Drivable Space Detection Using Unsupervised Driving Behavior Learning
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
There is a growing need to autonomously distinguish between drivable spaces and non-drivable spaces while driving, as the distinction can be tricky, especially in environments with multiple examples of both.
Innovation Solution
A method, system, and non-transitory computer readable medium for providing drivable spaces alerts, which involves obtaining visual and vehicle behavior information, processing it to classify spaces and behaviors, and generating identifiers for drivable and non-drivable spaces, allowing for autonomous determination and alerting of drivable spaces.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If autonomous distinction between drivable and non-drivable spaces is implemented, then safety and navigation accuracy are improved, but system complexity increases
Solution Approach 1:
The system segments the driving environment into discrete path portions and classifies each as drivable or non-drivable space. This segmentation approach allows the complex autonomous distinction task to be broken down into manageable classification decisions for individual path segments, improving reliability without overwhelming system complexity
Solution Approach 2:
The system adds a classification dimension to path perception by generating drivable space identifiers that mark specific path portions. This dimensional addition transforms raw path data into structured drivability information, enabling safe navigation through enhanced spatial understanding
2Measurement precision
If multiple examples of drivable and non-drivable spaces are distinguished, then navigation accuracy is improved, but difficulty of detecting and measuring increases
Solution Approach 1:
The system applies local quality classification to individual path portions rather than treating the entire path uniformly. Each path segment is independently analyzed and tagged with drivability characteristics, enabling precise navigation decisions for multiple drivable and non-drivable spaces while managing detection complexity through localized processing
3Speed
If drivable space identification is performed in real-time, then responsiveness is improved, but processing speed requirements increase
Solution Approach 1:
The system performs preliminary classification of path portions as drivable or non-drivable spaces before the vehicle reaches them. By pre-identifying and tagging path segments with drivability identifiers, the system prepares navigation decisions in advance, achieving real-time responsiveness without peak processing demands during critical driving moments
Data Source
AI summary
A method for unsupervised learning of drivable space, the method may include receiving, by a processing circuit, an image of an environment of a vehicle; searching, in the image and by the processing circuit, for a drivable space within the environment of the vehicle, wherein the drivable space is associated with a statistically significant drivable space behavior; wherein the association is learnt by applying an unsupervised learning process; and responding, by the processing circuit, to a finding of the drivable space.