CN115373399B - Ground robot path planning method based on air-ground coordination - Google Patents
Ground robot path planning method based on air-ground coordination Download PDFInfo
- Publication number
- CN115373399B CN115373399B CN202211108617.6A CN202211108617A CN115373399B CN 115373399 B CN115373399 B CN 115373399B CN 202211108617 A CN202211108617 A CN 202211108617A CN 115373399 B CN115373399 B CN 115373399B
- Authority
- CN
- China
- Prior art keywords
- path
- node
- map
- robot
- ground
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
- 238000000034 method Methods 0.000 title claims abstract description 43
- 238000005457 optimization Methods 0.000 claims abstract description 8
- 230000006870 function Effects 0.000 claims description 51
- 230000011218 segmentation Effects 0.000 claims description 16
- 238000012549 training Methods 0.000 claims description 16
- 238000009499 grossing Methods 0.000 claims description 14
- 230000008859 change Effects 0.000 claims description 13
- 230000008569 process Effects 0.000 claims description 9
- 238000012360 testing method Methods 0.000 claims description 6
- 230000004888 barrier function Effects 0.000 claims description 5
- 230000000694 effects Effects 0.000 claims description 5
- 230000005540 biological transmission Effects 0.000 claims description 4
- 238000006073 displacement reaction Methods 0.000 claims description 4
- 238000011478 gradient descent method Methods 0.000 claims description 4
- 239000013598 vector Substances 0.000 claims description 4
- 238000013527 convolutional neural network Methods 0.000 claims description 3
- 230000009191 jumping Effects 0.000 claims description 3
- 238000004519 manufacturing process Methods 0.000 claims description 3
- 230000007246 mechanism Effects 0.000 claims description 3
- 230000009471 action Effects 0.000 abstract description 3
- 230000033001 locomotion Effects 0.000 description 16
- 230000015572 biosynthetic process Effects 0.000 description 10
- 238000005381 potential energy Methods 0.000 description 7
- 238000010586 diagram Methods 0.000 description 6
- 238000010276 construction Methods 0.000 description 4
- 238000005516 engineering process Methods 0.000 description 4
- XLYOFNOQVPJJNP-UHFFFAOYSA-N water Substances O XLYOFNOQVPJJNP-UHFFFAOYSA-N 0.000 description 4
- 230000008901 benefit Effects 0.000 description 2
- 238000009826 distribution Methods 0.000 description 2
- 230000007613 environmental effect Effects 0.000 description 2
- 238000011156 evaluation Methods 0.000 description 2
- 238000000605 extraction Methods 0.000 description 2
- 230000004927 fusion Effects 0.000 description 2
- 230000008447 perception Effects 0.000 description 2
- 238000010845 search algorithm Methods 0.000 description 2
- 230000009466 transformation Effects 0.000 description 2
- 206010000369 Accident Diseases 0.000 description 1
- 238000004458 analytical method Methods 0.000 description 1
- 238000013459 approach Methods 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 238000013135 deep learning Methods 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 238000013507 mapping Methods 0.000 description 1
- 238000013178 mathematical model Methods 0.000 description 1
- 239000011159 matrix material Substances 0.000 description 1
- 238000012545 processing Methods 0.000 description 1
- 238000011160 research Methods 0.000 description 1
- 238000006467 substitution reaction Methods 0.000 description 1
- 238000013519 translation Methods 0.000 description 1
- 238000004804 winding Methods 0.000 description 1
Classifications
-
- G—PHYSICS
- G05—CONTROLLING; REGULATING
- G05D—SYSTEMS FOR CONTROLLING OR REGULATING NON-ELECTRIC VARIABLES
- G05D1/00—Control of position, course, altitude or attitude of land, water, air or space vehicles, e.g. using automatic pilots
- G05D1/02—Control of position or course in two dimensions
- G05D1/021—Control of position or course in two dimensions specially adapted to land vehicles
- G05D1/0231—Control of position or course in two dimensions specially adapted to land vehicles using optical position detecting means
- G05D1/0238—Control of position or course in two dimensions specially adapted to land vehicles using optical position detecting means using obstacle or wall sensors
- G05D1/024—Control of position or course in two dimensions specially adapted to land vehicles using optical position detecting means using obstacle or wall sensors in combination with a laser
-
- G—PHYSICS
- G05—CONTROLLING; REGULATING
- G05D—SYSTEMS FOR CONTROLLING OR REGULATING NON-ELECTRIC VARIABLES
- G05D1/00—Control of position, course, altitude or attitude of land, water, air or space vehicles, e.g. using automatic pilots
- G05D1/02—Control of position or course in two dimensions
- G05D1/021—Control of position or course in two dimensions specially adapted to land vehicles
- G05D1/0212—Control of position or course in two dimensions specially adapted to land vehicles with means for defining a desired trajectory
- G05D1/0214—Control of position or course in two dimensions specially adapted to land vehicles with means for defining a desired trajectory in accordance with safety or protection criteria, e.g. avoiding hazardous areas
-
- G—PHYSICS
- G05—CONTROLLING; REGULATING
- G05D—SYSTEMS FOR CONTROLLING OR REGULATING NON-ELECTRIC VARIABLES
- G05D1/00—Control of position, course, altitude or attitude of land, water, air or space vehicles, e.g. using automatic pilots
- G05D1/02—Control of position or course in two dimensions
- G05D1/021—Control of position or course in two dimensions specially adapted to land vehicles
- G05D1/0212—Control of position or course in two dimensions specially adapted to land vehicles with means for defining a desired trajectory
- G05D1/0221—Control of position or course in two dimensions specially adapted to land vehicles with means for defining a desired trajectory involving a learning process
-
- G—PHYSICS
- G05—CONTROLLING; REGULATING
- G05D—SYSTEMS FOR CONTROLLING OR REGULATING NON-ELECTRIC VARIABLES
- G05D1/00—Control of position, course, altitude or attitude of land, water, air or space vehicles, e.g. using automatic pilots
- G05D1/02—Control of position or course in two dimensions
- G05D1/021—Control of position or course in two dimensions specially adapted to land vehicles
- G05D1/0231—Control of position or course in two dimensions specially adapted to land vehicles using optical position detecting means
- G05D1/0246—Control of position or course in two dimensions specially adapted to land vehicles using optical position detecting means using a video camera in combination with image processing means
- G05D1/0251—Control of position or course in two dimensions specially adapted to land vehicles using optical position detecting means using a video camera in combination with image processing means extracting 3D information from a plurality of images taken from different locations, e.g. stereo vision
-
- G—PHYSICS
- G05—CONTROLLING; REGULATING
- G05D—SYSTEMS FOR CONTROLLING OR REGULATING NON-ELECTRIC VARIABLES
- G05D1/00—Control of position, course, altitude or attitude of land, water, air or space vehicles, e.g. using automatic pilots
- G05D1/02—Control of position or course in two dimensions
- G05D1/021—Control of position or course in two dimensions specially adapted to land vehicles
- G05D1/0257—Control of position or course in two dimensions specially adapted to land vehicles using a radar
-
- G—PHYSICS
- G05—CONTROLLING; REGULATING
- G05D—SYSTEMS FOR CONTROLLING OR REGULATING NON-ELECTRIC VARIABLES
- G05D1/00—Control of position, course, altitude or attitude of land, water, air or space vehicles, e.g. using automatic pilots
- G05D1/02—Control of position or course in two dimensions
- G05D1/021—Control of position or course in two dimensions specially adapted to land vehicles
- G05D1/0276—Control of position or course in two dimensions specially adapted to land vehicles using signals provided by a source external to the vehicle
-
- G—PHYSICS
- G05—CONTROLLING; REGULATING
- G05D—SYSTEMS FOR CONTROLLING OR REGULATING NON-ELECTRIC VARIABLES
- G05D1/00—Control of position, course, altitude or attitude of land, water, air or space vehicles, e.g. using automatic pilots
- G05D1/02—Control of position or course in two dimensions
- G05D1/021—Control of position or course in two dimensions specially adapted to land vehicles
- G05D1/0287—Control of position or course in two dimensions specially adapted to land vehicles involving a plurality of land vehicles, e.g. fleet or convoy travelling
- G05D1/0289—Control of position or course in two dimensions specially adapted to land vehicles involving a plurality of land vehicles, e.g. fleet or convoy travelling with means for avoiding collisions between vehicles
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Radar, Positioning & Navigation (AREA)
- Remote Sensing (AREA)
- Aviation & Aerospace Engineering (AREA)
- General Physics & Mathematics (AREA)
- Automation & Control Theory (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Electromagnetism (AREA)
- Multimedia (AREA)
- Optics & Photonics (AREA)
- Control Of Position, Course, Altitude, Or Attitude Of Moving Bodies (AREA)
Abstract
The invention discloses a ground robot path planning method based on air-ground cooperation, which comprises the following steps: based on the space-based image acquisition data of the target area, constructing a complete space three-dimensional model of the target area; constructing a moving path area of the ground mobile robot based on the space three-dimensional model, and rasterizing the moving path area to obtain a binary grid map of the robot accessible path; and establishing a mixed map combined with the Voronoi map based on the binary grid map, performing mixed A-algorithm global path planning on the mixed map, and performing path optimization based on multi-guide-point traction action to obtain a reliable moving path of the ground robot. According to the ground robot path planning method based on air-ground coordination, the path planned by the robot can be pulled to the vicinity of a plurality of set guide points, and a safe and reliable path can be successfully planned in a multi-obstacle environment.
Description
Technical Field
The invention belongs to the technical field of space-ground coordination and path planning, and particularly relates to a ground robot path planning method based on space-ground coordination.
Background
In recent years, space-ground collaboration is a leading-edge research hotspot problem of collaborative operation of heterogeneous unmanned systems. The collaborative operation of the aerial unmanned aerial vehicle group and the ground mobile robot group is more advantageous than the single ground robot group or the aerial unmanned aerial vehicle group. The unmanned plane has ultrahigh maneuvering flight capacity and wide air vision, can quickly provide global information for the ground robot, can closely observe the state (local information) of a target area, and can be applied to the tasks of fire rescue, environmental exploration, target search, military reconnaissance and the like by combining the advantages of the ground robot and the ground robot.
The method aims at the reconnaissance and rescue tasks of the heterogeneous unmanned system in a large-scale fire accident scene, and due to the complexity of the fire extinguishing environment and the requirement on the real-time performance of environment perception, in order to realize autonomous navigation of the ground multi-fire-fighting robot in an unknown environment, map construction and path planning are extremely important links in the navigation technology, and the problem that the map construction of the ground robot is time-consuming can be effectively solved by adopting a multi-unmanned-plane collaborative map construction mode. Therefore, the rapid high-precision three-dimensional space collaborative mapping of the fire scene environment is realized by adopting the multi-unmanned aerial vehicle based on the real-time three-dimensional reconstruction technology, a reliable moving area map is constructed for the ground robot, and the two-value of the ground robot is rasterized according to the extracted reliable moving area map to obtain a grid map which can be used for path planning. Currently, common two-dimensional path planning algorithms based on graph search include an a-algorithm, a Dijkstra algorithm, an RRT algorithm, a D-algorithm, and the like, and the path planning algorithms are based on position level search and do not consider the actual orientation problem of robot motion. If the method is directly applied to path planning of the intelligent fire-fighting robot, some problems tend to occur: 1) Because the load and the dead weight of the robot are large and generally reach hundreds of kilograms, accurate control is difficult to realize, in practical application, the direction of the robot needs to be adjusted in advance before the path tracking task is executed each time, the robot is consistent with the initial direction of a planned path as much as possible, otherwise, the speed change of the robot is easy to be overlarge, and potential safety hazards are caused; 2) Considering that when the fire-fighting robot drags the water belt to move, the problem that the robot rotates or retreats to a large extent to cause the water belt to bend and burst is avoided, so that the planned path needs to be smooth enough, the direction of the path is consistent with the direction of the robot as much as possible, and the two-dimensional path planning algorithm cannot guarantee the direction of the path; 3) If the group fire robots take the form of formation, the planned path needs to consider the safety of the whole formation, but the traditional two-dimensional path planning algorithm usually takes the shortest path as a criterion, the planned path is close to the obstacle, and the safety of the whole formation cannot be ensured.
Disclosure of Invention
In view of the above analysis, the invention aims to disclose a ground robot path planning method based on air-ground coordination, which is used for solving the problems that a scene high-precision environment map is difficult to acquire, time is very consuming to build according to the ground robot map, and the conventional path planning algorithm cannot meet the requirements of the ground multi-robot path safety and the executable performance.
The invention discloses a ground robot path planning method based on air-ground coordination, which comprises the following steps:
Step S101, constructing a complete space three-dimensional model of a target area based on space-based image acquisition data of the target area;
Step S102, constructing a moving path area of the ground mobile robot based on the space three-dimensional model, and rasterizing the moving path area to obtain a binary grid map of the robot accessible path;
And step S103, a mixed map combined with the Voronoi map is established based on the binary grid map, mixed A algorithm global path planning is conducted on the mixed map, path optimization is conducted based on multi-guide-point traction, and a reliable moving path of the ground robot is obtained.
Further, the step S103 includes:
S401, establishing a Voronoi map by utilizing a binary grid map, and combining the Voronoi map and the binary grid map to construct a hybrid map;
Step S402, adopting mixed A search in a mixed map to search out a rough path based on the kinematics of the ground robot and the minimum safety steering constraint, and meeting the continuity of the path;
Step S403, optimizing the mixed-search coarse path based on a track smoother of a gradient descent optimization algorithm according to the coarse path, and constructing a path which is suitable for safe execution and continuous smoothing of the ground multi-robot.
Further, in the step S402, the mixed a search specifically includes:
Step S501, inputting a mixing map;
Step S502, initializing search;
the initialization comprises the steps of establishing an open_set list and a closed_set list, wherein the two lists are empty lists when being initialized, the open_set list stores nodes to be expanded, and the closed_set list stores expanded nodes or nodes with barriers;
step S503, judging that the OpenSet list is a non-empty list, selecting a node with the smallest total cost from the OpenSet list as an expansion node to expand, obtaining a plurality of expansion sub-nodes continuous with the state of the expansion node, and removing the expansion node from the OpenSet list to the Closed-Set list;
step S504, judging whether the extension node is a target point; if yes, the search is ended; if not, continuing to expand the child node of the node;
Step S505, judging whether the expansion child node collides with the obstacle, if so, discarding the child node; if not, jumping to the next step;
Step S506, comparing the actual cost of the current expansion child node with the actual cost of the original node, when the actual cost of the current expansion child node is lower than the actual cost of the original node, associating the continuous state of the current expansion child node with the grid unit, setting the grid of the current expansion child node as the father grid of the child node, putting the expansion child node into an OpenSet list, and updating the total cost of the expansion child node;
Step S507, the steps S503 to S506 in the above process are circulated until the target node is searched, and the target node is put into an OpenSet list;
step S508, returning all traversed father nodes from the target node in the OpenSet list until the starting node, obtaining the searched path, and ending the algorithm after the path searching is finished; if the open_set list is empty, it indicates that no feasible path has been planned and the algorithm terminates.
Further, the total cost of each node in the searching process comprises an actual cost and a heuristic estimated cost;
The actual cost is the cost from the initial node to the current node, and comprises the cost from the parent node to the current node and the cost from the parent node to the current node, wherein the actual path length from the parent node to the current node, the steering cost from the parent node to the current node in the binary grid map and the cost from the parent node to the current node in the Voronoi map are included in the latter;
The heuristic estimated cost is the heuristic estimated cost from the current node to the target node and is determined by a larger value in a heuristic function based on incomplete constraint and a heuristic function based on complete constraint.
Further, in the path optimization based on the multi-guide-point traction effect, a path smoother based on the multi-guide-point effect is adopted to carry out smoothing treatment on the rough path so as to obtain a path which is suitable for safe execution and continuous smoothing of the ground robot.
Further, the gradient descent smoother establishes a minimized cost function based on the obstacle term, the curvature term, the smoothing term, the voronoi field term and the multi-guide point function term, and solves an optimal path by using a gradient descent method.
Further, when there is |x i-gi|＜ρmax for each search node x i, the cost function P gui of the multi-guide point function term is:
Wherein x i is the two-dimensional plane coordinate of the vertex on the path; g i is the position of the closest guidance point to node x i; ρ max is the threshold for the maximum distance of the guide point affecting the cost function; the guidance point weight w gui is a coefficient by which the guidance point affects the path change.
Further, the step S101 includes:
Step S201, dividing a target area into a plurality of target subareas, acquiring a sequence image of the target subareas by using at least one unmanned aerial vehicle, and transmitting an image shrinkage map to a ground server by using a map transmission device;
step S202, in a ground server, carrying out real-time three-dimensional reconstruction to recover dense three-dimensional point clouds on the surface of the target sub-region;
and step 203, integrally fusing the single unmanned aerial vehicle mechanism images by utilizing a multi-map splicing and fusing algorithm based on three-dimensional point cloud data, and constructing a complete space three-dimensional model of the target area.
Further, in step S102, a deep learning method based on a convolutional neural network prediction remote sensing image is adopted, a road surface region in the space three-dimensional model scene is extracted, a reliable moving path region of the ground mobile robot is constructed, and a binary grid map of the accessible path is obtained by performing rasterization interpolation.
Further, the step S102 specifically includes:
Step S301, manufacturing a point cloud data training set;
Marking the characteristics including buildings, vegetation, ground, vehicles, pedestrians, lakes, wall surfaces and roadbed according to the three-dimensional model of the live-action, and constructing a data training set;
Step S302, establishing a deep semantic segmentation network model for extracting a reliable moving path region;
step S303, training and testing the deep semantic segmentation network model by utilizing the data training set;
Step S304, extracting a complete space three-dimensional model of the target area by utilizing the trained depth semantic segmentation network model to obtain a reliable moving path area of the robot in the target area;
Step S305, according to the reliable moving path area of the robot, rasterizing interpolation is carried out on the reliable moving path area to obtain a binary raster map of the accessible path.
The invention can realize the following beneficial effects:
According to the ground robot path planning method based on air-ground coordination, the path planned by the robot can be pulled to the vicinity of a plurality of set guide points, and a safe and reliable path can be successfully planned in a multi-obstacle environment.
The method and the system can be suitable for planning the overall path of the multi-robot formation, and the planned path is close to the middle position of the road, so that the requirement of the overall width of the formation can be met, and the safety of the formation movement is ensured.
The positions and the directions of all the path points can be obtained based on a three-dimensional path planning algorithm, so that the method is more convenient in designing a path or track tracking controller and more accords with the motion control of a robot; the robot is not only suitable for intelligent automobiles, but also suitable for other robots with kinematics and steering constraint, avoids unnecessary steering movement, and can reach a designated position in an accurate direction.
The method is suitable for the field of task emergency, such as fire rescue, global information is provided for the ground robot by utilizing the rapid maneuvering capability of the unmanned aerial vehicle, a three-dimensional space model of a high-precision site is rapidly constructed, reliable moving path areas including areas of roads, obstacles and the like are further extracted, and a reference map is provided for path planning of the ground robot.
Drawings
The drawings are only for purposes of illustrating particular embodiments and are not to be construed as limiting the invention, like reference numerals being used to refer to like parts throughout the several views.
FIG. 1 is a flow chart of a ground robot path planning method based on air-ground coordination in an embodiment of the invention;
FIG. 2 is a flowchart of a method for constructing a complete spatial three-dimensional model of a target area in an embodiment of the invention;
FIG. 3 is a flowchart of a training and testing process of a deep semantic segmentation network model in an embodiment of the present invention;
Fig. 4 is a flowchart of a method for implementing reliable mobile path planning by using an improved hybrid a algorithm in an embodiment of the present invention;
Fig. 5 is a basic flowchart of a hybrid a algorithm in an embodiment of the present invention;
Fig. 6 is a diagram of an expansion mode of 3 forward search nodes of the hybrid a algorithm in an embodiment of the present invention;
Fig. 7 is a diagram of a hybrid a algorithm search example in an embodiment of the present invention.
Detailed Description
Preferred embodiments of the present application are described in detail below with reference to the attached drawing figures, which form a part of the present application and are used in conjunction with embodiments of the present application to illustrate the principles of the present application.
The embodiment of the invention discloses a ground robot path planning method based on air-ground coordination, which comprises the following steps as shown in fig. 1:
Step S101, constructing a complete space three-dimensional model of a target area based on space-based image acquisition data of the target area;
Step S102, constructing a moving path area of the ground mobile robot based on the space three-dimensional model, and rasterizing the moving path area to obtain a binary grid map of the robot accessible path;
And step S103, a mixed map combined with the Voronoi map is established based on the binary grid map, mixed A algorithm global path planning is conducted on the mixed map, path optimization is conducted based on multi-guide-point traction, and a reliable moving path of the ground robot is obtained.
As shown in fig. 2, step S101 specifically includes:
Step S201, dividing a target area into a plurality of target subareas, acquiring a sequence image of each target subarea by using at least one unmanned aerial vehicle, and transmitting an image shrinkage map to a ground server by using a map transmission device;
Because the perception capability and the maneuverability of the ground robot are generally limited, when the ground robot is in a large environment scene including a large fire rescue in a field environment, if a single or multiple ground robots are used for carrying cameras or laser radars to globally map a field area, a great deal of time is consumed, which is very undesirable in an emergency fire rescue process. While the drone has a wide field of view in the air and fast maneuvering capabilities.
According to the on-site range of the task, the real-time requirement of the map construction and the number of available unmanned aerial vehicles, the flight track of at least one unmanned aerial vehicle for shooting the image of the target area is determined, in order to construct a reliable moving area map for the ground robot to walk as soon as possible, the flight track can be selected as the starting position of the ground robot to the target end position of the target area, the sequence image of the target subarea is efficiently and accurately acquired during flight, and then the image shrinkage map is transmitted to the ground server through the map transmission device.
Step S202, in a ground server, carrying out real-time three-dimensional reconstruction to recover dense three-dimensional point clouds on the surface of the target sub-region;
in the embodiment, a mature application real-time three-dimensional reconstruction technology can be adopted, so that real-time three-dimensional reconstruction of the unmanned aerial vehicle on the task site becomes practical.
Step S203, a single unmanned mechanism diagram is integrally fused by utilizing a multi-map splicing fusion algorithm based on three-dimensional point cloud data, so that a complete space three-dimensional model of a target area is efficiently constructed;
According to the three-dimensional point cloud images of the sub-target areas obtained by each unmanned aerial vehicle, extracting edge pixel points of a point cloud map to be spliced, establishing a pixel point transformation mathematical model, converting a map splicing problem into data splicing and registration of the three-dimensional point cloud, and obtaining a translational rotation matrix of the pixel points; and according to the pixel point transformation result, carrying out corresponding rotation, translation and fusion on the map to be spliced, and obtaining the accurate splicing result of the complete space three-dimensional model of the target area.
Specifically, in step S102, a depth learning method based on a convolutional neural network prediction remote sensing image is adopted, a road area in the space three-dimensional model scene is extracted, a reliable moving path area of a ground mobile robot is constructed, and a binary grid map of an accessible path is obtained by performing gridding interpolation;
in the step S102, the binary grid map of the robot-accessible path includes:
Step S301, manufacturing a point cloud data training set;
Marking the characteristics including buildings, vegetation, ground, vehicles, pedestrians, lakes, wall surfaces and roadbed according to the three-dimensional model of the live-action, and constructing a data training set;
Step S302, establishing a deep semantic segmentation network model for extracting a reliable moving path region;
in the present embodiment, an image semantic segmentation model of a coder-decoder model of a classical structure is adopted as a depth semantic segmentation network model for reliable movement path region extraction.
Step S303, training and testing the deep semantic segmentation network model by utilizing the data training set;
And training the established depth semantic segmentation network model by using the training data set, inputting test data of the constructed complete target area three-dimensional model into the trained depth semantic segmentation network model, and obtaining an extraction result of a reliable moving path area of the remote sensing image.
Step S304, extracting a complete space three-dimensional model of the target area by utilizing the trained depth semantic segmentation network model to obtain a reliable moving path area of the robot in the target area;
Step S305, according to the reliable moving path area of the robot, rasterizing interpolation is carried out on the reliable moving path area to obtain a binary raster map of the accessible path.
The rasterization interpolation can be performed by adopting the existing map rasterization method.
As shown in FIG. 3, a flow chart of a training and testing process of the deep semantic segmentation network model is shown;
As shown in fig. 4, in step S103, specifically, the method includes:
S401, establishing a Voronoi map by utilizing a binary grid map, and combining the Voronoi map and the binary grid map to construct a hybrid map;
One disadvantage of the conventional path planning algorithm is that it plans a path too close to an obstacle, i.e. it chooses the shortest path without collision. In order to trade-off the contradiction between shortest path and distance from the obstacle, one common approach is to use a potential field to penalize the robot approaching the obstacle, however, there are several serious drawbacks to the conventional potential field.
First, the conventional potential field tends to create high potential regions in narrow channels, which makes the path through these channels computationally expensive.
Second, since the active potential field around an obstacle is typically defined as a function of distance from the obstacle, this means that the potential energy of all obstacles contained within the effective radius needs to be calculated when calculating a given potential field value, which can be computationally expensive.
To solve these problems, the present embodiment introduces Voronoi fields, readjusting the potential field distribution according to the geometry of the binary rasterized map.
Specifically, in a binary grid map, a certain distance threshold from an obstacle is selected by utilizing a Voronoi potential field function, and a two-dimensional Voronoi map of the field space is constructed;
The Voronoi field is used to define trade-off relationships between path length and approaching obstacles. The Voronoi potential field function is defined as follows:
Where d O and d V are the distance to the nearest obstacle and the distance to the edge of the nearest generalized Thiessen polygon (GVD), a ε [0, +#), The rate of decrease of the potential energy value and the range of control are controlled separately for the parameters to be adjusted. When (when)The time expression (1) holds; in other cases, ρ V (x, y) =0.
The potential energy field has the following properties:
(i) When (when) The potential energy is 0.
(Ii) The potential energy value ρ V (x, y) e [0,1] and the distribution is continuous.
(Iii) The potential energy value reaches a maximum value when a point (x, y) on the map is on or in an obstacle.
(Iv) When a point (x, y) on the map is on the side of the generalized voronoi diagram, its potential energy value reaches a minimum.
A key advantage of Voronoi fields over traditional potential fields is that the value of the field can be scaled in proportion to the navigated potential field. Thus, even a narrow U-shaped mouth can be navigated, which is not always the case for a conventional potential field. In addition, as the potential field at the middle position of the channel is zero, a reliable and safe path can be provided for the formation running of multiple robots.
Preferably, the Voronoi diagram is generated based on ROS (robot operating system) using the constructed Voronoi potential field function.
And superposing the Voronoi map and the binary grid map and combining to construct the mixed map.
And step S402, adopting mixed A search in a mixed map to search out a rough path based on the kinematics of the ground robot and the minimum safety steering constraint, and meeting the continuity of the path.
The heuristic search algorithm A is more general, but takes the defect that the vehicle kinematic constraint of the fire-fighting robot has discreteness with the traditional A into consideration, adopts a variant form of the A algorithm, and adopts the mixed A algorithm as a global path search technology.
Compared with the traditional A-algorithm comprising a two-dimensional search space of x-direction and y-direction position information, the spatial dimension searched by the hybrid A-algorithm is four-dimensional, wherein the current orientation information representing the robot is added, and the fourth dimension represents the mode of forward and backward movement of the robot, and the forward movement or the backward movement of the robot is added.
The conventional a-algorithm cannot directly execute the planned path by the robot due to the state discrete feature. To overcome this problem, the hybrid a is assigned to the corresponding discrete grid with the states of the continuous robots, so that the continuous coordinates are linked to be performed by the actual robot.
The mixed A algorithm is a search algorithm based on a heuristic function, path searching is carried out from a starting point to a target point in an extended node mode, and the total cost of each node in the searching process is measured by using an evaluation function. The evaluation function is f(s) =g(s) +h(s), where g(s) is expressed as the actual cost from the starting node to the current node, satisfying the following recurrence relation:
g(si)＝g(si-1)+cost(sk-1,sk)
＝g(si-1)+(1+a·turncost(sk-1,sk)+b·mapcost(sk-1,sk))·dist(sk-1,sk)
where s i is the current node (child node), and s i-1 is the previous node (parent node); g (s i) is the actual cost of the current node, g (s i-1) is the actual cost of the parent node of the current node; cost (s k-1,sk) is the cost from the parent node to the current node; a and b are weight factors; dist (s k-1,sk) is the path length from the parent node to the current node; turncost (s k-1,sk) is a steering cost from the parent node to the current node, and the value is 0 when the vehicle is traveling straight and 1 when the vehicle is steering. The function of the robot is to punish the planned path turning, so that the robot can keep straight as much as possible; mapcost (s k-1,sk) is the cost of the Voronoi diagram from the parent node to the current node, which is infinity within the obstacle safe distance, otherwise 0. The function of the method is to enable the planned path to avoid the obstacle and ensure the driving safety of the robot.
The second term h(s) represents the heuristic estimated cost from the current node to the target node, is determined by a larger value of a heuristic function based on incomplete constraint and a heuristic function based on complete constraint, and has the expression of h (s i)＝max h1(si),h2(si), wherein h 1 is a heuristic value obtained under the condition of ignoring environmental obstacle and considering the incomplete constraint of the robot, and the path length obtained by Reeds-Shepp or Dubin is generally selected; h 2 is a heuristic value obtained under the condition of neglecting the non-integrity constraint of the robot and considering the environment constraint, and the path length obtained by searching through an A-algorithm is generally selected.
Specifically, the basic flow of the hybrid a algorithm is shown in fig. 5: the method comprises the following steps:
Step S501, inputting a mixing map;
Step S502, initializing;
The initialization comprises the steps of establishing two lists of an open_set and a closed_set, wherein the two lists are empty lists when being initialized, the open_set list stores nodes to be expanded, and the closed_set list stores expanded nodes or nodes with barriers. Acquiring a start point s 0(x0,y0,θ0) and a target point s g(xg,yg,θg), and a kinematic model of the differential mobile robot. Putting the starting point into an OpenSet list;
step S503, judging that the OpenSet list is a non-empty list, selecting a node with the smallest total cost from the OpenSet list as an expansion node to expand, obtaining a plurality of expansion sub-nodes continuous with the state of the expansion node, and removing the expansion node from the OpenSet list to the Closed-Set list;
When the open_set list is searched as an empty list, the path search fails, and the map does not have a path which can reach the target position. If the node is not the empty list, selecting a node s i(xi,yi,θi with the smallest total cost from the OpenSet list) as an expansion node, expanding 3 child nodes which are continuous with the node state, and moving the node s i from the OpenSet list to the Closed Set list; x i,yi is a coordinate value, and θ i is a direction angle.
Preferably, the expansion direction is three directions of left turn, right turn and straight run of the node in the forward movement direction, and three continuous child nodes of expansion are shown in fig. 6. For a linear node, the robot moves along a linear motion direction; for the turning nodes, the robot moves along the arc movement direction, and the arc length of the minimum turning radius of the robot can be slightly larger during expansion.
Step S504, judging whether the extended node S i is a target point S g; if yes, the search is ended; if not, continuing to expand the child node s i+1 of the node s i;
Step S505, judging whether the expansion child node S i+1 collides with the obstacle (whether the expansion child node S i+1 is in the obstacle or not), if yes, discarding the child node; if not, jumping to the next step;
Step S506, comparing the actual cost of the current expansion child node S i+1 appearing in the same grid with the actual cost of the original node S open, when the actual cost g (S i+1) of the current expansion child node is lower than the actual cost g (S open) of the original node, associating the continuous state of the current expansion child node with the grid unit, setting the grid where the current expansion child node S i+1 is positioned as the parent grid of the grid where the child node is positioned, putting the expansion child node into an OpenSet list, and updating the total cost f (S i+1) of the expansion child node; otherwise, the child node currently searched is discarded.
Step S507, the steps S503 to S506 in the above process are circulated until the target node is searched, and the target node is put into an OpenSet list;
step S508, returning all traversed father nodes from the target node in the OpenSet list until the starting node, obtaining the searched path, and ending the algorithm after the path searching is finished; if the open_set list is empty, it indicates that no feasible path has been planned and the algorithm terminates.
Fig. 7 shows a search example of the hybrid a-x algorithm, taking the extended 3 forward nodes as an example.
Step S403, optimizing the mixed-search coarse path based on a track smoother of a gradient descent optimization algorithm according to the coarse path, and constructing a path which is suitable for safe execution and continuous smoothing of the ground multi-robot.
Although the above-mentioned hybrid a algorithm can output a continuous path for the robot to execute, the robot usually includes many unnecessary steering actions during execution, and in addition, in a scene of using the robot to perform fire extinguishing tasks by using the robot with a towing water belt, the problem of winding of the robot water belt caused by obstacles in the environment needs to be solved.
Therefore, in the present embodiment, a path smoother based on the action of multiple guide points is disclosed, and a path with better safety, reliability and better processing is obtained.
The gradient descent smoother establishes a minimized cost function based on the obstacle term, the curvature term, the smoothing term, the voronoi field term and the multi-guidance point function term, and solves an optimal path by using a gradient descent method.
Specifically, the minimization cost function P of the gradient descent smoother:
P＝Pobs+Pcur+Psmo+Pvor+Pgui (2)
(1) Barrier item
When there is |x i-oi|≤dmax for each search node x i, the cost function P obs is defined as:
Wherein N is the total node number; x i is the two-dimensional plane coordinates of the vertex on the path; o i is the position of the nearest obstacle to node x i; d max is the threshold of the maximum distance of the obstacle affecting the cost function; sigma obs is a quadratic penalty function, so that the penalty value is larger as the node is closer to the obstacle; the obstacle weight w obs is a coefficient by which the obstacle affects the path change. This allows the robot to effectively avoid collisions with obstacles.
(2) Curvature term
To ensure path executability, the curvature term sets an upper limit on the instantaneous curvature of each vertex, i.e., assumingDefining the cost function P cur as:
Where the displacement vector at vertex x i is defined as Deltax i＝xi-xi-1 and the change in the tangential angle of the vertex is expressed as The maximum allowable curvature is denoted by κ max; σ cur is a quadratic penalty function; the curvature weight w cur controls the effect on the path change. This term limits the instantaneous curvature of the trajectory at each node and enforces incomplete constraints of the vehicle.
(3) Smoothing items
The smoothing term evaluates the displacement vector between vertices. This term can smooth out nodes with uneven spacing and large direction change. Defining the cost function P smo as:
Where w smo denotes a smoothing weight, and determines the influence of the smoothing term on the path change.
(4) Veno potential field function term
For the followingDefine the cost function P vor as
Wherein dO denotes the distance the node has to reach the nearest obstacle; d edg denotes the distance that the node reaches the nearest edge of the generalized Thiessen polygon; Representing the maximum distance of the obstacle affecting the Voronoi potential field function; alpha controls the decay rate of the field; w vor is a weight coefficient representing the effect on the path. This effectively guides the path away from obstacles of narrow or wide passages. The closer to the obstacle, the greater the value of P vor, the closer to 1; the closer to the voronoi edge, the closer to 0 the value of P vor.
(5) Multi-guide point function term
When there is |x i-gi|＜ρmax for each search node x i, the cost function P gui is defined as:
Wherein x i is the two-dimensional plane coordinate of the vertex on the path; g i is the position of the closest guidance point to node x i; ρ max is the threshold for the maximum distance of the guide point affecting the cost function; the guidance point weight w gui is a coefficient by which the guidance point affects the path change. The value of P gui is greater as the path search node is closer to the boundary of the guide point distance ρ max; the closer to the boot point, the smaller the value of P gui. This term may be quickly guided by the path node to the vicinity of the guidance point.
After the cost function P is determined, an optimal path is solved by using a gradient descent method, in the practical application of a gradient descent algorithm, the absolute value of the gradient is generally selected as the standard of algorithm stopping, and the continuity of robot motion is ensured by limiting the maximum iteration times.
In summary, according to the ground robot path planning method based on space-ground coordination provided by the embodiment of the invention, a more efficient and accurate space three-dimensional model in a large-scale environment is established through space-based image acquisition data; the binary grid map is extracted through the depth semantic segmentation network model, and the path planning is performed by adopting an improved mixed A-based algorithm based on the binary grid map, so that the path planned by the robot can be pulled to the vicinity of a plurality of set guide points, and a safe and reliable path can be successfully planned in a multi-obstacle environment.
The embodiment of the invention can be suitable for planning the overall path of the multi-robot formation, and the planned path is close to the middle position of the road, so that the requirement of the overall width of the formation can be met, and the safety of the formation movement is ensured.
The positions and the directions of all the path points can be obtained based on a three-dimensional path planning algorithm, so that the method is more convenient in designing a path or track tracking controller and more accords with the motion control of a robot; the robot is not only suitable for intelligent automobiles, but also suitable for other robots with motion and steering constraint, avoids unnecessary steering motion, and can reach a designated position in an accurate direction.
The method is suitable for the field of task emergency, such as fire rescue, global information is provided for the ground robot by utilizing the rapid maneuvering capability of the unmanned aerial vehicle, a three-dimensional space model of a high-precision site is rapidly constructed, reliable moving path areas including areas of roads, obstacles and the like are further extracted, and a reference map is provided for path planning of the ground robot.
The present invention is not limited to the above-mentioned embodiments, and any changes or substitutions that can be easily understood by those skilled in the art within the technical scope of the present invention are intended to be included in the scope of the present invention.
Claims (6)
1. A ground robot path planning method based on air-ground coordination is characterized by comprising the following steps:
Step S101, constructing a complete space three-dimensional model of a target area based on space-based image acquisition data of the target area;
Step S102, constructing a moving path area of the ground mobile robot based on the space three-dimensional model, and rasterizing the moving path area to obtain a binary grid map of the robot accessible path;
Step S103, a mixed map combined with the Voronoi map is established based on the binary grid map, mixed A algorithm global path planning is conducted on the mixed map, path optimization is conducted based on multi-guide-point traction, and a reliable moving path of the ground robot is obtained;
the step S103 includes:
S401, establishing a Voronoi map by utilizing a binary grid map, and combining the Voronoi map and the binary grid map to construct a hybrid map;
Step S402, adopting mixed A search in a mixed map to search out a rough path based on the kinematics of the ground robot and the minimum safety steering constraint, and meeting the continuity of the path;
Step S403, optimizing the mixed-search coarse path based on a track smoother of a gradient descent optimization algorithm according to the coarse path, and constructing a continuous smooth path suitable for safe execution of multiple robots on the ground;
The track smoother establishes a minimized cost function based on the obstacle items, the curvature items, the smoothing items, the voronoi field items and the multi-guide point function items, and solves an optimal path by using a gradient descent method;
Specifically, the minimization cost function P of the trajectory smoother:
P＝Pobs+Pcur+Psmo+Pvor+Pgui；
(1) Barrier item
When there is |x i-oi|≤dmax for each search node, the cost function P obs is defined as:
Wherein N is the total node number; x i is the two-dimensional plane coordinates of the vertex on the path; o i is the position of the obstacle nearest to vertex x i; d max is the threshold of the maximum distance of the obstacle affecting the cost function; sigma obs is a quadratic penalty function, so that the penalty value is larger as the node is closer to the obstacle; the obstacle weight w obs is a coefficient of the obstacle influencing the path change;
(2) Curvature term
The curvature term sets the upper limit of the instantaneous curvature of each vertex, i.e. assumingDefining the cost function P cur as:
Wherein the displacement vector at the vertex x i is defined as Deltax i＝xi-xi-1, the change value of the tangential angle of the vertex is expressed as Deltaphi i, and the maximum allowable curvature is expressed as kappa max; σ cur is a quadratic penalty function; the curvature weight w cur controls the effect on the path change;
(3) Smoothing items
The smoothing term evaluates displacement vectors between vertices; the method can smooth nodes with uneven intervals and large direction change amplitude; defining the cost function P smo as:
wherein w smo represents a smoothing weight;
(4) Veno potential field function term
For the followingDefining the cost function P vor as:
Where d O represents the distance of the node to the nearest obstacle; d edg denotes the distance that the node reaches the nearest edge of the generalized Thiessen polygon; Representing the maximum distance of the obstacle affecting the Voronoi potential field function; alpha is the attenuation rate of the control field; w vor is a weight coefficient;
(5) Multi-guide point function term
When there is |x i-gi|＜ρmax for each search node, the cost function P gui is defined as:
Where g i is the position of the nearest guidance point from vertex x i; ρ max is the threshold for the maximum distance of the guide point affecting the cost function; the guidance point weight w gui is a coefficient by which the guidance point affects the path change.
2. The method according to claim 1, wherein in step S402, the mixed a search specifically includes:
Step S501, inputting a mixing map;
Step S502, initializing search;
the initialization comprises the steps of establishing an open_set list and a closed_set list, wherein the two lists are empty lists when being initialized, the open_set list stores nodes to be expanded, and the closed_set list stores expanded nodes or nodes with barriers;
step S503, judging that the OpenSet list is a non-empty list, selecting a node with the smallest total cost from the OpenSet list as an expansion node to expand, obtaining a plurality of expansion sub-nodes continuous with the state of the expansion node, and removing the expansion node from the OpenSet list to the Closed-Set list;
step S504, judging whether the extension node is a target point; if yes, the search is ended; if not, continuing to expand the child node of the node;
Step S505, judging whether the expansion child node collides with the obstacle, if so, discarding the child node; if not, jumping to the next step;
Step S506, comparing the actual cost of the current expansion child node with the actual cost of the original node, when the actual cost of the current expansion child node is lower than the actual cost of the original node, associating the continuous state of the current expansion child node with the grid unit, setting the grid of the current expansion child node as the father grid of the child node, putting the expansion child node into an OpenSet list, and updating the total cost of the expansion child node;
Step S507, the steps S503 to S506 in the above process are circulated until the target node is searched, and the target node is put into an OpenSet list;
step S508, returning all traversed father nodes from the target node in the OpenSet list until the starting node, obtaining the searched path, and ending the algorithm after the path searching is finished; if the open_set list is empty, it indicates that no feasible path has been planned and the algorithm terminates.
3. The method for planning a path of a floor robot according to claim 1, wherein,
The total cost of each node in the searching process comprises actual cost and heuristic estimated cost;
The actual cost is the cost from the initial node to the current node, and comprises the cost from the parent node to the current node and the cost from the parent node to the current node, wherein the actual path length from the parent node to the current node, the steering cost from the parent node to the current node in the binary grid map and the cost from the parent node to the current node in the Voronoi map are included in the latter;
The heuristic estimated cost is the heuristic estimated cost from the current node to the target node and is determined by a larger value in a heuristic function based on incomplete constraint and a heuristic function based on complete constraint.
4. The method for planning a path of a floor robot according to claim 1, wherein,
The step S101 includes:
Step S201, dividing a target area into a plurality of target subareas, acquiring a sequence image of the target subareas by using at least one unmanned aerial vehicle, and transmitting an image shrinkage map to a ground server by using a map transmission device;
step S202, in a ground server, carrying out real-time three-dimensional reconstruction to recover dense three-dimensional point clouds on the surface of the target sub-region;
and step 203, integrally fusing the single unmanned aerial vehicle mechanism images by utilizing a multi-map splicing and fusing algorithm based on three-dimensional point cloud data, and constructing a complete space three-dimensional model of the target area.
5. The method for planning a path of a floor robot according to claim 1, wherein,
In step S102, a depth learning method based on a convolutional neural network prediction remote sensing image is adopted, a road surface area in the space three-dimensional model scene is extracted, a reliable moving path area of the ground mobile robot is constructed, and a binary grid map of the accessible path is obtained by performing grid interpolation.
6. The method for planning a path of a floor robot according to claim 5, wherein,
The step S102 specifically includes:
Step S301, manufacturing a point cloud data training set;
Marking the characteristics including buildings, vegetation, ground, vehicles, pedestrians, lakes, wall surfaces and roadbed according to the three-dimensional model of the live-action, and constructing a data training set;
Step S302, establishing a deep semantic segmentation network model for extracting a reliable moving path region;
step S303, training and testing the deep semantic segmentation network model by utilizing the data training set;
Step S304, extracting a complete space three-dimensional model of the target area by utilizing the trained depth semantic segmentation network model to obtain a reliable moving path area of the robot in the target area;
Step S305, according to the reliable moving path area of the robot, rasterizing interpolation is carried out on the reliable moving path area to obtain a binary raster map of the accessible path.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202211108617.6A CN115373399B (en) | 2022-09-13 | 2022-09-13 | Ground robot path planning method based on air-ground coordination |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202211108617.6A CN115373399B (en) | 2022-09-13 | 2022-09-13 | Ground robot path planning method based on air-ground coordination |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN115373399A CN115373399A (en) | 2022-11-22 |
| CN115373399B true CN115373399B (en) | 2024-07-12 |
Family
ID=84072481
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202211108617.6A Active CN115373399B (en) | 2022-09-13 | 2022-09-13 | Ground robot path planning method based on air-ground coordination |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN115373399B (en) |
Families Citing this family (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN115562255B (en) * | 2022-09-13 | 2024-07-19 | 中国安全生产科学研究院 | Multi-intelligent fire-fighting robot fire hose anti-winding method based on air-ground coordination |
| CN115930969B (en) * | 2023-01-09 | 2023-06-02 | 季华实验室 | Path planning method and device for mobile robot, electronic equipment and storage medium |
| CN116311095B (en) * | 2023-03-16 | 2024-01-02 | 广州市衡正工程质量检测有限公司 | Pavement detection method based on region division, computer equipment and storage medium |
| CN116772880B (en) * | 2023-06-07 | 2024-01-23 | 中国人民解放军陆军装甲兵学院 | Unmanned aerial vehicle path planning method based on unmanned aerial vehicle vision |
| CN116433865B (en) * | 2023-06-15 | 2023-10-20 | 深圳大学 | Space-ground collaborative acquisition path planning method based on scene reconstructability analysis |
| CN116909268B (en) * | 2023-06-30 | 2024-05-28 | 广东省机场管理集团有限公司工程建设指挥部 | 5G-based path planning method, device, equipment and medium for walking robot |
| CN117685994B (en) * | 2024-02-04 | 2024-05-17 | 北京航空航天大学 | Unmanned vehicle path planning method for air-ground coordination |
| CN117930855A (en) * | 2024-03-25 | 2024-04-26 | 山东工程职业技术大学 | Robot inspection method and system for construction site |
Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106647769A (en) * | 2017-01-19 | 2017-05-10 | 厦门大学 | AGV path tracking and obstacle avoiding coordination method based on A* extraction guide point |
| CN114237232A (en) * | 2021-11-30 | 2022-03-25 | 武汉地大信息工程股份有限公司 | Path planning method and device |
Family Cites Families (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108898605B (en) * | 2018-07-25 | 2021-01-05 | 电子科技大学 | Grid map segmentation method based on map |
| CN111323037B (en) * | 2020-02-28 | 2022-07-05 | 海博(苏州)机器人科技有限公司 | Voronoi path planning algorithm for novel framework extraction of mobile robot |
| CN111998859B (en) * | 2020-09-16 | 2022-08-05 | 杭州海康消防科技有限公司 | Path planning method and device, electronic equipment and path indicating system |
| CN113504793B (en) * | 2021-07-12 | 2022-07-15 | 哈尔滨工程大学 | Unmanned aerial vehicle cluster path planning method based on Voronoi diagram |
| CN113759971B (en) * | 2021-08-30 | 2023-11-07 | 中国人民解放军国防科技大学 | Unmanned plane collaborative reconnaissance-oriented path planning method |
| CN115032999B (en) * | 2022-06-24 | 2024-09-13 | 中国安全生产科学研究院 | Dynamic optimization formation transformation method for multi-robot formation |
-
2022
- 2022-09-13 CN CN202211108617.6A patent/CN115373399B/en active Active
Patent Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106647769A (en) * | 2017-01-19 | 2017-05-10 | 厦门大学 | AGV path tracking and obstacle avoiding coordination method based on A* extraction guide point |
| CN114237232A (en) * | 2021-11-30 | 2022-03-25 | 武汉地大信息工程股份有限公司 | Path planning method and device |
Also Published As
| Publication number | Publication date |
|---|---|
| CN115373399A (en) | 2022-11-22 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN115373399B (en) | Ground robot path planning method based on air-ground coordination | |
| CN110531770B (en) | RRT path planning method and system based on improvement | |
| Reinhart et al. | Learning-based path planning for autonomous exploration of subterranean environments | |
| Zhang et al. | Maximum likelihood path planning for fast aerial maneuvers and collision avoidance | |
| Yang et al. | Graph-based topological exploration planning in large-scale 3d environments | |
| Garcia et al. | High-speed vision-based autonomous indoor navigation of a quadcopter | |
| Zhang et al. | P-CAP: Pre-computed alternative paths to enable aggressive aerial maneuvers in cluttered environments | |
| CN117406771B (en) | Efficient autonomous exploration method, system and equipment based on four-rotor unmanned aerial vehicle | |
| CN116795108B (en) | Intelligent unmanned vehicle distribution method based on multi-source sensing signals | |
| Pritzl et al. | Cooperative navigation and guidance of a micro-scale aerial vehicle by an accompanying UAV using 3D LiDAR relative localization | |
| Palazzolo et al. | Information-driven autonomous exploration for a vision-based mav | |
| Yang et al. | Path planning and collision avoidance methods for distributed multi-robot systems in complex dynamic environments | |
| Petit et al. | Tape: Tether-aware path planning for autonomous exploration of unknown 3d cavities using a tangle-compatible tethered aerial robot | |
| Lu et al. | Flight in GPS-denied environment: Autonomous navigation system for micro-aerial vehicle | |
| Zhang et al. | Collision Avoidance of Fixed-Wing UAVs in Dynamic Environments Based on Spline-RRT and Velocity Obstacle | |
| Guan et al. | Autonomous exploration using UWB and LiDAR | |
| Harun et al. | Collision avoidance control for Unmanned Autonomous Vehicles (UAV): Recent advancements and future prospects | |
| Saska et al. | Roads sweeping by unmanned multi-vehicle formations | |
| CN117889882A (en) | Road network planning method based on ground-air unmanned cluster collaborative situation assessment | |
| Berry et al. | A continuous local motion planning framework for unmanned vehicles in complex environments | |
| CN117036374A (en) | Laser radar point cloud segmentation and motion planning method for automatic driving | |
| CN115562255B (en) | Multi-intelligent fire-fighting robot fire hose anti-winding method based on air-ground coordination | |
| Adolf | Multi-query path planning for exploration tasks with an unmanned rotorcraft | |
| Meng et al. | A navigation framework for mobile robots with 3D LiDAR and monocular camera | |
| Rudnick-Cohen et al. | Planning unmanned aerial system (uas) takeoff trajectories to minimize third-party risk |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |