Multi-Layer Grid Path Planning for Straight-Back Parking
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional path planning algorithms for autonomous driving vehicles, such as the A-star searching algorithm, often result in control errors during complex maneuvers like parking due to the inability to revisit nodes, leading to curved backward movements that are difficult to follow.
Innovation Solution
A multi-layer grid searching algorithm is introduced, allowing each node in the grid to be visited twice by using a first and second layer grid, where the second layer grid corresponds to the first layer grid, enabling the vehicle to plan a path that includes backward movements with a straight line trajectory, reducing control errors.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of operation
If A-star searching algorithm is used for path planning, then the vehicle can navigate to various locations using onboard sensors, but the trajectory generated may be hard to follow and result in control errors due to curved backward movements
Solution Approach 1:
The patent segments the path planning into two distinct layers: a first layer for forward movements and a second layer for backward movements. This segmentation allows the algorithm to generate trajectories that are easier to follow by separating curved forward paths from straight backward paths, thereby reducing control errors while maintaining navigation capability.
2Adaptability or versatility
If conventional single-layer grid algorithm is used, then the path planning is computationally simpler, but the vehicle cannot revisit nodes which limits maneuverability in complex scenarios like parking
Solution Approach 1:
The patent introduces an additional dimension by creating a second layer grid that corresponds to the first layer. This dimensional expansion allows the vehicle to revisit positions by transitioning between layers, enabling complex maneuvers like parking while managing algorithmic complexity through structured layer correspondence.
3Manufacturing precision
If the algorithm allows backward movement with curve, then the vehicle can reach target position, but the last parking position requires extra attention and may not be acceptable
Solution Approach 1:
The patent applies different movement characteristics to different segments of the trajectory: forward movements can follow curved paths while backward movements are constrained to straight lines. This local differentiation ensures precise and acceptable final parking positions by making the backward approach segment easier to control and follow.
Data Source
AI summary
Path planning is performed using a multi-layer grid searching algorithm to position an ADV in a target position. A first layer grid including a first set of nodes is defined. A second layer grid is defined. The second layer grid includes a second set of nodes corresponding to at least a portion of the first set of nodes. From a start node until a goal node, following operations are iteratively performed. A set of next node candidates are identified by searching in the first set of nodes and the second set of nodes. For each next node candidate of the set of next node candidates, a cost is determined using a cost function. A next node having a lowest cost is selected from the set of next node candidates based on their respective costs. A path trajectory of the ADV is generated to position the ADV at the target position.