Vehicle Radar Control With ML Screening and Trajectory Verification
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current radar data processing methods for autonomous vehicles face challenges in providing efficient and reliable object detection, particularly in large areas, and often require extensive software development to ensure safety and compliance with standards like ISO26262, which can be resource-intensive and limit the application of advanced machine learning technologies.
Innovation Solution
A computer-implemented method that combines machine learning and conventional processing methods for vehicle control, where machine learning is used for initial radar data processing in a larger area and conventional methods, such as discrete Fourier transform, verify detections in a smaller area along the vehicle's trajectory, allowing for efficient and safe braking decisions without the need for extensive validation and testing.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If machine learning methods are used for radar data processing in large areas, then detection efficiency and coverage are improved, but computational complexity and resource consumption increase
Solution Approach 1:
The patent divides the radar detection area into a first area (larger coverage) processed by machine learning methods and a second area (smaller subarea along trajectory) processed by conventional methods. This segmentation allows efficient use of ML for broad coverage while using simpler conventional methods for detailed verification in critical zones, reducing overall computational complexity.
Solution Approach 2:
Different processing methods are applied to different spatial regions: machine learning provides probabilistic occupancy grids for the entire first area, while conventional methods provide deterministic verification for the second area along the trajectory. This local differentiation optimizes resource allocation by applying computationally intensive methods only where needed.
2Reliability
If conventional methods are used for verification in smaller areas, then detection reliability and legal compliance are improved, but processing time and computational resources increase
Solution Approach 1:
The machine learning method first processes the entire first area to generate probabilistic occupancy information before the conventional method verifies specific regions. This preliminary action filters potential objects of interest, allowing the conventional verification method to focus only on relevant areas along the trajectory, thus reducing processing time.
Solution Approach 2:
Instead of applying conventional verification to the entire large area, the patent applies it only to the smaller second area along the trajectory where verification is most critical. This partial application of the verification method reduces computational overhead while maintaining reliability where it matters most.
3Measurement precision
If machine learning methods are applied to autonomous vehicle control, then object detection performance is improved, but software development effort and safety qualification requirements increase
Solution Approach 1:
The patent introduces an intermediary verification layer using conventional methods that bridges the machine learning detection system and the safety-critical control system. This intermediary performs deterministic verification of ML detections, reducing the safety qualification burden on the ML component while maintaining high detection performance.
Solution Approach 2:
The system segments the detection pipeline into an ML-based first processing stage for high-performance detection and a conventional second processing stage for safety-critical verification. This segmentation allows each component to be optimized and qualified independently, reducing overall software development effort.
Data Source
Figure 1
Figure 2
Figure 3
AI summary
A computer implemented method for controlling a vehicle comprises the following steps carried out by computer hardware components: acquiring sensor data from a sensor; determining first processed data related to a first area around the vehicle based on the sensor data using a machine learning method; determining second processed data related to a second area around the vehicle based on the sensor data using a conventional method, wherein the second area comprises a subarea of the first area; and controlling the vehicle based on the first processed data and the second processed data.