Radar Height Estimation Using Sparse Reconstruction

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Current radar systems for automotive applications face challenges in accurately estimating the height of small objects at far distances, particularly between 80 m and 150 m, due to insufficient methods for power spectral density estimation and model dependencies on the number of scatterers and specular reflections, which affect accuracy and resolution.

Innovation Solution

The method employs sparse reconstruction and group-sparse models using a near-field multipath observation model that combines phase and amplitude information across arrays, and group-sparse estimation to account for incoherent measurements from different distances or arrays, enhancing the reliability of height estimation.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Measurement precision

If power spectral density estimation methods are used for height estimation, then the estimation can be performed, but the accuracy and resolution are insufficient for small objects at far distances

Engineering Contradiction:
Improveheight estimation accuracyVSAvoidestimation reliability
Core Design Contradiction:
Measurement precisionVSReliability

Solution Approach 1:

The patent transforms the height estimation problem from direct power spectral density analysis to a frequency estimation problem by changing the parameter domain. It uses the relationship between height and the frequency of interference patterns in the envelope of multi-path signals, applying inverse distance transformation and Fast Fourier Transform to extract frequency peaks that correspond to object height, thereby improving measurement precision for distant small objects

Inventive Principle:
Principle #35Parameter changes

Solution Approach 2:

The patent replaces traditional mechanical signal processing methods with spectral analysis techniques. Instead of directly analyzing signal amplitudes or time-domain characteristics, it substitutes the approach by transforming signals into the frequency domain using Fast Fourier Transform, where height information is encoded as frequency peaks, achieving superior accuracy and resolution

Inventive Principle:
Principle #28Mechanics substitution (Replace mechanical system)

2Measurement precision

If the periodicity of interference pattern method is used, then height estimation is possible, but the DC component introduces disturbance and the model is highly dependent on the number of scatterers

Engineering Contradiction:
Improveheight estimation accuracyVSAvoidmodel complexity
Core Design Contradiction:
Measurement precisionVSDevice complexity

Solution Approach 1:

The patent extracts only the essential frequency information from the interference pattern envelope using Fast Fourier Transform, separating the height-relevant frequency peaks from the problematic DC component and other irrelevant signal elements. This extraction approach eliminates disturbance from the DC component while maintaining height estimation accuracy

Inventive Principle:
Principle #2Taking out (Extraction)

Solution Approach 2:

The patent changes the modeling approach from directly using interference pattern periodicity to using frequency domain representation. By transforming to the frequency domain, the model becomes less dependent on the number of scatterers because frequency peaks directly correspond to height parameters, reducing model complexity and improving robustness

Inventive Principle:
Principle #35Parameter changes

3Ease of operation

If amplitude or power information is used instead of phases across an array, then the method is simpler, but extensions to the case of interest are limited

Engineering Contradiction:
Improvemethod simplicityVSAvoidmethod adaptability
Core Design Contradiction:
Ease of operationVSAdaptability or versatility

Solution Approach 1:

The patent creates a universal height estimation method that works across multiple scenarios by using frequency domain analysis. The same Fast Fourier Transform-based approach can estimate heights of various object types (small distant objects, larger objects) and adapt to different radar configurations, achieving both simplicity and versatility simultaneously

Inventive Principle:
Principle #6Universality (Multi-functionality)

Applied Scientific Principles

This section explains which scientific principles are used to turn an abstract innovation direction into a practical engineering solution.

Function Achieved in This Case

This approach provides improved accuracy and reliability in estimating the height of small objects at far distances, while also being robust to model uncertainties and variations in road conditions, and can estimate both small and larger objects effectively.

Implementation Method 1

radar signals are emitted to the scene and reflected radar signals from the scene are received

Methodology Applied
Scientific EffectRadar: Radar

Implementation Method 2

allow radial velocity estimation via Doppler measurements

Methodology Applied
Scientific EffectDoppler effect: Doppler Effect

Implementation Method 3

based on a near-field multipath observation model that combines phase and amplitude information for direct and ground-reflected path contributions

Methodology Applied
Scientific EffectSpecular reflection: Reflection

Data Source

PatentEP3588128B1Method for detection and height and azimuth estimation of objects in a scene by radar processing using sparse reconstruction with coherent and incoherent arrays
Publication Date: 2022.08.10 FRAUNHOFER GESELLSCHAFT ZUR FORDERUNG DER ANGEWANDTEN FORSCHUNG EV
  • EP3588128B1 patent drawingFigure 1~2C
  • EP3588128B1 patent drawingFigure 3
  • EP3588128B1 patent drawingFigure 4A~4B

AI summary

In a method for detection and height and azimuth estimation of objects in a scene by radar processing, radar signals are emitted to the scene and radar signals reflected from the scene are received using at least one multi-channel radar sensor, said radar sensor comprising a two-dimensional array of sensor elements with spatial diversity in horizontal and vertical axes. Measurement signals of the at least one radar sensor are processed to detect objects in the scene and a height and azimuth estimation of one or several detected objects is performed using compressed sensing. The compressed sensing is based on a combination of a near-field multipath observation model, which includes ground-reflected path contributions of the emitted and reflected radar signals for a flat surface, with possibly unknown reflection coefficient, for pairs of transmitting and receiving elements, and a group-sparse estimation model for measurement signals of mutually incoherent measurements taken at different distances from the detected object(s) or from mutually incoherent radar sensors or sensor elements.