Point Cloud Editing for Autonomous Vehicle Positioning Tests

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Current autonomous driving positioning systems face challenges in efficiently testing their performance under environmental changes due to the high cost and long mapping cycles of high-precision positioning maps, leading to positioning errors and low test efficiency, as simulating real environmental changes is difficult and can violate traffic laws.

Innovation Solution

A method and apparatus that process point cloud data from a real physical scenario to simulate environmental changes, allowing for the generation of processed point clouds that are sent to test vehicles to assess the stability of their positioning systems, enabling effective testing of environmental change scenarios without the need for real-world changes.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Reliability

If real-world environmental change scenarios are constructed for testing, then test coverage and verification capability are improved, but traffic safety is compromised and test efficiency is reduced due to the need to wait for natural occurrences

Engineering Contradiction:
Improvetest coverageVSAvoidtest efficiency
Core Design Contradiction:
ReliabilityVSProductivity

Solution Approach 1:

The patent creates a virtual copy of the real-world environment by processing point cloud data into a three-dimensional map. This virtual replica allows repeated testing of environmental change scenarios without physically modifying the real world, thereby improving test efficiency while maintaining comprehensive test coverage through simulation of various change scenarios.

Inventive Principle:
Principle #26Copying

Solution Approach 2:

The patent pre-processes point cloud data to generate three-dimensional maps before actual testing occurs. By preparing the virtual environment in advance and pre-configuring various environmental change scenarios, the system enables efficient testing without needing to wait for natural occurrences or construct physical scenarios, thus resolving the contradiction between test coverage and test efficiency.

Inventive Principle:
Principle #10Preliminary action

2Measurement precision

If high-precision positioning maps are acquired and drawn using lidar data, then positioning precision is improved, but acquisition cost and mapping cycle time increase significantly

Engineering Contradiction:
Improvepositioning precisionVSAvoidmapping cycle
Core Design Contradiction:
Measurement precisionVSLoss of time

Solution Approach 1:

The patent enables the system to use its own acquired point cloud data to generate three-dimensional maps automatically. By processing the point cloud data that the system already collects during normal operation, the system creates positioning maps without requiring separate, time-consuming high-precision mapping campaigns, thus reducing the mapping cycle while maintaining positioning precision.

Inventive Principle:
Principle #25Self-service

Solution Approach 2:

The patent changes the parameter representation from raw lidar point cloud data to a processed three-dimensional map format. This parameter transformation allows the system to utilize existing data more efficiently, converting point cloud information into a usable map representation that reduces acquisition time while preserving positioning accuracy through the maintained spatial relationships.

Inventive Principle:
Principle #35Parameter changes

3Productivity

If existing three-dimensional maps are used for positioning, then positioning can be performed, but the maps cannot be updated in real time leading to positioning errors when environments change

Engineering Contradiction:
Improvepositioning capabilityVSAvoidpositioning accuracy
Core Design Contradiction:
ProductivityVSReliability

Solution Approach 1:

The patent transforms the static three-dimensional map into a dynamic structure that can be updated in real-time. By continuously processing new point cloud data and comparing it with the existing map, the system dynamically updates the map to reflect current environmental conditions. This dynamic approach allows the positioning system to maintain accuracy despite environmental changes, resolving the contradiction between positioning capability and positioning accuracy.

Inventive Principle:
Principle #15Dynamics

Data Source

PatentEP3904829A1Method and apparatus for generating information, device, medium and computer program product
Publication Date: 2021.11.03 BEIJING BAIDU NETCOM SCI & TECH CO LTD
  • EP3904829A1 patent drawingFigure 1
  • EP3904829A1 patent drawingFigure 2
  • EP3904829A1 patent drawingFigure 3

AI summary

The present disclosure discloses a method and apparatus for generating information, a device, a medium and a computer program product, and relates to the field of autonomous driving. A specific implementation scheme is: receiving a point cloud of a target scenario, for example at a terminal device (201), and displaying a point cloud frame in the point cloud; receiving region selection information inputted by a user, the region selection information being sent by the user based on the displayed point cloud frame displayed by the terminal device (201); processing point data in a target region corresponding to the region selection information to obtain a processed point cloud; and sending the processed point cloud to a test vehicle, the processed point cloud being used for generation of positioning information. In this embodiment, the point cloud acquired from a real physical scenario is processed to simulate an environmental change of a real physical environment, and the processed point cloud is sent to the test vehicle (202), the processed point cloud being used for the generation of positioning information, thereby realizing stability test of a positioning system of the test vehicle (202). Preferably, the point data in the target region is segmented to segment point data corresponding to a target object and the point data corresponding to the target object is replaced with preset point data for replacement to obtain the processed point cloud.