Autonomous Driving Path Generation Region Expansion
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Autonomous driving systems face challenges in generating and selecting avoidance paths in real-time due to varying road conditions and vehicle behavior characteristics, which can lead to unsafe collisions with obstacles.
Innovation Solution
A method and apparatus for controlling an autonomous driving system that determines potential collisions with obstacles, generates multiple path candidates within an expanded path generation region, and selects the path with the lowest collision risk, adjusting vehicle velocity to avoid obstacles while maintaining the reference path if expansion is not possible.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If the path generation region is expanded to generate more path candidates, then the ability to avoid obstacles is improved, but the computational complexity and processing time increase
Solution Approach 1:
The path generation region is divided into multiple sub-regions or segments, and path candidates are generated in a staged manner. The system first generates paths within the original lane boundaries, then progressively expands to adjacent lanes or regions only when necessary, breaking down the complex global path generation into manageable local segments.
Solution Approach 2:
The system pre-establishes the reference path and generates initial path candidates within the standard path generation region before actual obstacle detection occurs. When an obstacle is detected, the system can quickly expand the region and generate additional candidates based on pre-computed road geometry and environmental data, reducing real-time computational burden.
2Reliability
If multiple path candidates are generated and evaluated in real-time, then the selection of optimal avoidance path is improved, but the processing time and computational load increase
Solution Approach 1:
The system dynamically adjusts evaluation parameters such as collision risk threshold, path smoothness weight, and velocity adjustment ratio based on the specific situation. By changing these parameters adaptively, the system can quickly filter out suboptimal paths and focus evaluation on the most promising candidates, reducing processing time while maintaining selection accuracy.
Solution Approach 2:
The system implements a hierarchical evaluation process where paths are quickly screened based on basic criteria (collision freedom, lane boundary compliance), and only paths passing this initial filter undergo detailed evaluation. This allows the system to rapidly eliminate clearly inferior options and rush through the evaluation of promising candidates.
3Reliability
If the vehicle velocity is adjusted dynamically to avoid obstacles, then the collision avoidance effectiveness is improved, but the driving stability and comfort may deteriorate
Solution Approach 1:
The system implements dynamic velocity adjustment that adapts to the specific avoidance scenario. Instead of applying fixed deceleration or stop commands, the velocity profile is dynamically calculated based on the selected path, obstacle distance, vehicle state, and road conditions. This allows smooth, context-appropriate velocity changes that maintain stability while achieving avoidance.
Solution Approach 2:
The system applies velocity adjustment in advance before the vehicle reaches critical proximity to the obstacle. By cushioning the velocity change early in the avoidance maneuver rather than making abrupt last-minute adjustments, the system maintains driving stability and comfort while still achieving effective collision avoidance.
Data Source
AI summary
A method for controlling a path of an autonomous driving system includes: determining whether or not an obstacle and an autonomously driven vehicle will collide when a sensor detects an obstacle in a proximity of the vehicle while the vehicle is being autonomously driven; generating a plurality of path candidates within a path generation region when it is determined that the detected obstacle and the vehicle will collide; determining whether or not an expansion of the path generation region is possible, expanding the path generation region when it is possible, and regenerating the plurality of path candidates within the expanded path generation region when there is no path candidate in which the collision between the obstacle and the vehicle does not occur; and selecting a path from among the path candidates in which the collision between the obstacle and the vehicle does not occur.