Vehicle Collision Avoidance Using Virtual Boundary Detection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current vehicle collision avoidance systems require extensive data and computations to predict and mitigate collisions, often relying on machine learning programs that are complex and resource-intensive, making them less efficient in determining whether to steer or brake.
Innovation Solution
A system that identifies lateral and forward virtual boundaries based on sensor data, determining constraint values to quickly and efficiently decide on steering and braking actions by decoupling lateral and longitudinal movements, allowing for faster and more accurate collision avoidance without the need for extensive data processing.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If machine learning programs are used to predict and mitigate collisions, then collision avoidance accuracy is improved, but system complexity and computational requirements increase
Solution Approach 1:
The patent segments the collision avoidance problem into two independent components: lateral boundary identification and forward boundary identification. By decoupling lateral and longitudinal movements, the system processes each dimension separately using simplified linear equations rather than complex machine learning models, reducing computational complexity while maintaining accuracy
Solution Approach 2:
The patent extracts and removes the complex machine learning computation layer from the collision avoidance system. Instead, it uses direct mathematical calculations based on sensor data to determine virtual boundaries and constraint values, eliminating the need for resource-intensive neural networks while preserving collision prediction capability
2Measurement precision
If extensive data processing is performed to determine collision avoidance actions, then decision accuracy is improved, but processing time increases
Solution Approach 1:
The patent replaces the mechanical data processing system (extensive computations and iterations) with a more efficient mathematical approach. By using direct calculation methods to compute virtual boundaries and constraint values from sensor inputs, the system achieves accurate collision avoidance decisions with minimal processing time, eliminating the need for iterative machine learning inference
3Reliability
If machine learning programs are used for collision prediction, then prediction capability is improved, but computational resources required increase
Solution Approach 1:
The patent employs computationally inexpensive mathematical calculations instead of resource-intensive machine learning models. By using simple linear equations to determine virtual boundaries and constraint values, the system achieves effective collision prediction with minimal computational resource consumption, making it suitable for real-time implementation in resource-constrained vehicular environments
Data Source
AI summary
A lateral virtual boundary for a host vehicle is identified based on a lateral distance between the host vehicle and a target vehicle, a longitudinal distance between the host vehicle and the target vehicle, and a speed of the target vehicle relative to the host vehicle. A forward virtual boundary for the host vehicle is identified based on the longitudinal distance between the host vehicle and the target vehicle. A lateral constraint value of the lateral virtual boundary and a forward constraint value of the forward virtual boundary are determined. A longitudinal acceleration and a steering angle are determined based on the lateral and forward virtual boundaries and the lateral and forward constraint values. One or both of a steering component or a brake are actuated based on the longitudinal acceleration and the steering angle.