Imaging Radar 5D Neural Processing for Adverse-Weather Detection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Autonomous devices face limitations in perception and navigation due to the poor performance of light-based sensors in adverse weather conditions, while radar systems are less affected but require improved object detection and classification methods for accurate environmental awareness.
Innovation Solution
Integration of a 5D deep neural network (DNN) architecture that processes four-dimensional radar reception data as sequential input, considering 3D spatial features and radial velocity over time to localize and classify dynamic objects, such as pedestrians, and their activities, using two concatenated deep neural networks to learn spatial and temporal features.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If light-based sensors (cameras, LIDAR) are used for autonomous perception, then high-resolution environmental mapping is achieved, but performance deteriorates under poor visibility and inclement weather conditions
Solution Approach 1:
The patent combines multiple sensor types (light-based sensors and radar systems) into an integrated perception system. The camera and LIDAR provide high-resolution environmental mapping, while the radar system provides reliable detection in adverse weather conditions. The fusion of these sensor modalities allows the autonomous device to maintain both high mapping precision and reliable operation across all weather conditions.
2Reliability
If radar systems are used for autonomous perception, then reliability in adverse weather conditions is improved, but object detection and classification accuracy deteriorates without advanced processing methods
Solution Approach 1:
The patent introduces a radar processing device as an intermediary component that bridges the radar system and the autonomous device controller. This processing device performs sophisticated signal processing, including FFT operations, Doppler processing, and object classification algorithms, to transform raw radar returns into accurate object detections and classifications. The intermediary processing layer enables the radar system to achieve high measurement precision while maintaining its reliability in adverse weather conditions.
3Measurement precision
If complex radar signal processing is implemented to improve object classification, then detection accuracy is improved, but processing time and computational complexity increase
Solution Approach 1:
The patent implements preliminary signal processing operations directly in the radar processing device, performing FFT transformations and Doppler frequency analysis before data is passed to the autonomous device controller. By pre-processing the radar signals and extracting key features (range, velocity, angle) beforehand, the system reduces the computational burden on the main controller and enables faster object classification decisions, thereby reducing overall processing time while maintaining high detection accuracy.
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
Enables rapid and accurate detection and classification of objects and activities, improving environmental awareness in complex scenarios like dense urban areas with low latency, suitable for autonomous vehicles and robots.
Implementation Method 1
Radio detection and ranging (Radar) systems, however, may be comparatively undisturbed by inclement weather and/or reduced visibility.
Implementation Method 2
The radar processor may transmit a plurality of transmit wireless signals within an environment and receive a plurality of receive wireless signals including reflections of the transmit wireless signals
Data Source
AI summary
According to various embodiments, a radar device is described comprising a processor configured to generate a scene comprising an object based on a plurality of receive wireless signals, generate a ground truth object parameter of the object and generate a dataset representative of the scene and a radar detector configured to determine an object parameter of the object using a machine learning algorithm and the dataset, determine an error value of the machine learning algorithm using a cost function, the object parameter, and the ground truth object parameter and adjust the machine learning algorithm values to reduce the error value.