Sensor Data Association Using Multi-Frame Velocity Estimates
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing data association methods in sensor systems, particularly in advanced driver assistance systems and autonomous driving, face inaccuracies in determining the association of measurement data due to the use of preset and potentially incorrect velocity vectors during track initialization, leading to imprecise data association.
Innovation Solution
A data association method that determines a velocity estimate based on position and velocity measurement data to calculate a metric value for association hypotheses, allowing for accurate determination of the association relationship without relying on prior velocity values, using smoothed, filtered, or predicted velocity estimates.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of operation
If a preset velocity vector is used for track initialization, then the data association process can be performed, but the accuracy of data association deteriorates due to potential inaccuracies in the preset velocity vector
Solution Approach 1:
The patent performs preliminary actions by establishing a track with a position vector before data association is needed. The position vector is updated using position measurement data from multiple frames, and a velocity estimate is calculated based on the change in position vectors between frames. This preliminary track establishment with accurate position and velocity estimates enables accurate data association without relying on potentially incorrect preset velocity vectors.
Solution Approach 2:
The patent implements feedback by using position measurement data from multiple frames to update the position vector and calculate the velocity estimate. The velocity estimate is derived from the actual movement observed in the measurement data rather than being preset, creating a feedback loop that continuously refines the track parameters based on actual measurements, thereby improving data association accuracy.
2Productivity
If a state vector with velocity vector is obtained in track maintenance stage, then prediction of state vector in next frame becomes possible, but the initial inaccuracy from preset velocity propagates through subsequent predictions
Solution Approach 1:
The patent performs preliminary actions by establishing a track with accurate position vector updates before prediction is needed. The position vector is continuously updated using actual position measurement data from multiple frames, creating an accurate baseline for subsequent predictions that does not depend on potentially incorrect preset velocity vectors.
Solution Approach 2:
The patent applies dynamics by making the velocity estimate adaptive rather than static. The velocity estimate is calculated dynamically from the actual change in position vectors between frames, allowing it to adjust to the real movement characteristics of the target. This dynamic approach ensures that predictions are based on actual observed behavior rather than fixed preset values.
3Measurement precision
If measurement data from multiple frames is used to determine association, then data association accuracy improves, but the complexity of processing increases
Solution Approach 1:
The patent applies segmentation by dividing the data association process into distinct stages: track maintenance stage where position vectors are updated and velocity estimates are calculated, and track initiation stage where data association is performed. This segmentation allows the complex task of using multiple frames of data to be broken down into manageable steps, improving accuracy while controlling processing complexity through structured approach.
Data Source
Figure 1~2
Figure 3~4
Figure 5
AI summary
Disclosed is a data association method, which relates to the field of sensor technologies, and includes: obtaining a plurality of pieces of measurement data (501), where each of the plurality of pieces of measurement data includes position measurement data and velocity measurement data; determining, based on position measurement data included in at least two of the plurality of pieces of measurement data, a velocity estimate corresponding to an association hypothesis of the at least two pieces of measurement data (502); determining, based on the velocity estimate corresponding to the association hypothesis and velocity measurement data in the at least two pieces of measurement data, a metric value corresponding to the association hypothesis (503); and determining an association relationship of the plurality of pieces of measurement data based on a metric value corresponding to at least one association hypothesis of the plurality of pieces of measurement data (504). It can be learned that, this application may be applied to the field of autonomous driving, intelligent driving, tracking, navigation, positioning, or the like, thereby improving accuracy of determining an association relationship of a plurality of pieces of measurement data.