OSM-based slam map global alignment system

Abstract

A kind of OSM-based SLAM map global alignment system of class map, comprising: prior OSM-based map processing module, SLAM map data processing module, global map rough alignment module and fine alignment module based on graph optimization, the topological structure shared in prior map is used as key point, topological map is constructed as the transfer structure of map alignment, reach the effect of aligning perception map and road network map, give SLAM map global coordinate, relatively high precision can be realized to complete the global coordinate calibration of SLAM map, i.e. longitude and latitude coordinates in geographical sense, and the semantic information association after SLAM map and OSM alignment, for the application of future map has good application expansibility.

Description

Technical Field

[0001] This invention relates to a technology in the field of intelligent driving, specifically a global map alignment system based on Localization and Mapping Technology (SLAM) of open-source maps (OpenStreet Map). Background Technology

[0002] Current localization and mapping technologies typically create maps with only local information in the vehicle's coordinate system, lacking global information aligned with the world coordinate system. This is far from sufficient for real-world driving scenarios. In completely unfamiliar environments, without surveying or high-definition map assistance, SLAM technology is essentially unable to obtain global coordinates, while high-definition maps... Figure One Generally, obtaining global coordinates for SLAM maps requires significant upfront investment and precise surveying, which undoubtedly increases the cost of SLAM map applications. Therefore, an automatic map alignment method that can assign global coordinates to SLAM maps at low cost is essential for the industrial application of SLAM maps. Summary of the Invention

[0003] This invention addresses the shortcomings of existing SLAM technologies, which only provide local information in the vehicle coordinate system and lack global information aligned with the world coordinate system. This leads to the inability to obtain global coordinate positions in unfamiliar scenes or without the assistance of surveying or high-definition maps. The invention proposes a global alignment system for SLAM maps based on OSM-like maps. Utilizing the common topological structures of prior maps as key points, a topological map is constructed as a transit structure for map alignment. This achieves alignment between the perceived map and the road network map, endowing the SLAM map with global coordinates. It can achieve relatively high accuracy in SLAM map global coordinate calibration, i.e., geographical latitude and longitude coordinates. Furthermore, the semantic information association between the SLAM map and OSM alignment provides good application extensibility for future map applications.

[0004] This invention is achieved through the following technical solution:

[0005] The application relates to an OSM-based SLAM map global alignment system, comprising a prior OSM-based map processing module, a SLAM map data processing module, a global map coarse alignment module and a fine alignment module based on graph optimization, wherein: the prior OSM-based map processing module extracts a prior topological map with semantic nodes meaningful for driving from original map data; the SLAM map data processing module obtains SLAM map data through a SLAM algorithm, generates a SLAM topological information map after sparse processing and key point extraction; the global data coarse alignment module performs coarse alignment processing on the prior topological map and the SLAM topological information map, and filters out topological map node pairs meeting the graph optimization fine alignment condition; the graph optimization fine alignment module constructs a graph optimization system according to the map node data to be aligned, optimizes and adjusts the initially constructed SLAM topological information map, obtains an optimized SLAM topological information map, and obtains a SLAM map containing global coordinates through inverse processing of the SLAM map data processing module.

[0006] The prior topological map G map ={V map ,E map} wherein V map and E map represent nodes and edges in the topological map respectively.

[0007] The SLAM topological information map G SLAM ={V SLAM ,E SLAM} wherein V SLAM and E SLAM represent nodes and edges in the SLAM topological map respectively.

[0008] The original map data refers to a prior OSM-based road network map, that is, map data in the OpenStreetMap format, comprising latitude and longitude coordinates of a point in a map and semantic information Nodes, marked as Road or contour information WAYS composed of the Nodes, marked as And storage of possible association information Relations between various information.

[0009] The extraction refers to that the prior OSM map processing module analyzes latitude and longitude coordinates of a point in a map and semantic information and road or contour (such as a fence) information According to TAGS N seg and TAGSW seg extractable road after extracting the road as V map , semantic information is retained and relations are established according to the associated information topological relations E between each other map , a topological semantic map G is formed map = {V map , E map}.

[0010] The SLAM map data refers to a SLAM positioning trajectory and surrounding environment information established by an existing SLAM algorithm (ORB-SLAM of a camera sensor, LOAM of a radar sensor, etc.), mainly in the form of a point cloud, and the map contains local coordinates of map points.

[0011] The sparse processing refers to extracting fewer key points in places without curvature changes in the obtained SLAM map data to achieve the effect of sparse processing and reduce the amount of data.

[0012] The key point extraction refers to selecting a fixed time length T as a sliding window, taking the orientation of the first positioning point X0 in T time as the positive direction, fitting the positioning points in T time as a curve, and marking the extreme points of the curve as key points V SLAM , that is, the true meaning of inflection points, and sparsely selecting data according to the curvature size to reduce the computational burden of later optimization.

[0013] The coarse alignment processing includes topological information alignment and semantic information alignment.

[0014] The topological information alignment refers to comparing the degrees of the to-be-matched V map and V SLAM , and the structure of a certain level of sub-tree developed from the root node itself, and taking the difference in degrees and the similarity of the sub-tree structure (using the WL-Test method to detect) as the first layer of judgment for alignment.

[0015] The semantic information alignment refers to judging whether the to-be-matched V SLAM may have the semantic information in the case that G map contains semantic information, for example, V map with a road intersection semantic label is more likely to be matched to V SLAM with greater curvature.

[0016] The to-be-aligned map node data refers to topological map node pairs (V match map , V match SLAM), that is, the result of coarse alignment

[0017] The topological map node pair (V match map , V match SLAM ) wherein V match map and V match SLAM The number of elements of the two node sets is the same and one-to-one matching relationship.

[0018] The optimization adjustment refers to: according to the graph optimization system G match = {V match , E match} wherein V match = V SLAM ∪ V match map , keep the node attribute in V match map fixed as one of the optimization references; E match = Relation (V match map , V match SLAM ) ∪ E SLAM , the edges in the SLAM topological information map obtained by the SLAM map data processing module are again input to the graph optimization fine alignment module, and iterative optimization is performed until the maximum number of iterations I max or the optimization error is less than the error threshold Error th , to obtain the optimized SLAM topological information map.

[0019] The optimization error refers to: the value of the loss function that satisfies E match as the optimization target. When there is no error, it means that there is a set of solutions that can perfectly satisfy all the relational expressions of E match , but it is generally impossible to achieve. The optimization error is the value of the loss function at the current iteration.

[0020] The inverse processing refers to: restoring the optimized SLAM topological information map into the original SLAM map format, including the coordinate position changes in the SLAM trajectory and environmental information, and the addition of global coordinate information.

[0021] Technical effects

[0022] The application uses the structure of a graph to match two heterogeneous map data of OSM type road network map and SLAM perception map, and combines a graph optimization algorithm to optimize the SLAM perception map and give the SLAM mapping result global position coordinates. The application uses lightweight original OSM type road network map data to achieve the effect of giving the SLAM mapping result global position coordinates. BRIEF DESCRIPTION OF DRAWINGS

[0023] Figure 1 The system schematic diagram of the application is shown in the figure.

[0024] Figure 2 The effect schematic diagram of the embodiment is shown in the figure.

[0025] In the figure: (a) is the visualization of the prior topological map; (b) is the visualization of the SLAM topological map, in which the yellow thickened nodes are key nodes; (c) is the visualization result after alignment. DETAILED DESCRIPTION

[0026] As shown in the figure, the application relates to a SLAM map global alignment system based on an OSM type map, which comprises a prior OSM type map processing module, a SLAM map data processing module, a global map coarse alignment module and a fine alignment module based on graph optimization.

[0027] The prior OSM type map processing module comprises a data splitting unit and a graph generating unit, wherein the data splitting unit splits the data according to public map data, map data obtained in a previous experiment process or prior information D map , to obtain latitude and longitude coordinates, semantic information Nodes, road or contour information Ways and association information Relations in the map, the graph generating unit analyzes and processes the information according to the Nodes, Ways and Relations, extracts the vehicle drivable road and the road category information, extracts the vehicle drivable road N seg and the road W seg according to the TAGS , then extracts the road on the road as a graph node V map , retains the semantic information and establishes the topological relationship between V map and V map according to the Relations, as an edge E map , to obtain a prior topological map G map = {V map , E SLAM} result.

[0028] The SLAM map data processing module includes a key point extraction unit and a graph generation unit. The key point extraction unit, based on the SLAM map information, selects a fixed time length T as a sliding window, takes the orientation of the first location point X0 within time T as the positive direction, fits the location points within time T into a curve, and marks the extreme points of the curve as key points V. SLAM This refers to the true inflection point, and data is sparsely selected based on the curvature to reduce the computational burden of later optimization. The graph generation unit is based on the key point V. SLAM and key point V in the SLAM map SLAM Based on relevant semantic information and local location information, an initial SLAM topology map G is constructed. SLAM ={V SLAM E SLAM}

[0029] The input to the global map coarse alignment module is the output of the first two modules {G}. map G SLAM This module will perform coarse alignment on the input data and filter out topological map node pairs (V) that meet the conditions for fine alignment in graph optimization. match map V match SLAM V match map and V match SLAM The two node sets have the same number of elements and a one-to-one matching relationship. The alignment implementation logic includes two parts: topological information and semantic information. Topological information alignment compares the V to be matched. map and V SLAM The degree of the node and the structure of subtrees of a certain level extending from itself as the root node are used as the first level of alignment judgment based on the difference in degree and the similarity of the subtree structures (detected using the WL-Test method); semantic information alignment is then performed on G. map When semantic information is included, the treatment of matching V SLAM Determine whether the semantic information is likely to be present, such as a V with an intersection semantic label. map More likely to match a V with a larger curvature SLAM .

[0030] After coarse alignment (V) match map V match SLAM The node pairs have already assigned global coordinates to the SLAM map to some extent, but further optimization of the alignment results is still needed.

[0031] The graph optimization-based fine alignment module constructs a graph optimization system G based on the map node data to be aligned. match= {V match ,E match} is optimized and adjusted based on the original SLAM data, wherein: match = V SLAM ∪ V match map , and the node attributes in the set V match map are fixed as one of the bases for optimization; E match = Relation(V match map , V match SLAM ∪ E SLAM , the edges of the output of the above alignment module and the edges of the SLAM topology graph are involved in the optimization process. The optimization process will be iterated until the maximum number of iterations I max or the optimization error is less than the error threshold Error th . Finally, a new set of SLAM topology information maps G' SLAM = {V' SLAM , E' SLAM} is output, and the topology information map is restored to a new SLAM map containing global coordinates through the inverse process of the logic in the SLAM map data processing module.

[0032] The embodiment is directed to the global map alignment method for SLAM technology of the above-mentioned system. First, the prior OSM-like map and SLAM perception data are processed to obtain the corresponding topology graph as shown in (b). Figure 2 After node matching, the node information in the SLAM topology graph is optimized through graph optimization, and the global position coordinate information in the prior map topology graph is assigned to the SLAM node to obtain the alignment effect in (c). Figure 2 Figure 2

[0033] Compared with the prior art, the method adds global coordinate information to the SLAM map with only local information without surveying and mapping.

[0034] The above specific embodiments can be adjusted in different ways by those skilled in the art without departing from the principles and purposes of the present application. The protection scope of the present application is subject to the claims and is not limited by the above specific embodiments. Each implementation within the scope is subject to the constraints of the present application.​​

Claims

1. A global alignment system for SLAM maps based on OSM-like maps, characterized in that, include: The system comprises a prior OSM-type map processing module, a SLAM map data processing module, a global map coarse alignment module, and a graph optimization-based fine alignment module. Specifically: the prior OSM-type map processing module extracts a prior topology map containing semantically meaningful nodes for driving from the original map data; the SLAM map data processing module obtains SLAM map data using the SLAM algorithm, performs sparsity processing and keypoint extraction, and generates a SLAM topology map based on the keypoints; the global data coarse alignment module performs coarse alignment processing on the prior topology map and the SLAM topology map, filtering out topology map node pairs that meet the graph optimization fine alignment conditions; the graph optimization fine alignment module constructs a graph optimization system based on the map node data to be aligned, optimizing and adjusting the initially constructed SLAM topology map to obtain an optimized SLAM topology map, which is then inversely processed by the SLAM map data processing module to obtain a SLAM map containing global coordinates. The aforementioned prior topology map middle, and These represent nodes and edges in the topological map, respectively. The SLAM topology information map middle, and These represent nodes and edges in the SLAM topology map, respectively. The original map data refers to: a priori OSM-type road network map, which is similar to OpenStreetMap format map data, including: latitude and longitude coordinates of a point on the map and semantic information Nodes, marked as... ;Ways, which contain road or outline information composed of Nodes, are tagged as And the relationships between the various pieces of information; The coarse alignment process includes: topological information alignment and semantic information alignment; Topological information alignment refers to: comparing the parts to be matched and The degree of the subtree and the structure of the subtrees at a certain level with itself as the root node are used as the first level of alignment judgment based on the difference in degree and the similarity of the subtree structure. Semantic information alignment refers to: in When semantic information is included, the matching process Determine whether the semantic information is likely to be present; The map node data to be aligned refers to: topological map node pairs This is the result of coarse alignment, where: and The two node sets have the same number of elements and are matched one-to-one; The aforementioned optimization and adjustment refers to: optimizing the system based on the graph. middle, ,Keep The node attributes in the set are fixed, serving as one of the benchmarks for optimization; The edges in the SLAM map data obtained from the graph optimization fine alignment module and the SLAM topology information map obtained from the SLAM map data processing module are used as input to the graph optimization fine alignment module again for iterative optimization until the maximum number of iterations is reached. Or the optimization error is less than the error threshold So far, we have obtained an optimized SLAM topology information map.

2. The global alignment system for SLAM maps based on OSM-like maps according to claim 1, characterized in that, The extraction mentioned above refers to the prior OSM map processing module analyzing the latitude and longitude coordinates and semantic information of points in the original map data. and road or outline information ,according to and Extracting the road where the vehicle can drive Afterwards, extract the road As Preserve semantic information and establish relationships based on relational information (Relations). Topological relationships between each other To form a topological semantic map .

3. The SLAM map global alignment system based on OSM-type maps according to claim 1, characterized in that, The aforementioned sparsity processing refers to extracting fewer key points from areas of the obtained SLAM map data where there is no curvature change in order to achieve the effect of sparsity processing and reduce the amount of data.

4. The SLAM map global alignment system based on OSM-type maps according to claim 1, characterized in that, The key point extraction refers to: selecting a fixed time length. As a sliding window, The first positioning point within the time period The orientation is positive, and the fitting... The location points within a time period are represented by a curve, and the extreme points of the curve are marked as key points. This refers to the true inflection point, and data is sparsely selected based on the curvature to reduce the computational burden of later optimization.

5. The SLAM map global alignment system based on OSM-type maps according to claim 1, characterized in that, The inverse processing refers to restoring the optimized SLAM topology map to the original SLAM map format, including changes in coordinate positions in the SLAM trajectory and environmental information, as well as the addition of global coordinate information.

6. A global map alignment method for SLAM technology based on any one of the OSM-based SLAM map global alignment systems described in claims 1-5, characterized in that, First, the prior OSM-type map and SLAM perception data are processed to obtain the corresponding topology map. After node matching, the node information in the SLAM topology map is optimized through graph optimization, and the global position coordinate information in the prior map topology map is assigned to the SLAM nodes.

Images