Point Cloud Free Space Estimation Under Sensor Noise
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Autonomous vehicles face challenges in efficiently determining free space due to sensor noise, limitations, and varying environmental conditions, which affect the accuracy of obstacle detection and navigation.
Innovation Solution
A system and method that process point cloud data from sensors to calculate the probability of obstacles by considering sensor noise, availability, obstacle heights, and distance, using a grid-based approach to segment and classify data, and applying noise factors to refine obstacle detection probabilities.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If free space is estimated from stereo camera data or radar data, then obstacle detection capability is improved, but sensor noise and measurement uncertainty increase
Solution Approach 1:
The patent segments the point cloud data into multiple regions based on distance from the sensor, creating near field, mid field, and far field zones. Each zone has different probability calculation parameters, allowing the system to account for varying sensor reliability at different distances while maintaining overall detection accuracy
Solution Approach 2:
The patent dynamically adjusts probability parameters based on sensor availability, measurement significance, and distance factors. By changing parameters like initial probability values and noise factors according to environmental conditions and sensor performance, the system maintains reliable obstacle detection despite varying sensor noise levels
2Reliability
If complex criteria are used to determine free space, then navigation safety is improved, but computational complexity increases
Solution Approach 1:
The patent divides the environment into a grid of cells and processes each cell independently with standardized probability calculations. This segmentation allows complex safety criteria to be applied systematically across the entire scene without requiring a single monolithic complex algorithm, improving both safety and computational efficiency
Solution Approach 2:
The patent calculates occupancy probabilities for all grid cells even though not all cells contain obstacles. By performing calculations for the entire grid and then filtering results, the system ensures no potential obstacle is missed while maintaining a uniform computational approach that simplifies the overall system design
3Speed
If real-time free space estimation is performed, then navigation speed is improved, but measurement accuracy decreases due to processing time constraints
Solution Approach 1:
The patent segments point cloud data into distance-based zones and processes each zone with simplified probability calculations tailored to that range. This allows real-time processing by breaking down the complex task into smaller, faster sub-tasks while maintaining accuracy through zone-specific parameters
Solution Approach 2:
The patent pre-calculates probability parameters, noise factors, and measurement significance values based on sensor characteristics and environmental conditions. By preparing these parameters in advance, the system can perform rapid real-time probability updates without compromising accuracy, as the computationally intensive parameter derivation is done beforehand
4Reliability
If sensor limitations are accounted for in probability calculations, then detection reliability is improved, but computational load increases
Solution Approach 1:
The patent applies different probability calculation parameters and noise factors to different spatial regions and distance zones based on local sensor performance characteristics. By tailoring calculations to local conditions rather than applying a uniform complex model everywhere, the system improves detection reliability in critical areas while reducing computational load in less critical regions
Data Source
AI summary
A system and method for estimating free space and assigning free space probabilities in point cloud data associated with an autonomous vehicle traveling on a surface, including taking into account sensor noise, sensor availability, obstacle heights, and distance of obstacles from the sensor. System and method can include determining surface planes and classifying point cloud points according to whether or not the points fall on surface planes, among other factors.