Autonomous Vehicle Path Planning for Static Obstacle Avoidance
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Autonomous vehicles face challenges in generating paths that balance passenger comfort and safety while avoiding static obstacles, particularly in maintaining proximity to the lane centerline and avoiding potential collisions.
Innovation Solution
A path planning system utilizing a station-lateral (SL) coordinate system and nonlinear optimization to generate control signals for autonomous driving vehicles, considering cost functions that minimize cumulative lateral distances, first-order, and second-order lateral rates of change, while adhering to constraints such as threshold lateral jerk and static obstacle avoidance.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of operation
If the vehicle maintains proximity to the lane centerline, then passenger comfort is improved, but collision risk with static obstacles increases
Solution Approach 1:
The path planning system applies different lateral distance preferences to different spatial regions. In regions without static obstacles, the vehicle is guided to maintain proximity to the lane centerline for comfort. In regions with static obstacles, the system locally adjusts the preferred lateral position to increase distance from obstacles, creating zone-specific path guidance that balances comfort and safety.
Solution Approach 2:
The preferred lateral position is dynamically adjusted based on the presence and position of static obstacles. The path planning system continuously modifies the desired trajectory by computing adjusted lateral offsets from the lane centerline when obstacles are detected, allowing the vehicle to transition smoothly between comfort-oriented and safety-oriented positioning.
2Reliability
If the vehicle avoids static obstacles by adjusting lateral position, then collision risk is reduced, but passenger comfort deteriorates
Solution Approach 1:
The system applies obstacle avoidance adjustments only partially - only when and where static obstacles are present. The lateral position adjustment is excessive only to the extent necessary to clear the obstacle, rather than maintaining a constant offset. This partial application minimizes comfort degradation while ensuring safety.
Solution Approach 2:
The path planning system proactively adjusts the lateral position in advance of potential collision risks by incorporating static obstacle information into the preferred path computation. This preliminary adjustment prevents the need for abrupt corrective maneuvers that would degrade passenger comfort.
3Reliability
If the path deviates from the lane centerline, then obstacle avoidance is improved, but path optimality deteriorates
Solution Approach 1:
The system modifies the path parameters (lateral position, curvature) by computing adjusted offsets from the ideal lane centerline path. These parameter changes are calculated to minimize deviation from the optimal path while ensuring adequate clearance from static obstacles, thereby maintaining path optimality under safety constraints.
Solution Approach 2:
The path planning problem is transformed from a two-dimensional path following task to a three-dimensional optimization problem by incorporating the lateral offset dimension. The preferred path is computed as a series of adjusted lateral positions relative to the lane centerline, allowing the system to navigate around obstacles while minimizing overall path deviation.
Data Source
AI summary
A method of generating a path for an autonomous driving vehicle (ADV) includes obtaining a plurality of path inputs including a lateral and a longitudinal starting state, a threshold lateral jerk, and a set of static obstacle boundaries with respect to a reference line, obtaining a plurality of path constraints related to the threshold lateral jerk, avoidance of static obstacles, and a threshold lateral velocity, obtaining a cost function associated with a path objective, the cost function comprising first, second, and third terms relating to cumulative lateral distances, to cumulative first order lateral rates of change, and to cumulative second order lateral rates of change, respectively, generating a plurality of planned ADV states as path results based on the plurality of path inputs, the plurality of path constraints, and the cost function and generating control signals to control the ADV based on the plurality of planned ADV states.