Radar Filter Parameter Control for Unreliable Multi-Target Scenes
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing radar systems face challenges in reliably tracking multiple targets, particularly when the scene includes an unexpected number of objects, leading to unreliable tracking results due to suboptimal parameter settings in digital filters.
Innovation Solution
A radar device equipped with a machine learning logic that includes a policy network to set digital filter parameters and a critic network to provide reward values, detecting out-of-distribution scenes to ensure reliable tracking by analyzing the distribution of reward values.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If digital filter parameters are set using conventional methods, then the system is simple to operate, but tracking reliability deteriorates in scenes with unexpected numbers of objects
Solution Approach 1:
The patent implements feedback by using a critic network that evaluates the output of the policy network and provides reward values. The system analyzes the distribution of reward values from multiple heads in the critic network to detect out-of-distribution scenes, then feeds this information back to adjust filtering parameters or trigger alerts, thereby improving tracking reliability in complex scenarios.
Solution Approach 2:
The patent changes parameters by using a policy network to dynamically adjust digital filter parameters based on the input scene characteristics. Instead of using fixed conventional parameters, the system learns optimal parameters through machine learning and adapts them in real-time, improving reliability while managing complexity through automated parameter optimization.
2Measurement precision
If machine learning logic with multiple critic heads is implemented, then tracking accuracy in complex scenarios improves, but device complexity increases
Solution Approach 1:
The patent applies segmentation by dividing the critic network into multiple independent heads, each evaluating different aspects of the tracking performance. This segmentation allows the system to analyze specific failure modes separately and combine their assessments to detect out-of-distribution scenes more accurately, improving tracking precision while maintaining manageable complexity through modular architecture.
3Adaptability or versatility
If conventional digital filtering is used, then the system is computationally efficient, but it cannot adapt to varying numbers of targets in different scenes
Solution Approach 1:
The patent implements dynamics by making the filter parameters adaptive rather than static. The policy network dynamically adjusts parameters based on the current scene characteristics, allowing the system to adapt to varying numbers of targets. This dynamic adaptation improves scene versatility while the incremental nature of the adjustments helps manage computational energy consumption.
Data Source
AI summary
A radar device includes a radar front end configured to send radar signals and to receive reflected radar signals, processing circuitry configured to provide digital radar data based on the received reflected radar signals, and a digital filter configured to process the digital radar data to obtain information about objects which reflected the radar signals. The device further comprises machine learning logic with a policy network configured to set the parameters of the digital filter based on the digital radar data, and a reward value generating network including a plurality of heads, each head configured to provide a respective expected reward value for a setting of parameters by the policy network. The radar device is further configured to detect that a scene captured by the radar device is not reliably processable based on a distribution of the expected reward values generated by the plurality of heads.