Radar Inter-Pulse Doppler Phase Generation Using BVH Micro-Step Scene Interpolation

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Current radar simulation methods are computationally intensive and impractical for generating simulated radar reflection data quickly and efficiently, making them unsuitable for training machine learning or artificial intelligence systems effectively.

Innovation Solution

The proposed method employs an interpolation framework that reduces computational burden by simulating radar reflections using micro-steps and a bounding volume hierarchy (BVH) data structure, limiting simulations to specific reference objects, and using Doppler phase shifts to accelerate the simulation process, allowing for faster generation of coherent measurements and Doppler velocities.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Measurement precision

If traditional radar simulation methods are used to generate simulated radar reflection data, then measurement precision and reliability are maintained, but computational time and resources increase significantly

Engineering Contradiction:
Improvesimulated radar reflection data accuracyVSAvoidcomputational time
Core Design Contradiction:
Measurement precisionVSLoss of time

Solution Approach 1:

The patent segments the continuous scene into discrete micro-steps (e.g., 100 micro-steps per scene) and processes radar reflections at specific intervals rather than continuously. This segmentation allows the system to maintain measurement precision at key points while dramatically reducing computational burden by skipping intermediate calculations, directly resolving the contradiction between accuracy and computational time.

Inventive Principle:
Principle #1Segmentation

Solution Approach 2:

The patent performs preliminary actions by pre-calculating and storing scene geometry data, object positions, and radar parameters before the actual simulation. By preparing bounding volume hierarchies and spatial indexes in advance, the system reduces real-time computational requirements while maintaining data accuracy, effectively reducing computational time without sacrificing measurement precision.

Inventive Principle:
Principle #10Preliminary action

2Quantity of substance

If traditional radar simulation methods simulate all objects in the scene, then comprehensive data coverage is achieved, but device complexity and computational resources increase

Engineering Contradiction:
Improvedata coverageVSAvoidsimulation complexity
Core Design Contradiction:
Quantity of substanceVSDevice complexity

Solution Approach 1:

The patent applies local quality by focusing computational resources on specific reference objects that are most relevant to the radar simulation rather than uniformly processing all objects. By identifying and prioritizing objects based on their radar cross-section, position, and motion characteristics, the system achieves comprehensive data coverage for critical targets while reducing device complexity by excluding less important objects from detailed simulation.

Inventive Principle:
Principle #3Local quality

Solution Approach 2:

The patent implements partial action by simulating radar reflections for only a subset of objects (reference objects) rather than all objects in the scene. This selective approach provides sufficient data coverage for training machine learning systems while significantly reducing simulation complexity and computational resources required, as the patent demonstrates that simulating all objects is excessive for the intended application.

Inventive Principle:
Principle #16Partial or excessive action

3Measurement precision

If full scene re-rending is performed at pulse rate for Doppler phase coherent return signals, then signal accuracy is maintained, but productivity decreases

Engineering Contradiction:
ImproveDoppler phase coherenceVSAvoidsimulation speed
Core Design Contradiction:
Measurement precisionVSProductivity

Solution Approach 1:

The patent implements periodic action by performing full Doppler phase coherent simulations only at specific pulse intervals rather than at every pulse. The system maintains measurement precision by ensuring phase coherence at these periodic intervals while improving productivity by using interpolated or approximated values for intermediate pulses, thereby reducing the frequency of computationally intensive full scene re-rending operations.

Inventive Principle:
Principle #19Periodic action

Solution Approach 2:

The patent uses copying by creating simplified representations or proxies of the full scene for intermediate pulse calculations. Instead of re-rendering the complete scene at every pulse, the system copies and interpolates from previously computed phase coherent data, maintaining adequate signal accuracy for training purposes while dramatically improving simulation speed and productivity.

Inventive Principle:
Principle #26Copying

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 significantly reduces computational time and resources required for simulating radar reflections, enabling faster and more efficient training of AI systems to identify objects encountered by autonomous vehicles.

Implementation Method 1

motion data that identifies start times, start locations, end times, and end locations of a set of reference objects in a scene-scenario... simulations may be performed to estimate the appearance of reflected radar signals... These simulations may identify patterns of reflected radar energy

Methodology Applied
Scientific EffectDoppler effect: Doppler Effect

Data Source

PatentUS12072437B2Radar inter-pulse doppler phase generation using performant bounding volume hierarchy micro-step scene interpolation
Publication Date: 2024.08.27 GM CRUISE HOLDINGS LLC
  • US12072437B2 patent drawing
  • US12072437B2 patent drawing
  • US12072437B2 patent drawing

AI summary

The present disclosure is directed to simulating patterns of reflected radar energy off of reference objects using motion data associated with these reference objects. This motion data may identify start times, start locations, end times, and end locations of a limited number reference objects in a set of discrete scenes. Each of these discrete scenes may also have a same time duration. Motion of these specific objects between a start time and an end time of each discrete scene may be interpolated. Once the locations of the objects are interpolated for a given scene, simulations may be performed to estimate the appearance of reflected radar signals that would be received by a radar apparatus. These simulations may identify patterns of reflected radar energy after radar signals have been emitted from the radar apparatus and these patterns may then be provided to train a machine learning apparatus.