Per-Point Radar Representations for AV Object Detection

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

Existing autonomous vehicle (AV) systems face challenges in effectively processing and integrating radar data for accurate scene understanding and object detection due to noise, ambiguity, and limited contextual information, which hinders efficient navigation and safety.

Innovation Solution

A method involving sparse radar point clouds from multiple sensors, combined with machine learning techniques, to generate fused radar point clouds, extract per-point features, and apply shallow and deep global features for instance segmentation, enabling accurate object classification and tracking.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Measurement precision

If traditional radar data processing methods are used, then the system complexity is low, but the object detection accuracy and scene understanding are insufficient due to noise and ambiguity

Engineering Contradiction:

Improveobject detection accuracyVSAvoiddata processing complexity

Core Design Contradiction:

Measurement precisionVSDevice complexity

Solution Approach 1:

The patent segments radar data processing into multiple specialized modules: point cloud generation from received signals, feature extraction at per-point level, instance segmentation for object classification, and tracking. This segmentation allows each module to focus on specific aspects of data processing, improving detection accuracy while making the overall complex system manageable through modular architecture

Inventive Principle:

Principle #1Segmentation

Solution Approach 2:

The patent introduces machine learning models as intermediary components between raw radar data and final object detection results. These ML intermediaries process the noisy radar signals, extract meaningful features, and produce reliable object classifications, thereby improving measurement precision while the modular ML architecture keeps processing complexity manageable

Inventive Principle:

Principle #24Intermediary (Mediator)

2Measurement precision

If multiple radar sensors are integrated to improve scene understanding, then the object detection accuracy improves, but the data processing time and computational load increase

Engineering Contradiction:

Improvescene understanding accuracyVSAvoiddata processing time

Core Design Contradiction:

Measurement precisionVSLoss of time

Solution Approach 1:

The patent merges data from multiple radar sensors into a unified point cloud representation, combining the strengths of different sensors to improve scene understanding and object detection accuracy. This merging is achieved through efficient point cloud fusion algorithms that integrate multi-sensor data while managing computational load

Inventive Principle:

Principle #5Merging (Combining)

Solution Approach 2:

The patent performs preliminary processing of radar data from multiple sensors before full integration, including initial point cloud generation and feature extraction. This preliminary action prepares the data in advance, reducing the computational burden during final scene understanding and object detection, thereby reducing processing time

Inventive Principle:

Principle #10Preliminary action

3Reliability

If traditional radar processing is used, then the computational resources required are low, but the signal-to-noise ratio and detection reliability are poor

Engineering Contradiction:

Improvedetection reliabilityVSAvoidcomputational energy consumption

Core Design Contradiction:

ReliabilityVSUse of energy by moving object

Solution Approach 1:

The patent replaces traditional mechanical radar signal processing methods with machine learning-based processing. This substitution enables more reliable object detection and scene understanding by leveraging ML algorithms that can distinguish meaningful signals from noise more effectively, improving detection reliability while the efficient ML architecture manages computational energy consumption

Inventive Principle:

Principle #28Mechanics substitution (Replace mechanical system)

Data Source

PatentUS12504532B2Object detection and state estimation from deep learned per-point radar representations

Publication Date: 2025.12.23 GM CRUISE HOLDINGS LLC

AI summary

A current set of radar data may be combined with previous sets of radar data to create a combined set of radar data. Each of these sets of radar data and the combined set of radar data may include various data points and each of these data points may be associated with certain specific features or classifications. The set of combined data may then be pre-processed to identify distances to associate with certain data points, locations to associate with those data points, and velocities associated with the data points such that each of the respective data points may be mapped and tagged with data that identifies identify positions, velocities, and other physical details of the respective data points. This pre-processed data may then be processed by a machine learning process to identify kinematic information that may then be provided to a tracking system of an AV.