Radar Stationary Object Detection via ML Range-Time Maps
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing radar systems struggle to accurately and efficiently detect smaller stationary objects while a vehicle is in motion, often failing to identify essential features which can lead to erratic vehicle behavior.
Innovation Solution
A method for a radar system using a machine-learned model for stationary object detection, involving the processing of radar data to generate range-azimuth maps, interpolated range-azimuth maps, and range-time maps, which are then used to extract features for input into a machine-learned model configured for stationary object detection.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If traditional radar systems use point-cloud representation to detect stationary objects, then large stationary objects can be detected, but smaller stationary objects cannot be detected with sufficient accuracy and speed
Solution Approach 1:
The patent segments the detection process into multiple specialized stages: generating Doppler beam vectors from time-series frames, creating range-azimuth maps through super-resolution operations, generating interpolated range-azimuth maps with precise timing, and finally detecting stationary objects using machine-learned models. This segmentation allows each stage to optimize for specific requirements, achieving both high accuracy for small objects and maintained detection speed.
Solution Approach 2:
The patent transforms radar data from traditional point-cloud representation into multiple dimensional representations including Doppler beam vectors, range-azimuth maps, and interpolated range-azimuth maps. This dimensional transformation enables the system to capture subtle features of small stationary objects that are lost in point-cloud formats, improving detection accuracy while maintaining processing efficiency through specialized operations in each dimensional space.
2Measurement precision
If radar systems process data to improve detection accuracy of stationary objects, then detection precision improves, but processing complexity increases
Solution Approach 1:
The patent performs preliminary actions by pre-generating Doppler beam vectors from time-series frames and pre-computing range-azimuth maps using super-resolution operations before the actual stationary object detection. These preliminary processed data structures are then reused in the detection stage, improving accuracy while avoiding redundant computations that would increase processing complexity.
Solution Approach 2:
The patent introduces interpolated range-azimuth maps as an intermediary data structure between raw radar data and final stationary object detection. These interpolated maps serve as a mediator that bridges the gap between coarse radar measurements and fine-grained object detection requirements, improving detection precision without requiring the final detection stage to directly process complex raw data.
3Reliability
If radar systems use machine-learned models for stationary object detection, then detection reliability improves, but computational requirements increase
Solution Approach 1:
The patent performs preliminary processing to generate optimized input data structures (Doppler beam vectors, range-azimuth maps, interpolated range-azimuth maps) before feeding data to machine-learned models. This preliminary action ensures that the models receive pre-processed, high-quality inputs that require less computational energy to process while maintaining high detection reliability for autonomous driving applications.
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 improves the accuracy and speed of detecting smaller stationary objects, independent of vehicle velocity, enhancing the reliability of autonomous and semi-autonomous driving systems.
Implementation Method 1
receiving radar data that comprises multiple time-series frames associated with electromagnetic, EM, energy reflected by one or more objects in a roadway and received at an antenna of a radar system
Implementation Method 2
generating, using the time-series frames of the radar data, a Doppler beam vector of the EM energy
Data Source
Figure 1
Figure 2
Figure 3
AI summary
This document describes techniques and systems related to a radar system using a machine-learned model for stationary object detection. The radar system includes a processor that can receive radar data as time-series frames associated with electromagnetic (EM) energy. The processor uses the radar data to generate a range-time map of the EM energy that is input to a machine-learned model. The machine-learned model can receive as inputs extracted features corresponding to the stationary objects from the range-time map for multiple range bins at each of the time-series frames. In this way, the described radar system and techniques can accurately detect stationary objects of various sizes and extract critical features corresponding to the stationary objects.