Radar Object State Estimation Using Association-Weighted Measurements
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current radar sensor systems for driver assistance in vehicles face challenges in accurately estimating object states due to high systematic deviations and limited angular resolution, leading to unreliable orientation and extension assignments of radar measurements.
Innovation Solution
A method that uses weighted estimation to adapt state parameters based on association probabilities, incorporating multiple measurements to enhance robustness, and employs an unscented transform to estimate uncertainties, thereby improving the accuracy of object state estimation and reducing systematic deviations.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of operation
If a clustering algorithm is used to assign radar measurements to objects independently from existing object knowledge, then the method is easier to operate and does not require high model knowledge, but it results in leaps in object parameters such as orientation or extension
Solution Approach 1:
The patent applies preliminary action by performing a weighted fit of model parameters before final object parameter determination. The system pre-processes radar measurements by associating them with object points and calculating weighted fits, which prepares the data in advance to prevent parameter leaps during object tracking. This preliminary processing step ensures smoother transitions in orientation and extension parameters.
2Reliability
If multiple measurements are incorporated in parameter estimation of state parameters, then the data amount is decreased and robustness is increased, but the device complexity increases due to weighted estimation and association probability calculations
Solution Approach 1:
The patent merges multiple radar measurements into a unified parameter estimation through weighted fitting. By combining association probabilities with measurement data, the system integrates multiple information sources into a single robust object state estimation. This merging approach increases robustness while managing complexity through systematic integration rather than separate processing of each measurement.
Solution Approach 2:
The system changes parameters by introducing association probabilities as weighting factors in the parameter estimation process. Instead of treating all measurements equally, the system dynamically adjusts the influence of each measurement based on its association probability, thereby improving robustness without requiring proportional increases in device complexity.
3Measurement precision
If association probabilities are used as weightings for locating measurements during state parameter adaptation, then systematic deviations are reduced and measurement precision is improved, but the calculation complexity increases
Solution Approach 1:
The patent applies parameter changes by transforming association probabilities into weighting factors for measurement adaptation. This parameter transformation allows the system to improve measurement precision by giving appropriate weights to different measurements based on their reliability, while managing computational complexity through efficient probability-to-weight conversion algorithms.
Data Source
AI summary
A method for evaluating locating measurements of a surroundings sensor for a motor vehicle. The method includes: associating locating measurements with an object described by an estimated object state, for the locating measurements in each case an association probability being determined for the association of the locating measurement with the object; estimating instantaneous state parameters of the object, including an adaptation of the state parameters to the locating measurements associated with the object, weightings of the locating measurements associated with the object being taken into consideration during the adaptation, for the locating measurements in each case the weighting being dependent on the determined association probability for the association of the particular locating measurement with the object; and transferring the estimated instantaneous state parameters of the object to a state estimator for updating the estimated state of the object. A sensor system is also described.