Map data updating method, apparatus, device, and readable storage medium

Patent metadata

Field	Value
Publication number	US-11859998-B2
Application number	US-202117207542-A
Country	US
Kind code	B2
Filing date	Mar 19, 2021
Priority date	Sep 4, 2020
Publication date	Jan 2, 2024
Grant date	Jan 2, 2024

Abstract

The present application discloses a map data updating method, an apparatus, a device and a readable storage medium. The specific implementation solution is: after receiving road information reported by an electronic device, a server obtains multiple sequences according to the road information, and each road information belonging to the same sequence has the same type and location. After that, the server inputs each road information contained in the sequences to a pre-trained neural network model, so that the neural network model outputs a recognition result according to the sequences. The server updates map data according to the recognition result. With such solution, valid road information is recognized by combining context of each road information in the sequences and the neural network technology, and the map data is updated, which achieves the purpose of accurately updating the map data.

First claim

What is claimed is:

1. A map data updating method, comprising:
receiving road information reported by an electronic device, wherein the road information is road information broadcast to the electronic device by a roadside unit (RSU);
determining at least one sequence according to the road information, wherein road information belonging to the same sequence in the at least one sequence has the same type and occurrence location;
inputting road information contained in each sequence in the at least one sequence into a pre-trained neural network model to obtain a recognition result of a corresponding sequence, wherein the recognition result is used to indicate whether the road information belonging to the corresponding sequence is valid, and when the road information belonging to the corresponding sequence is valid, the road information belonging to the corresponding sequence is real road information; and
updating map data by utilizing the road information belonging to the corresponding sequence if the road information belonging to the corresponding sequence is valid.

2. The method according to claim 1 , wherein before the inputting road information contained in each sequence in the at least one sequence into a pre-trained neural network model to obtain a recognition result of a corresponding sequence, the method further comprises:
acquiring a sample set, wherein samples in the sample set comprise positive samples and negative samples, the positive samples are real road information, and the negative samples are false road information;
dividing the samples in the sample set to obtain at least one sample sequence, wherein samples belonging to the same sample sequence in the at least one sample sequence have the same type and occurrence location; and
training an initial model according to the at least one sample sequence to obtain the neural network model.

3. The method according to claim 2 , wherein the training an initial model according to the at least one sample sequence to obtain the neural network model comprises:
determining, for an i th sample sequence, a feature vector of each sample in the i th sample sequence in an embedding layer of the initial model, wherein the i th sample sequence is obtained according to any sample sequence of the at least one sample sequence;
learning, by utilizing a long-short-term memory recurrent neural network layer of the initial model, the feature vector of each sample in the i th sample sequence to obtain multiple context vectors, wherein each context vector of the multiple context vectors is used to indicate relationships among samples in the i th sample sequence; and
training a Concatenate layer, a Fully Connected layer and a loss function layer of the initial model according to the multiple context vectors to obtain the neural network model.

4. The method according to claim 3 , wherein the training a Concatenate layer, a Fully Connected layer and a loss function layer of the initial model according to the multiple context vectors to obtain the neural network model comprises:
concatenating the multiple context vectors in the Concatenate layer of the initial model to obtain a concatenating vector; and
learning, by utilizing the concatenating vector, the Fully Connected layer and the loss function layer of the initial model to obtain the neural network model.

5. The method according to claim 3 , wherein the i th sample sequence is any sample sequence in the at least one sample sequence; or
the i th sample sequence is a subsequence of any sample sequence in the at least one sample sequence.

6. The method according to claim 3 , wherein the determining, for an i th sample sequence, a feature vector of each sample in the i th sample sequence comprises:
extracting at least one of an electronic device feature, a roadside unit feature, and a road information feature corresponding to the each sample in the i th sample sequence, wherein the electronic device feature is used to characterize an electronic device that reports the each sample, the RSU feature is used to characterize an RSU that broadcasts the each sample to the electronic device, and the road information feature is used to characterize the each sample; and
generating, for the each sample in the i th sample sequence, the feature vector of the each sample according to at least one of the electronic device feature, the RSU feature, and the road information feature corresponding to the each sample.

7. The method according to claim 6 , wherein the electronic device feature comprises an identification of the electronic device, the number of times that the electronic device reports the each sample, or the number of times that the electronic device reports a valid sample, and the method further comprises:
de-duplicating the each sample reported by the electronic device to determine the number of times that the electronic device reports a non-repetitive sample; and
de-duplicating the valid sample reported by the electronic device to determine the number of times that the electronic device reports a non-repetitive valid sample.

8. The method according to claim 6 , wherein the RSU feature comprises an identification of the RSU, the total number of times that the RSU broadcasts the each sample, and the number of times that the RSU broadcasts the valid sample, and the method further comprises:
removing the number of times that the RSU repeatedly broadcasts the each sample from the total number of times that the RSU broadcasts the each sample; and
removing the number of times that the RSU repeatedly broadcasts the valid sample from the number of times that the RSU broadcasts the valid sample.

9. The method according to claim 6 , wherein the road information feature comprises at least one of the following features: a sample type, a sample location, a sample start time, a sample end time, and a time when the electronic device receives the each sample, and wherein the sample location is used to characterize a geographic location where the each sample occurs.

10. A non-transitory computer-readable storage medium storing computer instructions, wherein the computer instructions are used to cause a computer to execute the method according to claim 1 .

11. An electronic device, comprising:
at least one processor; and
a memory communicatively connected with the at least one processor; wherein
the memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor to enable the at least one processor to execute the following steps:
receiving road information reported by an electronic device, wherein the road information is road information broadcast to the electronic device by a roadside unit (RSU);
determining at least one sequence according to the road information, wherein road information belonging to the same sequence in the at least one sequence has the same type and occurrence location;
inputting road information contained in each sequence in the at least one sequence into a pre-trained neural network model to obtain a recognition result of a corresponding sequence, wherein the recognition result is used to indicate whether the road information belonging to the corresponding sequence is valid, and when the road information belonging to the corresponding sequence is valid, the road information belonging to the corresponding sequence is real road information; and
updating map data by utilizing the road information belonging to the corresponding sequence if the road information belonging to the corresponding sequence is valid.

12. The electronic device according to claim 11 , the at least one processor is fu

Assignees

• Beijing Baidu Netcom Sci & Tech Co Ltd

Inventors

• Xia Deguo
• Zhang Liuhui
• Huang Jizhou
• Zhao Hui
• Lu Zhen
• Bai Hongxia
• Liu Yuting

Classifications

• G06N3/0442

  characterised by memory or gating, e.g. long short-term memory [LSTM] or gated recurrent units [GRU] · CPC title

• G06N3/09

  Supervised learning · CPC title

• G01C21/3893 Primary

  Transmission of map data from distributed sources, e.g. from roadside stations · CPC title

• G01C21/3815

  Road data · CPC title

• G01C21/3837

  Data obtained from a single source · CPC title