CN115437375A - Three-dimensional path planning method based on big data platform distributed tile pyramid - Google Patents
Three-dimensional path planning method based on big data platform distributed tile pyramid Download PDFInfo
- Publication number
- CN115437375A CN115437375A CN202211033770.7A CN202211033770A CN115437375A CN 115437375 A CN115437375 A CN 115437375A CN 202211033770 A CN202211033770 A CN 202211033770A CN 115437375 A CN115437375 A CN 115437375A
- Authority
- CN
- China
- Prior art keywords
- tile
- data
- layer
- distributed
- pyramid
- 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.)
- Granted
Links
- 238000000034 method Methods 0.000 title claims abstract description 33
- 238000012545 processing Methods 0.000 claims abstract description 14
- 230000008569 process Effects 0.000 claims description 5
- 238000012216 screening Methods 0.000 claims description 4
- 239000004744 fabric Substances 0.000 claims description 2
- 238000005070 sampling Methods 0.000 claims description 2
- 238000011439 discrete element method Methods 0.000 claims 9
- 238000003672 processing method Methods 0.000 claims 1
- 239000010410 layer Substances 0.000 description 74
- 208000022417 sinus histiocytosis with massive lymphadenopathy Diseases 0.000 description 22
- 238000010586 diagram Methods 0.000 description 8
- 238000004364 calculation method Methods 0.000 description 6
- PEDCQBHIVMGVHV-UHFFFAOYSA-N Glycerine Chemical compound OCC(O)CO PEDCQBHIVMGVHV-UHFFFAOYSA-N 0.000 description 3
- 238000006243 chemical reaction Methods 0.000 description 3
- 238000011160 research Methods 0.000 description 2
- 238000012360 testing method Methods 0.000 description 2
- 241000282414 Homo sapiens Species 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 230000008859 change Effects 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 230000000694 effects Effects 0.000 description 1
- 238000005516 engineering process Methods 0.000 description 1
- 238000001914 filtration Methods 0.000 description 1
- 230000006870 function Effects 0.000 description 1
- 239000011229 interlayer Substances 0.000 description 1
- 238000012804 iterative process Methods 0.000 description 1
- 238000013507 mapping Methods 0.000 description 1
- 230000007246 mechanism Effects 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000006467 substitution reaction Methods 0.000 description 1
Images
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
- G05D1/0246—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
- G05D1/0212—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
- G05D1/0212—Control of position or course in two dimensions specially adapted to land vehicles with means for defining a desired trajectory
- G05D1/0223—Control of position or course in two dimensions specially adapted to land vehicles with means for defining a desired trajectory involving speed control of 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
- G05D1/0276—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
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/10—File systems; File servers
- G06F16/18—File system types
- G06F16/182—Distributed file systems
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T17/00—Three dimensional [3D] modelling, e.g. data description of 3D objects
- G06T17/05—Geographic models
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Remote Sensing (AREA)
- Automation & Control Theory (AREA)
- Radar, Positioning & Navigation (AREA)
- Aviation & Aerospace Engineering (AREA)
- Theoretical Computer Science (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Geometry (AREA)
- Software Systems (AREA)
- General Engineering & Computer Science (AREA)
- Multimedia (AREA)
- Electromagnetism (AREA)
- Computer Graphics (AREA)
- Databases & Information Systems (AREA)
- Data Mining & Analysis (AREA)
- Processing Or Creating Images (AREA)
- Image Generation (AREA)
Abstract
The invention discloses a three-dimensional path planning method based on a large data platform distributed tile pyramid, which comprises the steps of carrying out distributed tile pyramid processing on large three-dimensional terrain data DEM of an area to be explored, and storing a generated distributed tile pyramid into a Hadoop Distributed File System (HDFS); reading DEM data of the distributed tile pyramid from top to bottom in the HDFS by using a Spark distributed computing cluster, and planning a three-dimensional path of an area to be explored by adopting a path planning method from coarse granularity to fine granularity in an iterative mode. According to the method, the path searching efficiency of the distributed tile pyramid is improved, the running speed of a long-distance path planning task of large DEM data is increased, and meanwhile, path planning from coarse granularity to fine granularity is adopted, so that the path planning running time is reduced.
Description
Technical Field
The invention belongs to the technical field of distributed computing, and particularly relates to a three-dimensional path planning method based on a large data platform distributed tile pyramid.
Background
The exploration of the moon by the lunar vehicle is the first step of exploring space by human beings, and the problem of path planning of the lunar vehicle is a research hotspot of lunar exploration engineering all the time. Due to the rapid development of modern mapping technology and sensor hardware, the generated DEM has higher and higher precision and larger data volume, so that the time for performing path-finding calculation on large three-dimensional terrain Data (DEM) by using a traditional strategy is longer and longer.
The core idea of path planning is to find an optimal path from a current starting position to a target position in an unknown environment. In the problem of path planning for large data, many scholars propose different solutions. These schemes can be summarized as: (1) The efficiency of the path planning algorithm is improved by improving the original path planning algorithm; (2) When DEM data is converted into discrete point data, a MapReduce or Spark is used to run Dijkstra or A star path planning algorithm to solve the shortest path problem.
However, the efficiency of the path planning algorithm is improved by improving the original path planning algorithm, and is difficult to improve on large data, the efficiency is limited by the physical hardware influence of a single server, and the problems of memory overflow, insufficient disk storage space, exponential increase of search time along with the increase of the search data amount are often generated on the calculation of large DEM data; although the influence of physical hardware of a single server is solved by running dijkstra or a star path planning algorithm by using a distributed computing framework such as MapReduce or Spark, simple conversion of large DEM data into discrete point data is a time-consuming operation on the task of path planning of DEM raster data, and a large amount of useless discrete point data is stored in a distributed memory, so that the amount of computation is increased, and the time spent on computation is increased. Moreover, the DEM data is converted into the discrete data only by improving the distributed computing capability of the algorithm, but the data structure characteristic of the DEM is not considered, and the neighborhood characteristic of the raster data is ignored.
Disclosure of Invention
Aiming at the technical problems of time consumption, labor consumption and the like of path planning of the prior DEM raster data, a three-dimensional path planning method based on a large data platform distributed tile pyramid is provided, for a path planning task of large DEM data, the method solves the problems of insufficient disk storage space and memory overflow in a single machine strategy through an HDFS storage frame and a Spark calculation frame built by a cluster, and applies the tile pyramid to the distributed path planning task to solve the problems that the search time in the single machine strategy exponentially increases along with the increase of the search data amount and the calculation time in the distributed strategy increases due to a large amount of useless discrete data.
According to the three-dimensional path planning method based on the large data platform distributed tile pyramid, the large three-dimensional terrain data DEM is converted into the distributed tile pyramid for storage, and through the thought of dividing and solving and coarse-fine granularity conversion, by means of Spark distributed calculation price and Hadoop distributed storage capacity, time spent on path planning tasks on the large DEM data is reduced, and calculation speed is improved.
The invention can be realized by the following technical scheme:
a three-dimensional path planning method based on a large data platform distributed tile pyramid is characterized in that distributed tile pyramid processing is carried out on large three-dimensional terrain data DEM of an area to be explored, and the generated distributed tile pyramid is stored in a Hadoop distributed file system HDFS; and reading DEM data of the distributed tile pyramid from top to bottom in the HDFS by using a Spark distributed computing cluster, and planning a three-dimensional path of the area to be explored by adopting a path planning method iterated from coarse granularity to fine granularity.
Further, according to the information of the starting point and the end point, three-dimensional path node data of a current layer corresponding to the tile data of the current pyramid layer is planned in the distributed computing cluster of the Spark, intersecting tiles in the three-dimensional path node data of the current layer and the tile data of the next pyramid layer are screened, corresponding intersection points are computed, therefore, the three-dimensional path of the current layer is divided into a plurality of local paths, the intersection points respectively correspond to the starting point and the end point of each local path and are marked as the local starting point and the local end point, then local paths corresponding to each local starting point and the local end point in the tile data of the next pyramid layer are planned in an iterative mode, the next three-dimensional path node data of the corresponding starting point and the end point are formed together, the iterative process from planning of the coarse-granularity path to planning of the fine-granularity distributed path is repeated, and the three-dimensional path planning corresponding to the starting point and the end point in the tile data of the bottom layer is completed.
Further, the path planning method from coarse granularity iteration to fine granularity iteration comprises the following steps: the layers of the distributed scoring tile pyramid are respectively the 0 th layer (8230), the ith layer (8230), the nth layer (n),
reading the i = n-th layer DEM data of a distributed tile pyramid from an HDFS, and constructing corresponding tile RDD data in a memory;
step two, planning a corresponding ith layer tile path in the ith layer tile RDD data by adopting a Spark distributed computing architecture according to the starting point information and the end point information;
reading tile data of an i-1 layer pyramid of the distributed tile pyramid from the HDFS, constructing corresponding tile RDD data in a memory, screening tiles intersected in the i-1 layer tile path and the i-1 layer tile data, and calculating corresponding intersection points, wherein the start points and the end points corresponding to all local paths in the i-1 layer tile path are local start point and local end point sets respectively;
step four, planning local paths corresponding to each local starting point and local end point in the i-1 layer tile RDD data by adopting a Spark distributed computing architecture, and forming an i-1 layer three-dimensional path corresponding to the starting point and the end point;
and step five, recording i = i-1, and repeatedly executing the steps three to four until the three-dimensional path planning of the corresponding starting point and the end point in the i =0 layer of tile RDD data is completed.
Furthermore, when the data of the distributed tile pyramid of the DEM is read from the HDFS, a layer skipping reading mode is adopted.
Further, the large three-dimensional terrain data DEM of the area to be explored is subjected to distributed tile pyramid processing, the processing mode is that a series of up-sampling is carried out on the original large three-dimensional terrain data DEM, the obtained image information and the original large three-dimensional terrain data DEM jointly generate a pyramid, each layer of data of the pyramid is cut into a plurality of tiles with the same size and rectangular shape, so that a distributed tile pyramid is obtained, finally the generated distributed tile pyramid is stored in a Hadoop distributed file system HDFS, the storage mode is that tile information of each layer is serialized into tile byte blocks and tile metadata corresponding to the pyramid layer, and the tile byte blocks and the tile metadata corresponding to the pyramid layer are stored in data nodes of the Hadoop distributed file system HDFS.
The beneficial technical effects of the invention are as follows:
(1) The DEM data is stored in the Hadoop distributed file system by utilizing the storage characteristics of the distributed server and the model characteristics of the tile pyramid through the storage model of the distributed tile pyramid, and the storage efficiency in the distributed cluster is improved.
(2) And establishing a flow framework of a Spark distributed processing tile pyramid, and switching from a coarse-grained tile path planning task to a fine-grained distributed tile path planning task in an iterative manner, so that the time spent on performing the path planning task on large DEM data is reduced.
(3) In the iteration process of the distributed path planning task, a layer-skipping processing mode is added to the distributed reading of the tile pyramid, the iteration times of the cluster operation task are reduced, and therefore the path planning operation time is reduced.
Drawings
FIG. 1 is a schematic structural diagram of a distributed tile pyramid of the DEM of the present invention;
FIG. 2 is a schematic diagram of a process of generating a distributed tile pyramid by DEM data and storing the pyramid in an HDFS according to the present invention;
FIG. 3 is a schematic diagram of a three-dimensional path planning process for realizing a distributed tile pyramid DEM by Spark according to the invention;
FIG. 4 is a schematic diagram of the three-dimensional path planning principle based on the distributed tile pyramid of the present invention;
fig. 5 is a schematic diagram of a Spark cluster structure and a task distribution method according to the present invention.
FIG. 6 is a schematic diagram of Spark implementation of the three-dimensional path planning method for a distributed tile pyramid of the present invention;
FIG. 7 is a schematic view of the orientation of the current node of the present invention on a grid map.
Detailed Description
The following detailed description of the preferred embodiments of the invention refers to the accompanying drawings.
As shown in fig. 1, the invention provides a three-dimensional path planning method based on a large data platform distributed tile pyramid, which is characterized in that large three-dimensional terrain data DEM of an area to be explored is subjected to distributed tile pyramid processing, and the generated distributed tile pyramid is stored in a Hadoop distributed file system HDFS; reading DEM data of the distributed tile pyramid from top to bottom in the HDFS according to a Spark distributed computing principle, and performing three-dimensional path planning on an area to be explored by adopting a path planning method from coarse granularity to fine granularity in an iterative mode. Therefore, by establishing a storage model of the distributed tile pyramid of the DEM and a flow framework of the Spark distributed processing tile pyramid, the path searching efficiency of the distributed tile pyramid is improved, and the running speed of a remote path planning task of large DEM data is increased. The method comprises the following specific steps:
1. distributed tile pyramid storage of large DEM data
Aiming at the actual storage requirement of the large DEM data, the invention provides that the distributed tile pyramid operation is carried out on the large DEM data, and the generated tile pyramid is stored in a Hadoop distributed file system HDFS.
The distributed tile pyramid is established by performing a series of upsampling on original DEM data to obtain image data, forming the image data and the original DEM data into a pyramid together, and cutting each layer of data of the pyramid into a plurality of rectangular tiles with the same size. As shown in fig. 1 (a), the original DEM data has a pixel size of 2048 × 2048, and is used as the 0 th layer, which is the bottom layer of the tile pyramid, and the tile pyramid is generated by performing a series of upsampling on the pixel data, as shown in fig. 1 (b), the pyramid of each layer is cut into tiles of uniform size, each tile has a tile row number corresponding to the layer, and the tile row number is used for quickly positioning the position of the tile on the pyramid model of the layer.
In order to accelerate the distributed processing of the tile pyramid, the tile pyramid of DEM data is stored by a Hadoop Distributed File System (HDFS), the HDFS is a distributed file system which is designed to be suitable for running on general hardware, the HDFS can provide a high-throughput data access and data redundancy mechanism, and the method is very suitable for being applied to a large-scale data set. The HDFS cluster is a typical master-slave operation mode, distributed storage of the cluster is controlled and managed through a master node and data nodes, the master node is responsible for storing metadata of distributed file blocks, and the metadata is used for positioning storage positions of real data in the data nodes; the data nodes are responsible for storing and redundantly backing up real data. The master node communicates with the data nodes over a network. Fig. 2 is a schematic diagram of generating a distributed tile pyramid according to DEM data and storing the generated tile pyramid in an HDFS, and taking a DEM with an image element size of 2048 × 2048 as an example, this diagram shows how DEM data is converted into a distributed tile pyramid with an image element size of 512 × 512 of tiles and stored in an HDFS, and the steps are as follows:
1. original DEM data with 2048 x 2048 pixel size is subjected to a series of upsampling to obtain image data, and the image data and the original DEM data generate a 3-layer pyramid structure together.
2. Each layer of pyramid is cut into tiles of 512 x 512 pel size, respectively.
3. And serializing the tile information of each layer into tile byte blocks and tile metadata of a corresponding pyramid layer.
4. And storing the tile byte blocks and the tile metadata of the corresponding pyramid layer into the data nodes of the HDFS.
Wherein, the tile byte block is a serialized file of the tile block, and the invention generally uses Z-step filling curve or Hilbert filling curve serialization; the tile metadata includes pyramid level information, tile row and column numbers, and corresponding tile storage nodes and storage location information, that is, storage information corresponding to the current byte block.
2. Three-dimensional path planning based on distributed tile pyramid
Apache Spark is a fast general-purpose computing engine designed specially for large-scale data processing, which ensures high fault tolerance of tasks in a distributed computing environment and reduces time spent by writing tasks into disks by constructing an elastic distributed data set RDD and a directed acyclic graph DAG in a memory, wherein the RDD is a most basic data processing model in Spark and represents an elastic, invariable, partitionable and parallel computing memory set, the DAG is a combination of a group of vertexes and edges, the vertexes are used for representing RDDs, the edges are used for representing operational relationships among the RDDs, and the DAG in Spark is a key for ensuring that distributed computing can execute tasks in sequence. Researches show that Spark is higher than MapReduce in computational efficiency, and the main reason is that operations in various stages of MapReduce are independent, so that the MapReduce operation can be written into a disk every time, and DAG of Spark can judge whether the current operation needs to be written into the disk according to the operation type between RDDs, so that the computational efficiency is improved by reducing the number of times of writing into the disk.
The invention provides a three-dimensional path planning of a distributed tile pyramid, which adopts a path planning method iterated from coarse granularity to fine granularity, a large-scale path planning task is disassembled into distributed partial path planning subtasks which are carried out on tiles, the starting point and the end point of each layer of tiles for carrying out the path planning task are determined by utilizing pyramid characteristics on DEM data, namely, a Spark distributed computing framework is adopted according to the starting point and the end point information, a corresponding current layer three-dimensional path in the tile data of the current pyramid layer is planned, tiles which are intersected in the current layer three-dimensional path and the tile data of the next pyramid layer are screened, corresponding intersection points are computed, the current layer three-dimensional path is divided into a plurality of local paths, the intersection points respectively correspond to the starting point and the end point of each local path and are marked as the local starting point and the local end point, then a Spark distributed computing principle is adopted, local paths which correspond to each local starting point and the local end point in the tile data of the next pyramid layer are planned, a lower layer three-dimensional path corresponding to the starting point and the local end point is jointly formed, and the process is repeated until the three-dimensional paths which correspond to the starting point and the end point in the bottom layer tile data are planned. As shown in fig. 5, the Spark divides the path planning task into a plurality of subtasks to be distributed to the working node for execution through a master-slave mode, and the working node obtains the tile RDD data from the HDFS and executes the subtasks distributed by the driving node, and finally writes a result to the HDFS. As shown in fig. 3, the details are as follows:
the layers of the score cloth tile pyramid are respectively the 0 th layer 8230, the ith layer 8230and the nth layer from the bottom layer to the top layer,
reading i = n-th-layer DEM data of a distributed tile pyramid from an HDFS (Hadoop distributed file system), and constructing corresponding tile RDD data in a memory;
step two, planning a corresponding ith layer tile path in the ith layer tile RDD data by adopting a Spark distributed computing architecture according to the starting point information and the end point information;
reading tile data of an i-1 layer pyramid of the distributed tile pyramid from the HDFS, constructing corresponding tile RDD data in a memory, screening tiles intersected in the i-1 layer tile path and the i-1 layer tile data, and calculating corresponding intersection points, wherein the start points and the end points corresponding to each local path in the i-1 layer tile path are local start point and local end point sets respectively;
fourthly, planning local paths corresponding to each local starting point and local end point in the tile RDD data of the (i-1) th layer by adopting a Spark distributed computing architecture, and forming an (i-1) th layer three-dimensional path corresponding to the starting point and the end point together;
and step five, recording i = i-1, and repeatedly executing the steps three to four until the three-dimensional path planning corresponding to the starting point and the end point in the i =0 layer of tile RDD data is completed.
According to the method, the coarse-grained tile path planning task is switched to the fine-grained distributed tile path planning task in an iteration mode, so that the time spent on the path planning task on a large-scale DEM is reduced. In order to reduce the iteration time for path planning on the distributed tile pyramid in the distributed cluster, layer skipping processing, namely interlayer reading, is adopted in the reading operation from the upper layer data of the tile pyramid to the lower layer data of the tile pyramid, for example, coarse-grained path planning is carried out from the uppermost layer (n = 4) of the tile pyramid, the starting point and the end point of each tile in the n-2 layers of tile pyramids are calculated from the result of the coarse-grained path planning, and then the distributed path planning is carried out on the n-2 layers of tile pyramids until the result of the path planning of the bottom layer of the tile pyramid is calculated.
In the path planning of the distributed tile pyramid realized by Spark, a driving node is responsible for broadcasting a starting point set and an end point set of each tile and controlling the iteration times, and a working node cluster is responsible for a sub-path planning task of each tile. As shown in fig. 6, first, the driver node and the worker node cluster construct tile RDDs and tile metadata of the topmost pyramid (i = n) from the HDFS, and the driver node broadcasts the start point and the end point of the path planning to the worker node cluster. And then filtering the RDD of the tiles which do not participate in the path planning by the working node cluster, and performing sub-path planning on the rest tiles according to the set starting point and the set end point. And finally judging whether the layer is the lowest layer (i = 0) according to the tile metadata. If yes, saving the obtained path node RDD to the HDFS, and if not, performing iterative operation. And the iterative operation is that the driving node collects path nodes RDD obtained from the working node cluster, calculates a starting point set and an end point set of each tile of the i-2 layer, finally informs the HDFS to construct tile RDD and tile metadata of the pyramid of the i-2 layer to Spark and executes distributed path planning of the i-2 layer.
To verify the feasibility of the three-dimensional path planning method of the present invention, we used the ChangE's two-dimensional CCD stereo camera DEM-20m dataset from the ground application system of China moon and deep space exploration project (China national space agency, 2020).http：//moon.bao.ac.cn) And (3) carrying out path planning on the partial DEM data of the distributed tile pyramid, wherein the total size of the selected DEM pixels is 32768 multiplied by 32768, and the resolution is 20m. And (4) setting a feasible slope threshold value of whether the trolley passes through the path planning as 20 degrees by referring to the crawler-type lunar vehicle, and applying the feasible slope threshold value to a DEM raster path planning task. In the test, assuming that the projected area of the vehicle is one grid cell, according to grid cell (i, j) (corresponding to elevation Z) i，j ) And 8 grid cells around, forming a 3 × 3 window, as shown in fig. 7, and using this information, the slope in the vertical/horizontal direction and the slope in the diagonal direction are calculated using equation (1). Wherein CellSize is the size of each grid:
the algorithm for local path planning on each tile of the invention adopts an A star algorithm, and the heuristic function is shown as equation 2:
F(P)＝G(P)+H(P) (2)
where G (P) is the distance cost from node P to the starting point and H (P) is the distance cost from node S to the end point. In this study, the euclidean distance was used to calculate G (P), as shown in equation 3; the Manhattan distance is used to calculate H (P), as shown in equation 4, where G (P ') represents the actual distance from the starting point to P ', G (P) is the actual distance from the starting point to point P through P ', P end The position of the end point is shown, and x, y and z respectively show the horizontal position and the vertical position of the node and the corresponding elevation value.
The table below shows four sets of points required for the following tests, all of which simulated the selection of long, medium, and short paths.
In the A star algorithm, four groups of starting points and end points with different distances are recorded, and the time overhead of path planning in a single machine serial mode and a distributed tile pyramid parallel mode is tested, as shown in the following table:
compared with a single machine serial computing strategy and a distributed tile pyramid parallel computing strategy, under the condition of short-distance path search, the time cost of the distributed tile pyramid parallel computing strategy is larger than that of the single machine serial computing strategy due to the fact that the time of cluster starting and RDD operator conversion exceeds the time of path planning, but under the condition that a path finding path is long, the time cost of the distributed tile pyramid parallel computing strategy is smaller than that of the single machine serial computing strategy.
In the A star algorithm, four groups of starting points and end points with different distances are recorded, and the time expenditure for path planning in a distributed parallel computing mode and a distributed tile pyramid parallel mode is tested, as shown in the following table:
the result shows that in the distributed computing environment, because the neighborhood information of each pixel is reserved in the distributed tile pyramid parallel computing strategy, the time cost for path planning by using the distributed tile pyramid parallel computing strategy is much lower than that for distributed path planning for DEM data by using only a cluster, wherein under the condition of long-distance path search, the speed of using the distributed tile pyramid parallel computing strategy is 113 times that of using only the cluster for distributed path planning for DEM data.
The above examples are only intended to illustrate the technical solution of the present invention, and not to limit it; although the present invention has been described in detail with reference to the foregoing embodiments, it should be understood by those of ordinary skill in the art that: the technical solutions described in the foregoing embodiments may still be modified, or some technical features may be equivalently replaced; and such modifications or substitutions do not depart from the spirit and scope of the corresponding technical solutions of the embodiments of the present invention.
Claims (5)
1. A three-dimensional path planning method based on a large data platform distributed tile pyramid is characterized by comprising the following steps: carrying out distributed tile pyramid processing on large three-dimensional terrain data DEM of an area to be explored, and storing a generated distributed tile pyramid into a Hadoop distributed file system HDFS; reading DEM data of the distributed tile pyramid from top to bottom in the HDFS by using a Spark distributed computing cluster, and planning a three-dimensional path of an area to be explored by adopting a path planning method from coarse granularity to fine granularity in an iterative mode.
2. The big data platform distributed tile pyramid based three-dimensional path planning method according to claim 1, wherein: planning corresponding current-layer three-dimensional path node data in current pyramid-layer tile data in a distributed computing cluster of Spark according to the information of a starting point and an end point, screening tiles intersected in the current-layer three-dimensional path node data and next pyramid-layer tile data, computing corresponding intersection points, dividing the current-layer three-dimensional path into a plurality of local paths, respectively corresponding to the starting point and the end point of each local path, marking as the local starting point and the local end point, then planning local paths corresponding to the local starting points and the local end points in the next pyramid-layer tile data, jointly forming next-layer three-dimensional path node data corresponding to the starting point and the end point, and repeating the iteration process from coarse-granularity path planning to fine-granularity distributed path planning until the three-dimensional path planning corresponding to the starting point and the end point in the bottom-layer tile data is completed.
3. The big data platform distributed tile pyramid-based three-dimensional path planning method of claim 2, wherein: the path planning method from coarse granularity iteration to fine granularity iteration comprises the following steps: the layers of the score cloth tile pyramid are respectively the 0 th layer 8230, the ith layer 8230and the nth layer from the bottom layer to the top layer,
reading the i = n-th layer DEM data of a distributed tile pyramid from an HDFS, and constructing corresponding tile RDD data in a memory;
step two, planning a corresponding ith layer tile path in the ith layer tile RDD data by adopting a Spark distributed computing architecture according to the starting point information and the end point information;
reading the i-1 layer DEM data of the distributed tile pyramid from the HDFS, constructing corresponding tile RDD data in a memory, screening the tiles intersected in the i-1 layer tile path and the i-1 layer DEM data, and calculating corresponding intersection points, wherein the starting points and the end points corresponding to all local paths in the i-1 layer tile path are local starting point and local end point sets respectively;
fourthly, planning local paths corresponding to each local starting point and local end point in the tile RDD data of the (i-1) th layer by adopting a Spark distributed computing architecture, and forming an (i-1) th layer three-dimensional path corresponding to the starting point and the end point together;
and step five, recording i = i-1, and repeatedly executing the steps three to four until the three-dimensional path planning of the corresponding starting point and the end point in the i =0 layer of tile RDD data is completed.
4. The big data platform distributed tile pyramid-based three-dimensional path planning method of claim 1, wherein: and when reading DEM data of the distributed tile pyramid from the HDFS, adopting a layer jump reading mode.
5. The big data platform distributed tile pyramid based three-dimensional path planning method according to claim 1, wherein: the distributed tile pyramid processing method includes the steps that distributed tile pyramid processing is conducted on large three-dimensional terrain data DEMs in areas to be explored, a series of up-sampling is conducted on the original large three-dimensional terrain data DEMs, obtained image information and the original large three-dimensional terrain data DEMs jointly generate pyramids, each layer of data of the pyramids are cut into a plurality of tiles with the same size and rectangular shape, the distributed tile pyramids are obtained, the generated distributed tile pyramids are stored in a Hadoop distributed file system HDFS, the storage mode is that tile information of each layer is serialized into tile byte blocks and tile metadata corresponding to pyramid layers, and the tile byte blocks and the tile metadata corresponding to the pyramid layers are stored in data nodes of the Hadoop distributed file system HDFS.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202211033770.7A CN115437375B (en) | 2022-08-26 | 2022-08-26 | Three-dimensional path planning method based on big data platform distributed tile pyramid |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202211033770.7A CN115437375B (en) | 2022-08-26 | 2022-08-26 | Three-dimensional path planning method based on big data platform distributed tile pyramid |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN115437375A true CN115437375A (en) | 2022-12-06 |
| CN115437375B CN115437375B (en) | 2024-05-24 |
Family
ID=84244723
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202211033770.7A Active CN115437375B (en) | 2022-08-26 | 2022-08-26 | Three-dimensional path planning method based on big data platform distributed tile pyramid |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN115437375B (en) |
Citations (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20030195672A1 (en) * | 2002-04-12 | 2003-10-16 | Gang He | Terrain augmented 3d flight path display for flight management systems |
| US20140257687A1 (en) * | 2013-03-08 | 2014-09-11 | Qualcomm Incorporated | Pyramid mapping data structure for indoor navigation |
| CN107092978A (en) * | 2017-04-05 | 2017-08-25 | 武汉大学 | A kind of shortest path hierarchical reconfiguration planning method of the Virtual earth |
| CN108446785A (en) * | 2018-01-31 | 2018-08-24 | 南京师范大学 | A kind of optimal visual overlay path planing method based on landform visible range |
| CN109272151A (en) * | 2018-09-06 | 2019-01-25 | 中南大学 | A kind of vehicle path planning algorithm optimization method based on Spark |
| CN110019610A (en) * | 2017-11-23 | 2019-07-16 | 杭州海康机器人技术有限公司 | Map tile generation method, apparatus and system |
| CN113495876A (en) * | 2020-03-19 | 2021-10-12 | 中科星图股份有限公司 | Spark-based image pyramid distributed slicing system and method |
| US20220215621A1 (en) * | 2021-09-24 | 2022-07-07 | Nanjing University Of Aeronautics And Astronautics | Three-dimensional model-based coverage path planning method for unmanned aerial vehicles |
-
2022
- 2022-08-26 CN CN202211033770.7A patent/CN115437375B/en active Active
Patent Citations (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20030195672A1 (en) * | 2002-04-12 | 2003-10-16 | Gang He | Terrain augmented 3d flight path display for flight management systems |
| US20140257687A1 (en) * | 2013-03-08 | 2014-09-11 | Qualcomm Incorporated | Pyramid mapping data structure for indoor navigation |
| CN107092978A (en) * | 2017-04-05 | 2017-08-25 | 武汉大学 | A kind of shortest path hierarchical reconfiguration planning method of the Virtual earth |
| CN110019610A (en) * | 2017-11-23 | 2019-07-16 | 杭州海康机器人技术有限公司 | Map tile generation method, apparatus and system |
| CN108446785A (en) * | 2018-01-31 | 2018-08-24 | 南京师范大学 | A kind of optimal visual overlay path planing method based on landform visible range |
| CN109272151A (en) * | 2018-09-06 | 2019-01-25 | 中南大学 | A kind of vehicle path planning algorithm optimization method based on Spark |
| CN113495876A (en) * | 2020-03-19 | 2021-10-12 | 中科星图股份有限公司 | Spark-based image pyramid distributed slicing system and method |
| US20220215621A1 (en) * | 2021-09-24 | 2022-07-07 | Nanjing University Of Aeronautics And Astronautics | Three-dimensional model-based coverage path planning method for unmanned aerial vehicles |
Non-Patent Citations (4)
| Title |
|---|
| HONG ZHONGHUA等: "A Fast Large-Scale Path Planning Method on Lunar DEM Using Distributed Tile Pyramid Strategy", IEEE JOURNAL OF SELECTED TOPICS IN APPLIED EARTH OBSERVATIONS AND REMOTE SENSING, vol. 16, 5 December 2022 (2022-12-05), pages 344 - 355 * |
| 吴华意等: "RS-ODMS:一种分布式遥感数据在线管理与服务框架", 武汉大学学报信息科学版, vol. 45, no. 12, 31 December 2020 (2020-12-31), pages 1836 - 1846 * |
| 涂斌: "基于遥感大数据平台的月球路径规划研究", 中国优秀硕士学位论文全文数据库, 11 January 2024 (2024-01-11), pages 2 - 41 * |
| 王艳丽: "基于栅格地形数据的最短路径算法研究", 中国优秀硕士学位论文全文数据库基础科学辑, no. 2020, 29 February 2020 (2020-02-29), pages 008 - 177 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN115437375B (en) | 2024-05-24 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN105528955B (en) | Generate the method and its device of road network | |
| JP4669659B2 (en) | How to organize map data | |
| JP6230213B2 (en) | Navigation device, method for determining altitude coordinates, and method for generating database | |
| CN108038249B (en) | Global map data storage organization method and call method | |
| JP5950673B2 (en) | Database for navigation device, method for outputting three-dimensional display of terrain, and method for generating database | |
| CN107092978B (en) | Shortest path layered planning method for virtual earth | |
| CN106251400B (en) | A kind of method and device based on more quadrilateral mesh building topographic map | |
| US20160239597A1 (en) | Apparatus and method for performing finite element computation | |
| WO2019019653A1 (en) | Device and method for extracting topographical boundary | |
| CN111695281B (en) | Particle rapid positioning method for tetrahedral mesh division finite element particle simulation | |
| CN104125582A (en) | Method of planning communication network | |
| CN115272637A (en) | Large-area-oriented three-dimensional virtual ecological environment visualization integration and optimization system | |
| Hong et al. | A fast large-scale path planning method on lunar DEM using distributed tile pyramid strategy | |
| CN115437375B (en) | Three-dimensional path planning method based on big data platform distributed tile pyramid | |
| CN110162903B (en) | Urban building windward surface density calculation method and system based on grid parallelism | |
| CN116467540B (en) | HBase-based massive space data rapid visualization method | |
| CN105138607A (en) | Hybrid granularity distributional memory grid index-based KNN query method | |
| CN112836872B (en) | Multi-GPU-based high-performance numerical solution method for pollutant convection diffusion equation | |
| CN112765292A (en) | Method for processing shp data by using tile technology | |
| Saupi Teri et al. | GPU utilization in geoprocessing BIG Geodata: A review | |
| Yoon | A study on terrain-surface modeling and searching algorithms for real-time simulation of off-road vehicles | |
| CN111476895A (en) | Earth ellipsoid surface gridding method | |
| CN111539375B (en) | Large-scale crowd behavior aided planning method | |
| CN110706354B (en) | Data organization method suitable for three-dimensional dynamic display of flood risk graph | |
| TWI786874B (en) | Method of the digital grid model and system thereof |
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 |