planning based on task points, replanning, and return. The global path planning module has an I / O data interface, which can receive vehicle location information and send global planning results. The program is designed with a human-machine interface, which can realize real-time correction of planning requirements.

[0034] The global path planning map, road network, and task files are all saved as configuration files in TXT format within the program's file system. These are read upon program startup. The road network and task files can be modified and saved through the program's human-computer interaction interface. The global path planning module uses ROS to build an I / O data interface, receiving vehicle position signals from the vehicle control module. Based on this data, it uses a kd-tree nearest neighbor search algorithm to find the vehicle's closest position within the road network, serving as the starting reference point for sending the vehicle's guidance path. If the vehicle is more than 20 meters away from any position in the road network, it is considered too far, and the guidance path is not sent. When the nearest point to the vehicle's current position exists in the road network, 200 points are extracted forward from this point along the overall global path and sent to the local planning module as the basis for local planning. Simultaneously, when the vehicle passes through key task points, relevant speed, task attributes, and other information are also synchronously transmitted to the local planning module. This information is updated in real-time as the vehicle's position moves. The local planning module plans a guiding path within a small area in front of the vehicle based on global planning information and perception information. When an obstacle appears that cannot be avoided, it will plan to reverse after a certain period of time until it leaves the blocked road. For global path planning, the road where the current vehicle is located is marked as a blocked road and removed from the global road network. The planning algorithm is then triggered again to obtain a new guiding path. This is the implementation principle of the replanning function.

[0035] For satellite maps, drivable road sections are manually identified and marked using relevant software. This step directly affects the success of subsequent real-vehicle testing. Manually marked road sections should be located in the center of the road as much as possible. For the marked road sections, relevant software should be used to complete waypoint densification, smooth extraction, and correction. It is required that intersecting roads be disconnected from each other and become independent roads, and the coordinates of the waypoints at the endpoints of the two connected roads should be consistent.

[0036] The driving route is a complex off-road surface. The main geographical information, including passable roads, major obstacles that obviously affect driving, and forks, is identified manually. The road network is drawn manually, and the software assists in completing waypoint densification, smooth extraction, and correction of independent waypoints. Given the latitude and longitude coordinates of several points on the satellite map, the software can calculate the accurate latitude and longitude coordinates of all road network points.

[0037] After receiving the autonomous driving command, the program reads the road network information and the given task point information. Using a breadth-first search algorithm, it solves for the optimal driving route according to the order in which the task points appear. During the vehicle's journey, the program continuously receives vehicle position information and updates vehicle control data in real time until the vehicle successfully reaches the task endpoint. If a passable route exists within the solved optimal driving route, it is marked on the road network display interface. If there are special cases where the given task points cannot be connected, the task points are only marked on the road network display interface. The program allows mouse dragging to fine-tune the task point positions until a reasonable planned route is displayed. For corners at the intersection of two road segments in the route, Bézier curves are used for smoothing to facilitate vehicle turning. Simultaneously, the program receives the vehicle's current position information. When the closest distance from the vehicle to the road network is no more than 20 meters, the program marks the nearest point relative to the vehicle's current position on the road network and, based on the obtained route information, sequentially provides the latitude and longitude coordinates of the next 200 points, sending them to the local planning module as the basis for vehicle driving control.

[0038] After the vehicle reaches the mission endpoint, an interactive dialog box pops up on the display screen. The program allows you to choose between two return modes: return via the original route and return via the shortest route. When returning via the original route, the mission points are arranged in reverse order, and the breadth-first search algorithm is used again to find the return route. When returning via the shortest route, only the original mission endpoint is used as the return start point, and the original mission start point is used as the return endpoint. The A* algorithm is used to find the shortest return route. If the difference between the starting heading angle and the heading angle when the vehicle reaches the endpoint is greater than 180°, the program automatically takes four points near the vehicle's location to generate a Bézier curve. These sampled waypoints serve as the basis for the vehicle's autonomous U-turn. The two mission points on the Bézier curve can be dragged with the mouse to manually correct the U-turn route. During the vehicle's journey, the program continuously receives vehicle position information and updates vehicle control data in real time until the vehicle successfully returns to the mission starting point.

[0039] Example 2 illustrates the technical solution of the present invention in detail with specific examples.

[0040] 1. This invention primarily utilizes three algorithms: kd-tree nearest neighbor search algorithm, which constructs a tree based on all road network points, traverses each task point to find its nearest neighbor, marks the found nearest neighbor as a necessary point for the vehicle, and considers task points with a minimum distance greater than 20 meters from the road network as too far and not marked; and Breadth-First Search (BFS) algorithm, which establishes the connection relationships of each road based on the road network file, where connected roads contain road points with the same coordinates, including the road number and whether the connection is to the start or end point of that road. The final representation is map{(id, start)(id, start)…(id, end)}. Combining the road topology connections, starting from the starting point (sequence number 0), the BFS algorithm is applied to search along the edge connections, connecting the starting point to the first sub-target point with sequence number 1, and continuing sequentially until the task endpoint. This achieves global path planning. The a_star algorithm, as above, applies the a_star algorithm to the existing road topology connection relationship. The actual length of the roads already traversed is the actual cost g_cost, and the Euclidean distance to the road is the expected cost h_cost. The two are added together to form the total cost a_cost. Starting from the starting point, the algorithm gradually approaches the destination to obtain the actual shortest global driving guidance path.

[0041] 2. The program incorporates several practical functions designed to meet real-world application needs. Task point addition, deletion, and modification: Implemented using Qt, the program allows users to drag task points while holding down the left mouse button to change the position of key points the vehicle passes through. Selecting a task point allows modification of its attributes to specific options. Pressing the DELETE key deletes unnecessary task points. These operations can be performed simultaneously during driving. Road network correction: In actual driving, if a deviation in a road network is detected due to drawing accuracy issues, users can press the ALT key while dragging the mouse to select the offset road network. After selection, pressing the up, down, left, and right arrow keys will correct the road network. The program offers two return-to-home modes: returning along the original route and returning via the shortest route. Approximately 5 meters before the vehicle reaches its destination, a pop-up button appears on the interface to select the mode. After the program guides the vehicle to its destination and stops, either return-to-home mode is selected. If the vehicle's current heading differs from the initial expected heading by more than 180°, a U-turn is triggered. The U-turn trajectory is fitted using a Bézier curve and can be modified interactively by pressing the SHIFT key on the keyboard and dragging the control points of the Bézier curve with the mouse. Throughout this process, the vehicle remains parked. After all adjustments are complete, pressing the right mouse button starts the vehicle, and the selected return-to-home plan is executed.

[0042] This invention presents a global path planning and return-to-home technology for complex battlefield environments, independent of high-precision road network information. It primarily addresses material transport strategies in complex off-road battlefield environments. This solution eliminates the need for manual high-precision road network data collection. Instead, it manually marks drivable areas on satellite maps to form a road network. Based on this road network, the unmanned platform performs global path planning using a breadth-first search algorithm, referencing relevant essential mission points. Upon reaching the target destination, it triggers a return-to-home function. The unmanned platform can use the breadth-first search algorithm to find the original return route or the A* algorithm to find the shortest return route. This invention provides crucial reference for the transfer, storage, and transportation of combat supplies in complex and unknown battlefield environments.

[0043] This invention is not limited to the above-described embodiments. Anyone should know that any structural changes made under the guidance of this invention, and any technical solutions that are the same as or similar to this invention, fall within the protection scope of this invention.

[0044] The technologies, shapes, and structures not described in detail in this invention are all known technologies.

Claims

1. A global path planning and return-to-home method based on satellite maps, characterized in that, Specifically, the following steps are included: S1. The global path planning module in the autonomous driving system of unmanned vehicles includes building a road network based on satellite maps, global planning based on task points, replanning, and return. The global path planning module has an I / O data interface, which can receive vehicle location information and send global planning results. S2. Based on the satellite map, manually identify drivable road sections and manually mark the road sections using relevant software. This step directly affects the success of subsequent real vehicle testing. S3. After receiving the autonomous driving command, the program reads the road network information and the given task point information. Based on the breadth-first algorithm, it solves the optimal driving route according to the order in which the task points appear. During the vehicle's journey, the program continuously receives the vehicle's position information and updates the vehicle control data in real time until the vehicle successfully reaches the task endpoint. If there is a passable route in the solved optimal driving route, it is marked on the road network display interface. If there are situations where the given task points cannot be connected to each other, the task points are only marked on the road network display interface. The program allows the mouse to drag and fine-tune the position of the task points until a reasonable planned route is displayed. For the corners at the intersection of two road segments in the route, Bézier curves are used for smoothing to facilitate the vehicle's turning maneuvers. S4. After the vehicle arrives at the mission endpoint, an interactive dialog box pops up on the display screen. The program allows you to choose between two return modes: return along the original route and return along the shortest route. For return along the original route, the mission points are arranged in reverse order, and the breadth-first search algorithm is used again to solve for the return route. When choosing the shortest route, only the original mission endpoint is used as the return start point, and the original mission start point is used as the return endpoint. A... The algorithm solves for the shortest return route. When the difference between the starting heading angle and the heading angle when the vehicle arrives at the destination is greater than 180°, the program automatically takes four points near the vehicle's location to generate a Bézier curve. The sampled waypoints serve as the basis for the vehicle to turn around autonomously. The two task points of the Bézier curve can be dragged with the mouse to manually correct the turning route. During the vehicle's journey, the program continuously receives vehicle location information and updates vehicle control data in real time until the vehicle successfully returns to the mission starting point.

2. The global path planning and return method based on satellite maps according to claim 1, characterized in that, In step S2, the manually marked driving segments must be located in the center of the road. The marked segments need to be densified, smoothed, and corrected using relevant software. The intersecting roads must be disconnected from each other to form an independent road segment, and the coordinates of the endpoints of the two connected roads must be consistent.

3. The global path planning and return method based on satellite maps according to claim 2, characterized in that, The driving route is a complex off-road surface. The main geographical information, including passable roads, obstacles affecting driving, and forks, is identified manually. The road network is drawn manually, and the software assists in completing waypoint densification, smooth extraction, and correction of independent waypoints. Given the latitude and longitude coordinates of several points on the satellite map, the software can calculate the accurate latitude and longitude coordinates of all road network points.

4. The global path planning and return method based on satellite maps according to claim 1, characterized in that, In step S3, the program simultaneously receives the vehicle's current location information. When the nearest distance between the vehicle and the road network is no more than 20 meters, the program marks the nearest point relative to the vehicle's current location on the road network. Based on the obtained route information, the program sequentially provides the latitude and longitude coordinates of two hundred points ahead and sends them to the local planning module as the basis for vehicle driving control.

5. The global path planning and return method based on satellite maps according to claim 4, characterized in that, The local planning module has supporting local planning, vehicle control, and perception modules, which work together to be installed in the vehicle to complete autonomous driving control. The communication between the local planning, vehicle control, and perception modules is based on ROS.

6. The global path planning and return method based on satellite maps according to claim 1, characterized in that, The task point information in step S3 consists of the sequence number that the vehicle needs to pass through, latitude and longitude coordinates, speed, and task attributes. The task point information program uses the kd-tree nearest neighbor search algorithm, taking all road network points as the main body, to search for the coordinates of the task point. When the nearest distance from the task point to all road network points is no more than 20 meters, the nearest neighbor point of the task point among all road network points is obtained, and this is taken as a must-pass point for the vehicle. At the same time, the sequence number of the task point, speed, and various attributes of the task are assigned to this nearest neighbor point.

7. The global path planning and return method based on satellite maps according to claim 1, characterized in that, The breadth-first algorithm in step S3 starts from the task starting point and searches for a passable edge to the next task point each time. The edge with the minimum cost is used as the planned route for that segment. This process continues until the task endpoint.

8. The global path planning and return method based on satellite maps according to claim 1, characterized in that, The comprehensive cost of the A* algorithm is the sum of the actual length of the route already traversed and the Euclidean distance from the current position to the target destination. The route generated by triggering the shortest path return satisfies the strict shortest requirement.

9. The global path planning and return method based on satellite maps according to claim 1, characterized in that, The application for dragging task points with the mouse is built on Qt and uses signals and keys to complete human-computer interaction.

Images