3D Drivable Area Detection for Curbs and Vertical Obstacles
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing vehicle detection systems struggle to consistently identify physical features like curbs, speed bumps, and other obstructions around the vehicle due to limitations in radar, camera-based 3D scene understanding, and the difficulty of validating such features using machine learning across diverse environmental conditions.
Innovation Solution
A vehicle drivable area detection system utilizing a 3D sensor to scan and capture data, an electronic controller to process this data, and generate digital renderings of drivable areas without relying on object recognition techniques, by extracting vertical obstacles, estimating terrain, and detecting curbs through geometric methods.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Length of stationary object
If radar is used for detection, then detection range is extended, but detection precision of physical features like curbs and speed bumps deteriorates
Solution Approach 1:
The system combines radar and camera sensors to create a unified detection system. Radar provides long-range detection capability while cameras provide detailed visual information for identifying physical features. The fusion of these different sensor types allows the system to achieve both extended detection range and high precision in detecting curbs, speed bumps, and other road features.
2Measurement precision
If camera is used for detection, then detection of physical features is improved, but reliability under adverse environmental conditions deteriorates
Solution Approach 1:
Radar serves as an intermediary sensor that operates effectively in adverse environmental conditions (darkness, inclement weather) where camera performance degrades. The radar provides complementary detection data that compensates for camera limitations, ensuring reliable detection of physical features across all weather and lighting conditions.
3Adaptability or versatility
If machine learning is used for object recognition, then object identification capability is improved, but validation difficulty across global applications increases
Solution Approach 1:
The system replaces complex machine learning-based object recognition with a geometric analysis approach that uses 3D spatial relationships and physical feature characteristics. This substitution maintains the ability to identify objects and features while significantly reducing validation complexity, as geometric algorithms can be universally applied without requiring region-specific training data or validation.
4Loss of information
If 3D sensor data is processed using object recognition techniques, then scene understanding is improved, but system complexity increases
Solution Approach 1:
The system extracts only the essential geometric and spatial information needed for drivable area detection from the 3D sensor data, rather than performing comprehensive object recognition. By focusing specifically on extracting relevant geometric features and spatial relationships, the system achieves effective scene understanding while maintaining computational efficiency and reducing overall system complexity.
Applied Scientific Principles
This section explains which scientific principles are used to turn an abstract innovation direction into a practical engineering solution.
Function Achieved in This Case
Enables reliable detection of drivable areas and obstructions under various conditions, enhancing driver assist components with improved obstacle and terrain awareness, independent of environmental factors and without the need for specific training on diverse features.
Implementation Method 1
a 3D sensor installed to the vehicle and being configured to scan and capture data using laser imaging, detection and distance ranging
Data Source
AI summary
A vehicle drivable area detection system includes a vehicle, at least one 3D sensor and an electronic controller. The at least one 3D sensor is installed to the vehicle and is configured to scan and capture data using laser imaging, detection and distance ranging relative to the vehicle. The data collected represents ground surface features including vertical obstacles, non-vertical obstacles and a drivable area proximate the vehicle within a line-of-sight of the 3D sensor. The electronic controller is installed within the vehicle and is electronically connected to the 3D sensor and at least one driver assist component. The electronic controller conducts the following: a vertical obstacle extraction from the data; terrain estimating from the data; curb detection from the data; and generating a plurality of data elements identifying vertical obstacles including curbs and the drivable area to the at least one driver assist component.