Traversable Path Detection Using Smoothed Clusters and Open Polygons
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional methods for determining a traversable path for vehicles, such as vision-based road detection and edge detection, are limited to structured roads and require complex algorithm training, making them inefficient and unreliable for both structured and unstructured environments.
Innovation Solution
A method and apparatus that identify objects in a vehicle's environment, determine cluster points, smooth these points to form smoothed clusters, and specify the traversable path boundaries as open polygons by intersecting perpendicular lines with cluster boundaries, allowing for robust detection suitable for various road conditions.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If vision-based road detection techniques or edge detection and histograms are used to identify lane boundaries, then the system can control vehicle lane-keeping behavior, but these approaches are only suitable for structured roads and require complex algorithm training
Solution Approach 1:
The patent changes the fundamental parameters of detection from vision-based pixel analysis to sensor-based point cloud analysis. By transforming the detection space from 2D image coordinates to 3D spatial coordinates and using statistical parameters (standard deviation, distance metrics) to characterize point distributions, the system achieves reliable detection across both structured and unstructured roads without requiring complex training algorithms
Solution Approach 2:
The patent replaces the mechanical/vision-based detection system with a sensor-based system using laser range finders or ultrasonic sensors. This substitution eliminates the need for complex image processing and training algorithms while providing robust detection in various road conditions, as the sensor directly measures spatial positions of road boundaries
2Productivity
If conventional vision-based methods are used for road detection, then the system can identify lane boundaries on structured roads, but the approach becomes inefficient and unreliable for unstructured environments
Solution Approach 1:
The patent replaces inefficient vision-based image processing with direct sensor measurements of spatial positions. The sensor-based approach efficiently captures point cloud data that directly represents road boundaries, eliminating the need for complex image processing algorithms and providing reliable detection in unstructured environments where vision-based methods fail
3Measurement precision
If complex algorithm training is applied to obtain reliable boundary detection results, then detection accuracy improves, but the system complexity and training requirements increase
Solution Approach 1:
The patent replaces complex trained algorithms with straightforward statistical analysis of sensor data. By using standard deviation and distance metrics to characterize point cloud distributions, the system achieves accurate boundary detection without requiring any training phase, significantly reducing system complexity while maintaining high measurement precision
Solution Approach 2:
The sensor-based system is self-sufficient and does not require external training data or training processes. The algorithm automatically adapts to different road conditions by computing statistical parameters from the current sensor data, making the system both accurate and simple to operate in diverse environments
Data Source
Figure 1~2
Figure 3
Figure 4~5
AI summary
The present invention provides a determination of a traversable path for a vehicle. It is for this purpose, that elements in the environment of the vehicle are identified and further processed as smoothed clusters. The smoothed clusters may be considered as cluster points with a probability distribution having a particular diameter. Based on the smoothed clusters, the boundaries of the traversable path may be specified as open polygons. The vertices of the open polygons may be identified by intersecting the smoothed clusters with an axis perpendicular to an estimated movement of the vehicle.