Radar Direction-of-Arrival Super-Resolution via Autoregressive Extrapolation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Automotive radar systems face challenges in achieving accurate direction of arrival (DoA) estimation due to imperfect antenna placement and calibration errors, which reduce resolution and the number of detectable objects, despite the use of super-resolution algorithms that require ideal conditions like high SNR and precise antenna spacing.
Innovation Solution
A radar system employs an autoregressive model to extrapolate signal values, increasing the size of the covariance matrix and enhancing super-resolution algorithms to improve robustness against calibration errors, allowing for more accurate DoA estimation and detection of multiple objects.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If super-resolution algorithms are used to improve DoA estimation accuracy, then measurement precision is improved, but the system becomes more sensitive to calibration errors and antenna placement inaccuracies
Solution Approach 1:
The patent introduces an intermediary signal processing layer between the raw radar signals and the DoA estimation algorithm. This intermediary layer includes steps for generating range-Doppler maps, performing spatial smoothing, and creating covariance matrices that act as mediators to reduce the direct impact of calibration errors on the final DoA estimation, thereby improving robustness while maintaining precision
Solution Approach 2:
The patent changes several processing parameters including the size of the covariance matrix, the degree of spatial smoothing, and the resolution bandwidth to optimize the balance between DoA estimation precision and robustness. By adjusting these parameters, the system achieves high measurement precision while maintaining reliability in the presence of calibration errors
2Measurement precision
If the covariance matrix size is increased to improve resolution, then measurement precision is improved, but computational complexity increases
Solution Approach 1:
The patent segments the signal processing into distinct stages: range compression, Doppler processing, spatial smoothing, and DoA estimation. Each stage operates on a portion of the data independently, allowing the covariance matrix to be constructed from smaller, manageable segments rather than processing the entire signal at once, thus reducing overall computational complexity while maintaining high resolution
Solution Approach 2:
The patent applies partial spatial smoothing rather than complete smoothing of the entire signal. By applying smoothing only to the extent necessary to achieve the desired resolution improvement, the system avoids excessive computational complexity. The covariance matrix size is increased only partially to the minimum necessary level to achieve the required resolution, rather than maximizing it
3Measurement precision
If antenna spacing is increased to improve resolution, then measurement precision is improved, but manufacturing precision requirements become more stringent
Solution Approach 1:
The patent creates a virtual copy of the physical antenna array through signal processing. By generating a covariance matrix that represents the array response, the system achieves the resolution benefits of a larger virtual aperture without requiring physically larger or more precisely positioned antennas. This virtual array effect allows high resolution with relaxed manufacturing precision requirements
Solution Approach 2:
The patent transitions from the physical spatial dimension to the signal processing dimension. Instead of relying solely on physical antenna spacing for resolution, the system uses the covariance matrix and spectral estimation techniques to achieve resolution in the angular domain. This dimensional shift allows high resolution without proportionally increasing physical antenna spacing or manufacturing precision
Data Source
AI summary
A device includes a radar processor that transmits, at a first time, a first radar signal, receives a received signal, and processes the received signal to generate a range-Doppler data frame. The radar processor determines a first snapshot comprising a first plurality of values associated with a first range-Doppler bin of the range-Doppler data frame and processes the first plurality of values in the first snapshot to generate an autoregressive model based upon the first plurality of values. The radar processor uses use the autoregressive model to extrapolate a second snapshot, wherein the second snapshot includes the first plurality of values and a second plurality values generated using the autoregressive model, determines, using the second snapshot, a full rank covariance matrix, and identifies attributes of a plurality of objects using the full rank covariance matrix.