VRU Collision Avoidance via Predictive Path Overlap
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current VRU warning systems fail to accurately distinguish between harmless objects and Vulnerable Road Users (VRUs) in a vehicle's blind zone, leading to false warnings and a lack of effective collision avoidance, especially in scenarios where VRUs are temporarily immobile or obscured.
Innovation Solution
A method and system that predicts the future path of both the vehicle and VRUs using camera and sensor data, determining the probability of collision by measuring vehicle velocity, steering wheel angle, and VRU position, and issuing warnings or taking action only when a collision is probable, thereby reducing false alarms and enhancing traffic safety.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If simple ultrasonic sensors are used to detect objects in blind zones, then the system can identify the presence of objects, but it generates false warnings by unable to distinguish between harmless objects and VRUs
Solution Approach 1:
The patent introduces camera-based vision systems as intermediary devices to capture visual information about detected objects. The camera images serve as a mediator between the ultrasonic sensor detection and the final classification decision, enabling the system to distinguish VRUs from harmless objects through image analysis while maintaining the benefits of ultrasonic detection coverage
Solution Approach 2:
The patent replaces the purely mechanical/acoustic ultrasonic detection system with a hybrid system that incorporates optical/camera-based detection. This substitution enables visual classification of objects, allowing the system to differentiate between VRUs and stationary objects like lamp posts, thereby reducing false warnings while maintaining detection capability
2Reliability
If path prediction is made too restrictive to reduce false warnings, then fewer false alarms occur, but dangerous situations may be ignored or delayed
Solution Approach 1:
The patent implements dynamic path prediction that adapts to the detected VRU behavior and environmental context. The prediction model adjusts its restrictiveness based on real-time conditions, such as VRU movement patterns, road geometry, and traffic conditions, allowing the system to maintain high accuracy while responding promptly to dangerous situations without unnecessary delays
Solution Approach 2:
The system incorporates feedback loops where the outcomes of path predictions and warning actions are continuously monitored and used to refine future predictions. This feedback mechanism allows the system to learn from actual VRU behaviors and environmental conditions, optimizing the balance between reducing false warnings and maintaining rapid response to genuine dangers
3Reliability
If the system warns for all detected objects to ensure safety, then no dangerous situations are missed, but driver trust is lost due to excessive false warnings
Solution Approach 1:
The patent applies different detection and warning strategies to different spatial zones around the vehicle. High-priority monitoring with immediate warnings is applied to blind zones where VRUs are most dangerous, while less restrictive monitoring is applied to areas with lower risk. This localized quality approach ensures critical safety is maintained while reducing unnecessary warnings in lower-risk areas, preserving driver trust
Data Source
Figure 1~2
Figure 3
Figure 4A~4B
AI summary
Method (600) and control unit (310), for avoiding a potential collision between the vehicle (100) and a Vulnerable Road User, VRU (200). The method (600) comprises: predicting (601) a future path (t1, t2, t3) of the vehicle (100); detecting (602) the VRU (200) and the position of the VRU (200); determining (603) velocity of the detected (602) VRU (200); predicting (604) a future position (210) of the detected (602) VRU (200), based on the VRU position upon detection (602) and the determined (603) VRU velocity; and performing (607) an action for avoiding a collision, when the predicted (604) future position (210) of the VRU (200) is overlapping (220) the predicted (601) future path (t1, t2, t3) of the vehicle (100).