Adaptive GPS Error Identification via Dead Reckoning and Pose Graphs
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current methods for identifying erroneous GPS observed values in autonomous driving systems are inadequate, particularly in outdoor environments where GPS data can be affected by multipath effects, leading to reduced accuracy in 3D scene map construction and vehicle positioning.
Innovation Solution
A method for adaptive identification of erroneous GPS observed values involves acquiring consecutive positioning data, deriving dead reckoning trajectories, and using a pose graph construction with covariance matrices and kernel functions to classify data levels and eliminate erroneous values through cost function analysis, ensuring high-precision real-time positioning.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If the identification method based on own GPS observed values (Kalman filter) is used, then the positioning accuracy can be improved through prediction models, but the system complexity increases and requires pre-planned motion trajectories and sample data which may not match actual driving conditions
Solution Approach 1:
The patent introduces an accelerometer as an intermediary device to obtain vehicle motion state information (acceleration, speed, heading angle). This intermediary provides additional measurement data that complements GPS observations, enabling the system to detect erroneous GPS values without requiring complex pre-planned trajectories or extensive sample data collection.
Solution Approach 2:
The patent changes the approach from using complex prediction models requiring trajectory data to using real-time motion state parameters (acceleration, speed, heading angle) from the accelerometer. This parameter change simplifies the system by replacing complex temporal-spatial prediction models with direct measurement-based error detection.
2Reliability
If the identification method based on combination of GPS observed values and accelerometer is used, then jumping erroneous values can be identified through speed and position variations, but gradual errors and step errors caused by multipath effects cannot be detected
Solution Approach 1:
The patent dynamically adjusts the error detection strategy by using the accelerometer to continuously track vehicle motion state. The system adapts to different error types (jumping errors, gradual errors, step errors) by comparing real-time motion state data with GPS observations, allowing flexible response to various multipath effect scenarios without fixed thresholds.
Solution Approach 2:
The patent implements feedback mechanisms where the accelerometer continuously provides motion state information that is compared with GPS-derived motion. When discrepancies are detected (indicating erroneous GPS values), the system uses this feedback to identify and eliminate erroneous observations, creating a closed-loop error detection system that adapts to different error patterns.
3Productivity
If pre-planned motion trajectories and sample data are used to obtain processing models, then the Kalman filter can analyze GPS data in real time, but the vehicle may not strictly follow preset tracks leading to inaccurate error analysis
Solution Approach 1:
The patent enables the system to self-adjust by using the accelerometer to directly measure vehicle motion state without requiring pre-planned trajectories. The system serves itself by obtaining motion information from onboard sensors rather than external preset data, eliminating the mismatch between planned and actual trajectories while maintaining real-time processing capability.
Data Source
AI summary
Disclosed is a method for adaptive identification of erroneous GPS observed value, including: acquiring positioning information of a vehicle from a GPS sensor, and extracting first observed value data; acquiring posture information and speed information of the vehicle to acquire dead reckoning trajectory data of the vehicle; eliminating the erroneous GPS observed values based on respective data on data status value, heading significant bit, the number of satellites used and horizontal dilution of precision in the first observed value data to obtain second observed value data; constructing pose graph data based on the second observed value data and acquiring processing result information; analyzing and optimizing the processing result information to eliminate the erroneous GPS observed values of which the cost function exceeds a preset cost function threshold to obtain third observed value data; and constructing a high-precision map based on the third observed value data and three-dimensional scene map data.