3D Ground Modeling for Drivable Surface Detection on Complex Terrain
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing systems struggle to accurately identify drivable surfaces in off-road or urban environments due to the failure of conventional methods in distinguishing between the ground and obstacles, especially when the ground is not planar or has a steep incline, leading to incorrect obstacle detection.
Innovation Solution
A method and system using a three-dimensional measurement sensor to capture data, compare it with a three-dimensional ground model, and determine a drivable area by fitting a smooth surface based on captured data, adjusting for slope thresholds and filling gaps with local slope information, and optionally using pre-generated ground models for improved accuracy.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If ground plane removal is used to identify obstacles, then obstacle detection works on planar road surfaces, but it fails in off-road environments or areas with ramps and curbs
Solution Approach 1:
The patent divides the environment into multiple ground models, each representing a specific terrain type (road, off-road, ramp, curb). Instead of using a single ground plane assumption, the system segments the detection space and applies appropriate ground models to different segments, enabling accurate obstacle detection across diverse environments.
Solution Approach 2:
The system dynamically selects and adapts ground models based on the detected environment. Rather than relying on a static planar ground assumption, the ground models are adjusted to match the actual terrain geometry, allowing the obstacle detection to remain accurate whether the vehicle is on a flat road, off-road terrain, or navigating ramps and curbs.
2Adaptability or versatility
If sensors are oriented to avoid measuring the ground, then planar ground identification is avoided, but low profile obstacles or obstacles in off-road environments cannot be correctly identified
Solution Approach 1:
The patent introduces three-dimensional ground models as intermediary structures between the sensor data and obstacle identification. These ground models serve as a reference framework that allows the system to measure and interpret ground terrain while simultaneously enabling the detection of low profile obstacles and off-road obstacles by comparing actual measurements against the expected ground model.
3Object-affected harmful factors
If ground measurements are filtered out as noise, then ground interference is reduced, but obstacles with low profile or in off-road environments are not correctly identified
Solution Approach 1:
Instead of uniformly filtering out all ground measurements as noise, the patent applies local quality differentiation by using three-dimensional ground models that capture the specific characteristics of different terrain types. The system identifies and processes ground regions versus obstacle regions based on their local properties, allowing low profile obstacles and off-road obstacles to be distinguished from the ground terrain in their respective locations.
4Measurement precision
If conventional ground plane methods are used, then simple planar surfaces are handled correctly, but steep inclines cause the ground to appear in the sensor field of view leading to incorrect identification
Solution Approach 1:
The patent transitions from two-dimensional ground plane assumptions to three-dimensional ground models that incorporate vertical dimension information. This dimensional expansion allows the system to accurately represent and handle steep inclines, ramps, and other complex terrain features that cannot be captured by planar surfaces, enabling correct drivable area identification on varied terrain.
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
Enhances the ability to accurately identify drivable areas by effectively distinguishing between ground and obstacles, even in complex terrains, thereby improving the safety and reliability of autonomous vehicles.
Implementation Method 1
The three-dimensional measurement sensor may be a LIDAR sensor
Data Source
AI summary
Systems and methods are provided for identifying a drivable area in front of a vehicle. A three-dimensional measurement sensor captures three-dimensional data corresponding to an area in front of the vehicle. Processing circuitry compares the captured three-dimensional data with a three-dimensional ground model to identify at least one obstacle and determine a drivable area in front of the vehicle based on the comparison.