Collision Avoidance System Using Map Data Correction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing collision avoidance systems face inaccuracies in predicting probability densities at longer time scales due to neglecting driver behavior and road construction effects, leading to false alarms or missed warnings, as they rely solely on mechanical models that fail to account for these factors.
Innovation Solution
A collision avoidance system that predicts preliminary probability densities based on vehicle dynamics and corrects them using a factor derived from map data, specifically adjusting the distribution across different regions to account for driver corrections and road features, thereby improving the accuracy of collision warnings and interventions.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of manufacture
If a mechanical model of vehicle motion is used to predict probability density, then the model is simple to implement, but the accuracy decreases at longer time scales due to neglecting driver behavior and road construction effects
Solution Approach 1:
The patent combines the simple mechanical vehicle motion model with additional data sources including map data (road construction information) and driver behavior patterns. This merging allows the system to maintain the computational simplicity of mechanical models while incorporating corrective factors from external data sources to improve prediction accuracy at longer time scales
Solution Approach 2:
The patent introduces map data and driver behavior models as intermediary elements that mediate between the simple mechanical model and the actual vehicle position prediction. These intermediaries provide correction factors that account for road construction effects and driver behavior patterns without requiring a complete replacement of the mechanical model
2Reliability
If probability density is overestimated for certain locations, then collision warnings are generated, but false alarms occur which compromise reliance on justified alarms
Solution Approach 1:
The patent employs feedback mechanisms where prediction results are continuously refined using map data and driver behavior patterns. The system uses feedback from multiple data sources to adjust probability density estimates, reducing overestimation that leads to false alarms while maintaining sensitivity to actual collision risks
Solution Approach 2:
The patent dynamically adjusts prediction parameters by incorporating map data and driver behavior models. These parameter changes allow the system to adapt probability density estimates to specific road conditions and driving patterns, reducing false alarms caused by rigid mechanical model predictions
3Object-generated harmful factors
If probability density is underestimated for certain locations, then false negative occurs, but collision warnings are missed
Solution Approach 1:
The patent merges multiple prediction approaches and data sources to compensate for underestimation issues. By combining mechanical model predictions with map data and driver behavior patterns, the system achieves more comprehensive coverage of potential collision scenarios, reducing false negatives while maintaining overall detection reliability
4Duration of action of moving object
If the time scale for collision avoidance is extended, then more time for intervention is available, but prediction accuracy decreases due to driver behavior unpredictability
Solution Approach 1:
The patent performs preliminary actions by incorporating map data and driver behavior patterns into the prediction model before actual collision risk assessment. This preliminary integration of corrective factors enables the system to extend the prediction time scale while maintaining accuracy by pre-accounting for driver behavior and road construction effects
Data Source
Figure 1~2
Figure 3
AI summary
Collision avoidance actions in a road vehicle are controlled based on a computed probability density of future positions of the vehicle. A preliminary probability density (32) is computed by means of a mechanical model by extrapolation from the detected state of the vehicle, including at least its detected position(31). Map data is used to define different regions in an area that contains the road, such as a first region that comprises a road part for traffic that moves in the direction of travel of the vehicle, a second region that comprises a road part for traffic in the opposite direction and a third region bordering on the road. The regions are used to obtain a correction factor of the computed probability density function in the first region (33). The correction factor is computed dependent on the aggregates of the probability density function in respective ones of the regions. In this way it is possible to compensate for errors that result from extrapolations that misinterpret swerving motion that suggest that the vehicle will move of the road, but are automatically corrected by the driver.