Scene-Based Path Planning for Autonomous Vehicles
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current autonomous driving technologies face inefficiencies in real-time path planning, as they often rely on standard reference lines that may not be optimal for safe and comfortable driving, and require dynamic calculations that consume resources and time.
Innovation Solution
Implementing a scene-based driving guideline that uses preconfigured path segments based on driving behaviors of various drivers in similar circumstances, stored in scene/path data structures, allowing for efficient path planning by matching route segments with predefined driving scenes and using preconfigured path segments when possible.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If standard reference lines are used for path planning, then the path planning process is simple, but the driving safety and comfort are reduced
Solution Approach 1:
The patent pre-calculates and stores optimal path segments for various driving scenes (straight, turning, lane changing) in a scene/path data structure before real-time operation. During actual path planning, the system only needs to match the current scene with pre-stored segments, avoiding complex real-time calculations while ensuring optimal safety and comfort.
2Reliability
If dynamic path calculations are performed in real-time, then the path optimality is improved, but the processing time and resource consumption increase
Solution Approach 1:
Optimal path segments are pre-calculated and stored in the scene/path data structure during system initialization or offline processing. The scene-based path planning module retrieves and combines these pre-stored segments during real-time operation, achieving optimal paths without real-time calculation overhead.
Solution Approach 2:
The patent divides the path planning process into discrete scene-based segments (straight, turning, lane changing). Each segment type has pre-calculated optimal paths stored in the data structure. The system segments the overall route into these standard scenes and combines the corresponding pre-stored segments, avoiding the need for continuous dynamic recalculation.
3Productivity
If standard reference lines are used, then the calculation speed is fast, but the driving comfort is reduced
Solution Approach 1:
Comfort-optimized path segments for different driving scenes are pre-calculated and stored in the scene/path data structure. These segments incorporate comfort considerations such as smooth transitions and natural driving patterns. During real-time operation, the system quickly retrieves these pre-optimized segments based on scene matching, achieving both high calculation speed and driving comfort.
Data Source
AI summary
According to one aspect, a set of driving scenes are defined including, but are not limited to, a driving straight scene, a turning scene, a nudge scene, and a lane changing scene. For each of the predefined driving scenes, a list of path segments corresponding to a set of one or more features (e.g., speed, curvature, turning radius, relative distance, lane width, etc.) are determined based on the driving statistics of a number of vehicles driving through the same or similar driving scenes (e.g., driving straight, making turns, changing lanes, nudging obstacles or another vehicle) in the past. The path segments are then stored in a driving scene-to-path (scene/path) data structure specifically configured for the corresponding driving scene. For example, there will be at least one scene/path data structure for each of the driving straight scene, turning scene, nudge scene, and lane changing scene.