Forward Collision Warning System Roadside Target Filtering
Solution Overview
Problem
Forward Collision Warning (FCW) systems often issue false warnings due to the inability to accurately distinguish between stationary roadside objects and potential collision threats, especially in curved road sections, where path prediction errors are prevalent, leading to decreased system reliability and potential driver habituation or system shutdown.
Innovation Solution
The method involves continuously measuring vehicle parameters such as steering input angle change rate and yaw rate to identify a Collision Threat Zone (CTZ) and apply width reductions to suppress false warnings, using a combination of look-back periods and threshold values to determine when to reduce the CTZ width, potentially to zero, thereby filtering out roadside objects as non-threats.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Device complexity
If radar-only detection is used without computer vision camera, then system complexity is reduced, but false warnings increase due to inability to classify stationary objects
Solution Approach 1:
The patent changes the parameters used for target classification from static radar data to dynamic trajectory parameters. By analyzing movement patterns, acceleration profiles, and positional changes over time, the system dynamically adjusts classification criteria to distinguish stationary roadside objects from moving vehicles without requiring additional sensors
Solution Approach 2:
The patent introduces trajectory analysis as an intermediary processing layer between radar detection and warning generation. This intermediate step processes raw radar positions to extract movement characteristics, serving as a mediator that enables vehicle classification without the computer vision camera
2Measurement precision
If path prediction is performed on curved roads, then collision detection accuracy is improved, but false warnings increase due to path prediction errors
Solution Approach 1:
The patent applies dynamic threshold adjustment based on road curvature and vehicle motion state. The clearance threshold is not fixed but adapts according to yaw rate, steering angle, and curve radius, allowing the system to maintain accurate collision detection while compensating for path prediction uncertainties on curved roads
Solution Approach 2:
The patent performs preliminary trajectory verification by checking if detected objects follow expected motion patterns before issuing warnings. The system pre-calculates valid trajectory ranges based on road geometry and vehicle dynamics, filtering out objects that fall outside these predicted paths before generating false warnings
3Measurement precision
If lateral clearance threshold is lowered to detect closer objects, then detection sensitivity is improved, but false warnings increase from roadside objects
Solution Approach 1:
The patent applies different clearance thresholds for different spatial zones and object types. Instead of a uniform threshold, the system adjusts lateral clearance requirements based on object classification, road curvature, and position relative to the vehicle path, allowing sensitive detection in safe zones while filtering roadside objects in critical zones
Data Source
AI summary
A method of filtering/rejecting targets detected by a forward collision warning system of a motor vehicle when entering a curved road segment. The width of a collision threat zone (CTZ) is reduced in one or more stages depending upon continuously-measured values of steering input angle (SIA), driver brake application (DBA), and a yaw rate. The measured values are used to find: a degree of near-past DBA variation during a first look-back period, a degree of near-past SIA change rate during a second look-back period, and a degree of far-past yaw rate change during a third look-back period longer than the first and the second look-back periods. A series of first, second and third width reductions of decreasing severity are applied to the CTZ based upon comparisons of the tracked variations and change rates with respective thresholds. The thresholds are tunable to achieve desired levels or false-target rejection.