Sensor Data Processing Alignment for Multi-Sensor Positioning Accuracy

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Current methods for positioning intelligent terminals, such as self-driving cars, using multiple sensors face challenges in merging data collected by different sensors, leading to low accuracy and poor applicability due to strict chronological data processing and linearization errors.

Innovation Solution

A method that determines a reference sensor and aligns observation data from various sensors using interpolation coefficients, allowing for flexible data processing and merging of rotation and translation data at different sampling moments, enabling out-of-order input and improving positioning accuracy.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Reliability

If linearization processing is performed strictly in chronological order using CKF-based method, then data merging can be implemented for multiple sensors, but positioning accuracy deteriorates due to gradual accumulation of errors and narrow application scope

Engineering Contradiction:
Improvepositioning accuracyVSAvoiddata processing complexity
Core Design Contradiction:
ReliabilityVSDevice complexity

Solution Approach 1:

The patent performs linearization processing in advance at each sensor's sampling moment before data merging, rather than during the merging process. This preliminary linearization reduces accumulated errors by establishing accurate local linear models at each time point, improving positioning accuracy while simplifying the subsequent merging operation.

Inventive Principle:
Principle #10Preliminary action

Solution Approach 2:

The patent introduces a flexible data processing manner that adapts to different sensor types (unary pose observation vs. binary pose observation) and different sampling moments. By dynamically selecting processing strategies based on sensor characteristics and timing, the system improves accuracy without requiring a universally complex processing framework.

Inventive Principle:
Principle #15Dynamics

2Adaptability or versatility

If strict chronological order is enforced for data merging, then processing simplicity is maintained, but applicability deteriorates due to inability to handle out-of-order sensor data

Engineering Contradiction:
ImproveapplicabilityVSAvoiddata processing ease
Core Design Contradiction:
Adaptability or versatilityVSEase of operation

Solution Approach 1:

By performing linearization processing in advance at each sensor's sampling moment, the patent prepares data in a form that can be easily merged regardless of arrival order. This preliminary preparation enables the system to accept out-of-order sensor data without complex reordering operations, improving applicability while maintaining processing simplicity.

Inventive Principle:
Principle #10Preliminary action

Solution Approach 2:

The patent changes the processing parameters by introducing different linearization strategies for unary and binary pose observations, and by allowing flexible association of sensor data with reference sensor sampling moments. This parameter flexibility improves adaptability to various sensor configurations and data arrival patterns.

Inventive Principle:
Principle #35Parameter changes

Data Source

PatentEP3680690B1Sensor data processing method and apparatus
Publication Date: 2024.04.17 HUAWEI TECH CO LTD
  • EP3680690B1 patent drawingFigure 1
  • EP3680690B1 patent drawingFigure 2
  • EP3680690B1 patent drawingFigure 3

AI summary

Embodiments of this application provide a sensor data processing method and apparatus. The method includes: determining a reference sensor from at least two sensors and sampling moments of the reference sensor; when obtaining observation data of a sensor K at a moment t1, determining a type of the sensor K; if the sensor K is a first sensor of a first type, determining a sampling moment T1 that is closest to t1 and an interpolation coefficient λ1, and calculating first rotation data and first translation data that are at T1, and a first covariance matrix; if the sensor K is a second sensor of a second type, determining two sampling moments T2 and T3 that are closest to t1 and interpolation coefficients λ2 and λ3, and calculating second rotation data and second translation data that are of the second sensor from T2 to T3, and a second covariance matrix; and merging rotation data, translation data, and covariance matrices that are at the sampling moments. According to the embodiments of this application, operational flexibility in data merging for a plurality of sensors and data processing efficiency can be improved.