Autonomous Vehicle Trajectory Planning via Static Image Extraction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current unmanned vehicle trajectory determination methods rely on extensive data processing, which is inefficient and prone to over-fitting due to the inclusion of dynamic objects, leading to suboptimal processing efficiency and accuracy.
Innovation Solution
The method involves acquiring a vehicle environment image, extracting a static environment image using image recognition technology, and using this static image as input for a trajectory planning model to plan the vehicle's trajectory, thereby avoiding over-fitting and reducing data requirements.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If extensive data processing is used for trajectory determination, then comprehensive environment information is obtained, but processing efficiency decreases and over-fitting occurs due to dynamic objects
Solution Approach 1:
The patent extracts only the static environment images from the complete vehicle environment images, separating the useful static background information from the dynamic objects that cause over-fitting. This extraction process removes distracting dynamic elements while preserving the essential static environmental context needed for accurate trajectory planning, thereby improving both accuracy and processing efficiency.
2Adaptability or versatility
If complete vehicle environment images are used for trajectory planning, then comprehensive environmental context is provided, but data volume increases leading to suboptimal processing efficiency
Solution Approach 1:
The system extracts only the static environment portions from complete vehicle environment images, removing dynamic objects that consume processing resources. This extraction maintains comprehensive environmental context for trajectory planning while significantly reducing the data volume that requires processing, thereby decreasing processing time without sacrificing adaptability.
Solution Approach 2:
The patent segments the vehicle environment image into static environment images and dynamic object regions, processing only the static portion for trajectory planning. This segmentation separates the permanent environmental structure from transient objects, allowing efficient processing of essential spatial information while ignoring time-varying elements that would increase processing time.
3Adaptability or versatility
If dynamic objects are included in trajectory determination data, then complete environment is captured, but over-fitting occurs reducing determination accuracy
Solution Approach 1:
The patent explicitly extracts and removes dynamic objects from the vehicle environment images, keeping only static environment images for trajectory determination. This extraction eliminates the source of over-fitting while preserving the complete static environmental context, thereby improving trajectory determination accuracy without sacrificing environmental completeness.
Solution Approach 2:
The system segments environment images into static and dynamic components, using only the static segment for trajectory planning. This segmentation strategy captures the complete environmental structure while excluding dynamic objects that cause over-fitting, thus resolving the contradiction between environment completeness and determination accuracy.
Data Source
AI summary
The present disclosure provides a method and a device for controlling a vehicle, a device and a storage medium, and relates to the field of unmanned vehicle technologies. The method includes: acquiring a vehicle environment image by an image acquirer during traveling of the vehicle; extracting a static environment image included in the vehicle environment image; obtaining a planned vehicle traveling trajectory by taking the static environment image as an input of a trajectory planning model; and controlling the vehicle to travel according to the planned vehicle traveling trajectory.