Cross-traffic detection using linear interpolation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing cross-traffic alert systems face challenges in accurately predicting when an approaching vehicle will pass behind a host vehicle, especially when initial heading estimates are erroneous or the road is curved, leading to late alerts and excessive computational resource usage.
Innovation Solution
A cross-traffic detection system that uses linear interpolation to determine a vehicle's path-history and estimates a crossing-distance by measuring the offset-distance relative to a linearly-interpolated line, allowing for efficient determination of whether a second vehicle is following the path-history and predicting potential collisions.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If polynomial or curved least-squares model is used to fit trajectories of approaching vehicles, then prediction accuracy is improved, but computational resource usage and memory requirements increase excessively
Solution Approach 1:
The patent segments the trajectory prediction problem by dividing the road into multiple sections and using separate linear models for each section. Instead of applying a single complex polynomial model to the entire trajectory, the system breaks down the path into manageable segments, each handled by simpler linear interpolation calculations. This segmentation maintains prediction accuracy while significantly reducing computational complexity and memory requirements.
Solution Approach 2:
The patent uses simple linear interpolation models that are computationally inexpensive and can be quickly calculated and discarded. Rather than maintaining complex polynomial models that require substantial computational resources and memory storage, the system employs lightweight linear models that can be rapidly computed for each trajectory segment and then discarded, reducing both computational overhead and memory usage.
2Measurement precision
If polynomial or curved least-squares model is used to fit trajectories, then prediction accuracy is improved, but memory requirements to store data and solutions increase
Solution Approach 1:
The patent segments the trajectory data storage by dividing the path into multiple smaller sections, each requiring minimal memory to store linear model parameters. Instead of storing large datasets required for polynomial fitting, the system stores compact linear model parameters for each segment, dramatically reducing memory requirements while maintaining the ability to accurately predict vehicle trajectories.
Solution Approach 2:
The patent employs lightweight linear interpolation models that require minimal memory storage. These simple models use only a few parameters per trajectory segment compared to the extensive data requirements of polynomial models. The computational simplicity allows the system to maintain accurate predictions with significantly reduced memory footprint.
3Ease of operation
If traditional heading estimation methods are used, then initial vehicle orientation is determined, but accuracy deteriorates when road is curved or initial estimates are erroneous
Solution Approach 1:
The patent performs preliminary actions by establishing a reference path using linear interpolation between known road geometry points before attempting to estimate vehicle heading. This preliminary path establishment provides an accurate geometric framework that guides subsequent heading calculations, ensuring accuracy even on curved roads or when initial estimates are erroneous. The system prepares the reference trajectory in advance to guide the heading estimation process.
Solution Approach 2:
The patent introduces a linearly-interpolated reference path as an intermediary between the raw vehicle position data and the final heading estimation. This intermediary reference path, derived from known road geometry, mediates the heading calculation process by providing an accurate geometric baseline. The vehicle's actual heading is then determined relative to this intermediary reference, improving accuracy on curved roads and correcting erroneous initial estimates.
Data Source
Figure 1
Figure 2
Figure 3
AI summary
A cross-traffic detection system (10) suitable for use on an automated vehicle includes an object-detector (14) and a controller (24). The object-detector (14) is used to determine locations (16) of a moving-object (18) relative to a host-vehicle (12). Each of the locations (16) is indicated by a lateral-distance (20) and a longitudinal-distance (22) of the moving-object (18) from the host-vehicle (12). The controller (24) is in communication with the object-detector (14). The controller (24) is configured to accumulate a plurality of first-longitudinal-distances (26) of a first-vehicle (30) at a plurality of predetermined-lateral-distances (28), and determine a path-history (40) of the first-vehicle (30) based on linear-interpolation between successive instances of the plurality of first-longitudinal-distances (26) at corresponding instances of the plurality of predetermined-lateral-distances (28). The controller (24) is also configured to determine that a second-vehicle (44) is following the path-history (40) of the first-vehicle (30) when the second-vehicle (44) passes within an offset-distance (46) to the path-history (40) that is less than an offset-threshold (48), and estimate a crossing-distance (54) from the host-vehicle (12) based on the path-history (40) and the offset-distance (46).