Hierarchical Path Planning for Moving-Obstacle Lane Boundary Shifts
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing autonomous driving systems struggle to handle complex scenarios, such as moving obstacles and emergencies, often resulting in stalled path decisions or hard-braking instead of proper dodging, limiting their ability to navigate safely and comfortably.
Innovation Solution
A hierarchical lane boundary determination system dynamically transitions between different lane boundary schemes to plan robust trajectories, evaluating and adjusting paths based on safety rules and obstacle movement predictions to ensure safe navigation.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If an existing path decision module is used, then simple scenarios can be handled, but complex scenarios such as moving obstacles and emergencies result in stalled path decisions or hard-braking
Solution Approach 1:
The system dynamically transitions between different lane boundary determination schemes (first scheme with strict lane boundaries, second scheme with relaxed boundaries) based on the driving scenario. This dynamic adaptation allows the system to handle both simple and complex scenarios reliably, avoiding stalled decisions or hard-braking by selecting the appropriate scheme for each situation.
2Manufacturing precision
If a first lane boundary determination scheme is used, then lane boundary accuracy is improved, but trajectory flexibility deteriorates in complex scenarios
Solution Approach 1:
The lane boundary determination is segmented into multiple schemes: a first scheme for normal conditions with strict lane boundary adherence, and a second scheme for complex scenarios with relaxed boundaries. This segmentation allows the system to maintain high accuracy when applicable while gaining flexibility when needed, resolving the contradiction between precision and adaptability.
Solution Approach 2:
The system dynamically switches between the first lane boundary determination scheme (high accuracy, low flexibility) and the second scheme (lower accuracy, high flexibility) based on scenario complexity. This dynamic transition enables the system to optimize the trade-off between boundary accuracy and trajectory flexibility for each specific driving situation.
3Reliability
If path decision is made to be conservative, then safety is improved, but navigation performance and comfort deteriorate
Solution Approach 1:
The system dynamically adjusts its conservatism level by switching between lane boundary determination schemes. In normal scenarios, it uses the first scheme with strict boundaries for high safety. In complex scenarios with moving obstacles, it transitions to the second scheme with relaxed boundaries, enabling agile maneuvers that improve navigation performance and comfort while maintaining safety through continuous evaluation.
Data Source
AI summary
According to one embodiment, during a first planning cycle, a first lane boundary of a driving environment perceived by an ADV is determined using a first lane boundary determination scheme (e.g., current lane boundary), which has been designated as a current lane boundary determination scheme. A first trajectory is planned based on the first lane boundary to drive the ADV to navigate through the driving environment. The first trajectory is evaluated against a predetermined set of safety rules (e.g., whether it will collide or get too close to an object) to avoid a collision with an object detected in the driving environment. In response to determining that the first trajectory fails to satisfy the safety rules, a second lane determination boundary of the driving environment is determined using a second lane boundary determination scheme and a second trajectory is planned based on the second lane boundary to drive the ADV.