Ground Profile Estimation Using Spline Fitting for Object Separation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Distinguishing objects from the ground in environments with changes in grade poses challenges for autonomous vehicles, affecting safety and accuracy in collision avoidance systems.
Innovation Solution
A collision avoidance system with a secondary perception component classifies sensor data as either ground or object using spline fitting and threshold distance calculations, improving accuracy and reducing computational resources.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If traditional ground modeling methods are used in environments with grade changes, then the system is simple to implement, but the accuracy of distinguishing objects from ground deteriorates
Solution Approach 1:
The ground profile estimation is segmented into multiple discrete depth values at specific longitudinal positions rather than using a continuous complex model. This segmentation allows accurate representation of grade changes while maintaining computational simplicity by processing only key depth measurements at critical locations.
Solution Approach 2:
The system applies different depth value analysis locally at each longitudinal position along the path, rather than using a single global ground model. This local quality approach enables accurate adaptation to varying grade conditions at different locations while keeping each local computation simple and efficient.
2Measurement precision
If comprehensive sensor data processing is performed to accurately differentiate objects from ground, then the accuracy improves, but the computational bandwidth required increases
Solution Approach 1:
The system extracts only the essential depth values at critical longitudinal positions from the complete sensor data set, rather than processing all sensor measurements. This extraction approach maintains high accuracy in object-ground differentiation while significantly reducing computational bandwidth by focusing only on the most relevant data points.
Solution Approach 2:
The system performs partial processing by analyzing only specific depth values at selected longitudinal positions rather than processing the entire sensor data set. This partial action approach provides sufficient accuracy for safety-critical object detection while consuming less computational bandwidth than complete data processing.
Data Source
Figure 1
Figure 2
Figure 3
AI summary
Classifying sensor data as being associated with ground (as opposed to an object) may comprise determining a number of channels of sensor data that have returns in them, setting a number of control points and a number of knots of a curve based at least in part on the number of channels that have returns, and fitting a curve having the number of control points and the number of knots to the sensor data. The curve may be used to distinguish sensor data associated with the ground from sensor data associated with an object. Determining the curve may additionally or alternatively include limiting an elevation value of a control point and/or knot based on elevation value(s) of the sensor data, weighting the sensor data based at least in part on elevation values associated with the sensor data, and/or adjusting knot spacing, et alia.