CN-122015888-A - Vehicle positioning method and system for multi-level overpass scene

CN122015888ACN 122015888 ACN122015888 ACN 122015888ACN-122015888-A

Abstract

The invention discloses a vehicle positioning method and system for a multi-level cross-line bridge scene, and the method comprises the steps of obtaining real-time positioning data comprising three-dimensional positions and positioning states, obtaining a three-dimensional road network model based on the three-dimensional positions, determining a candidate lane screening range and a candidate lane set by taking the positioning states as self-adaptive factors, calculating observation probabilities based on three-dimensional spatial relations between vehicle observation positions and candidate lanes, calculating transition probabilities based on three-dimensional topological connection relations between candidate lane pairs at adjacent moments, searching optimal paths through a Viterbi algorithm by utilizing the observation probabilities and the transition probabilities, and outputting lanes where vehicles are located and matching positions. According to the invention, the hidden Markov model is expanded to a three-dimensional space, the computational complexity is reduced through self-adaptive screening, and the matching precision is improved through multi-dimensional feature fusion, so that the method is suitable for three-dimensional traffic scenes such as viaducts, tunnels and the like.

Inventors

WU ZHOU
Xu Chaohan
YAO YIBIN
ZHANG LIANG
ZHANG BAO

Assignees

武汉大学

Dates

Publication Date: 20260512
Application Date: 20260410

Claims (10)

1. A vehicle positioning method for a multi-level overpass scene is characterized by comprising the following steps: Acquiring real-time positioning data of a vehicle, wherein the real-time positioning data comprises a three-dimensional position and a positioning state of the vehicle; based on the three-dimensional position, acquiring a three-dimensional road network model in a preset range; determining a candidate lane screening range by taking the positioning state as an adaptive factor; screening a candidate lane set at the current moment from the three-dimensional road network model according to the candidate lane screening range; Calculating an observation probability based on a three-dimensional spatial relationship between a vehicle observation position and each candidate lane in a candidate lane set, wherein the three-dimensional spatial relationship comprises a shortest distance, a direction difference, an elevation difference and a gradient difference; Calculating transition probability based on three-dimensional topological connection relations between candidate lane pairs at adjacent moments; searching an optimal path in a three-dimensional state sequence by using the observation probability and the transition probability through a Viterbi algorithm, and outputting a lane where a vehicle is and a matching position; And taking the current matching result as prior information of matching at the next moment, and returning to the step of executing the acquisition of the real-time positioning data of the vehicle until the positioning is finished.
2. The vehicle positioning method for the multi-level overpass scene as set forth in claim 1, wherein the positioning state comprises a fixed solution, a floating solution, a single-point solution or an inertial navigation solution only, and the candidate lane screening range is set to different values or functions according to the different positioning states.
3. The vehicle positioning method for a multi-level overpass scenario of claim 2, wherein the candidate lane screening range increases with time when the positioning state is inertial navigation only.
4. The method for locating a vehicle in a multi-level overpass scene as set forth in claim 1, wherein the observation probability is calculated by multiplying a first exponential decay function based on three-dimensional space distance, a second exponential decay function based on three-dimensional direction deviation, a third exponential decay function based on elevation deviation, and a fourth exponential decay function based on gradient deviation.
5. The vehicle positioning method for the multi-level overpass scene as set forth in claim 1, wherein said transition probability is calculated by multiplying a fifth exponential decay function based on distance connection consistency determined based on a difference in three-dimensional spatial distance between said candidate lane pairs by a sixth exponential decay function based on direction connection consistency determined based on a difference in three-dimensional direction between said candidate lane pairs.
6. A vehicle locating system for a multi-level overpass scenario, comprising: The positioning module is used for acquiring real-time positioning data of the vehicle, wherein the real-time positioning data comprise a three-dimensional position and a positioning state of the vehicle; the road network acquisition module is used for acquiring a three-dimensional road network model in a preset range based on the three-dimensional position; the self-adaptive screening module is used for determining a candidate lane screening range by taking the positioning state as a self-adaptive factor and screening a candidate lane set at the current moment from the three-dimensional road network model according to the candidate lane screening range; The system comprises an observation probability calculation module, a calculation module and a calculation module, wherein the observation probability calculation module is used for calculating the observation probability based on a three-dimensional spatial relationship between the vehicle observation position and each candidate lane in the candidate lane set, wherein the three-dimensional spatial relationship comprises a shortest distance, a direction difference, an elevation difference and a gradient difference; the transition probability calculation module is used for calculating the transition probability based on the three-dimensional topological connection relation between each candidate lane pair at adjacent moments; And the path decoding module is used for searching an optimal path in a three-dimensional state sequence through a Viterbi algorithm by utilizing the observation probability and the transition probability, outputting a lane where the vehicle is located and a matching position, taking a current matching result as prior information matched at the next moment, triggering the positioning module to acquire real-time positioning data of the vehicle again, and realizing real-time continuous positioning.
7. The vehicle positioning system for a multi-level overpass scenario of claim 6, wherein the positioning state comprises a fixed solution, a floating solution, a single-point solution, or an inertial navigation solution only, and the adaptive screening module sets different candidate lane screening ranges according to different positioning states.
8. The vehicle positioning system for a multi-level overpass scene as recited in claim 6, wherein said observation probability calculation module is further configured to multiply a first exponential decay function based on three-dimensional spatial distance, a second exponential decay function based on three-dimensional directional deviation, a third exponential decay function based on elevation deviation, and a fourth exponential decay function based on slope deviation.
9. The vehicle positioning system for a multi-level overpass scenario of claim 6, wherein said transition probability calculation module is further configured to multiply a fifth exponential decay function based on distance connection consistency with a sixth exponential decay function based on direction connection consistency, wherein said distance connection consistency is determined based on differences in three-dimensional spatial distances between said candidate lane pairs, and said direction connection consistency is determined based on differences in three-dimensional directions between said candidate lane pairs.
10. A computer readable storage medium having stored thereon a computer program which when executed by a processor implements a method of vehicle localization for a multi-level overpass scenario as claimed in any of claims 1 to 5.

Description

Vehicle positioning method and system for multi-level overpass scene Technical Field The invention belongs to the fields of intelligent transportation, vehicle navigation and high-precision map matching, and particularly relates to a vehicle positioning method and system for a multi-level overpass scene. Background With the acceleration of the urban process, multilayer three-dimensional traffic networks are becoming more common, and particularly, high-rise structure dense areas such as overpasses, overpasses and the like, vehicle positioning accuracy faces serious challenges. The traditional vehicle positioning method is mostly based on two-dimensional plane road network design, and is difficult to effectively process complex three-dimensional scenes such as bridge intersection, overlapping and the like, so that the problems of positioning drift, mismatching and the like are caused. The Hidden Markov Model (HMM) is used as a common map matching method, has good performance in a planar road network, but has the problems of sharply increased search space of candidate lanes, reduced matching efficiency, lower matching precision caused by single observation probability calculation characteristic and the like in a three-dimensional road network, and severely restricts the positioning precision and practicability of the hidden Markov model under a three-dimensional traffic environment. Therefore, how to realize continuous, accurate and efficient vehicle positioning in a complex three-dimensional road structure has become a key problem to be solved in the fields of intelligent transportation, automatic driving, high-precision navigation and the like. Disclosure of Invention In order to overcome the defects of the prior art, the invention provides a vehicle positioning method for a multi-level overpass scene, which expands a hidden Markov model which is conventionally applied to a planar road network into a three-dimensional space, so that complex road structures such as a three-dimensional intersection, a parallel laminated bridge and the like can be effectively processed, and a feasible solution is provided for continuous and accurate vehicle positioning under complex three-dimensional traffic scenes such as viaducts, tunnels, multi-layer intercommunication and the like. According to an aspect of the present disclosure, a vehicle positioning method for a multi-level bridge-crossing scene is provided, including: Acquiring real-time positioning data of a vehicle, wherein the real-time positioning data comprises a three-dimensional position and a positioning state of the vehicle; based on the three-dimensional position, acquiring a three-dimensional road network model in a preset range; determining a candidate lane screening range by taking the positioning state as an adaptive factor; screening a candidate lane set at the current moment from the three-dimensional road network model according to the candidate lane screening range; Calculating an observation probability based on a three-dimensional spatial relationship between a vehicle observation position and each candidate lane in a candidate lane set, wherein the three-dimensional spatial relationship comprises a shortest distance, a direction difference, an elevation difference and a gradient difference; Calculating transition probability based on three-dimensional topological connection relations between candidate lane pairs at adjacent moments; searching an optimal path in a three-dimensional state sequence by using the observation probability and the transition probability through a Viterbi algorithm, and outputting a lane where a vehicle is and a matching position; And taking the current matching result as prior information of matching at the next moment, and returning to the step of executing the acquisition of the real-time positioning data of the vehicle until the positioning is finished. As a further technical scheme, the positioning state comprises a fixed solution, a floating solution, a single-point solution or an inertial navigation solution only, and the candidate lane screening range is set to different values or functions according to different positioning states. As a further technical aspect, when the positioning state is inertial navigation only, the candidate lane screening range increases with time. As a further technical scheme, the observation probability is calculated by multiplying a first exponential decay function based on the three-dimensional space distance, a second exponential decay function based on the three-dimensional direction deviation, a third exponential decay function based on the elevation deviation and a fourth exponential decay function based on the gradient deviation. The calculation mode of the transition probability is that a fifth exponential decay function based on distance connection consistency is multiplied by a sixth exponential decay function based on direction connection consistency, wherein the distance connec