Intersection Driving Exit Prediction Using Likelihood Analysis
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
In automatic driving, accurately predicting the driving exit of another vehicle at an intersection is challenging due to complex vehicle flows and the absence of lane identifiers, leading to potential collisions and reduced traffic efficiency.
Innovation Solution
A method and apparatus for predicting the driving exit of a target vehicle at an intersection by obtaining its location and driving direction, calculating likelihoods for each driving exit based on reference points, and determining the exit with the highest likelihood using statistical methods and posterior probabilities.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Device complexity
If traditional intersection prediction methods are used without lane identifiers, then device complexity is reduced, but measurement precision of driving exit prediction deteriorates
Solution Approach 1:
The patent segments the intersection prediction problem into multiple independent probability calculations for each driving exit. Instead of treating the complex intersection as a single unit, the system divides it into N distinct driving exits, each with its own probability calculation based on track distance and angle deviations. This segmentation allows precise prediction for each exit while keeping the overall system manageable.
Solution Approach 2:
The patent introduces a probability dimension to the prediction system. Rather than simply determining which exit a vehicle will take, the system calculates likelihood probabilities for each exit by analyzing track distance and angle deviations. This probabilistic approach adds a new dimension of information that significantly improves prediction accuracy without requiring complex infrastructure.
2Reliability
If prediction accuracy is improved by calculating likelihoods for each exit, then reliability of collision avoidance is improved, but loss of time for calculation increases
Solution Approach 1:
The patent applies partial action by focusing calculations only on the most likely driving exits rather than exhaustively analyzing all possible paths. The system calculates probabilities for N exits but uses heuristics to prioritize exits with higher likelihoods, performing more detailed analysis only when necessary. This reduces calculation time while maintaining high reliability for the most critical predictions.
Solution Approach 2:
The system performs preliminary filtering of driving exits based on basic track analysis before conducting detailed probability calculations. By pre-identifying exits that are geometrically plausible based on the vehicle's current track, the system reduces the number of exits requiring full probability computation, thereby reducing calculation time while preserving accuracy for relevant exits.
3Reliability
If the system decelerates in advance based on predicted straight movement, then safety is improved, but productivity of traffic flow decreases
Solution Approach 1:
The patent implements dynamic prediction that adapts to the target vehicle's actual behavior. The system continuously updates the predicted driving exit based on real-time track analysis, allowing the host vehicle to dynamically adjust its speed and position. This dynamic approach enables the system to maintain safety by decelerating only when prediction confidence is high, while improving productivity by avoiding unnecessary deceleration when the target vehicle is likely to turn.
Data Source
AI summary
This application discloses a vehicle driving exit prediction method includes: obtaining a first location of the target vehicle at the target intersection and a driving direction of the target vehicle; obtaining N reference points respectively associated with N driving exits of the target intersection, where N is a positive integer; obtaining, based on the first location, the driving direction of the target vehicle, and the N reference points, likelihoods corresponding to the N driving exits respectively, where the likelihood indicates a probability that the target vehicle travels out of the target intersection from a corresponding driving exit; and obtaining a driving exit with a largest likelihood in the N driving exits as the driving exit of the target vehicle.