Unified Driver Behavior Model for Path Prediction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing driver assistance systems face challenges in accurately predicting driver intentions, particularly in complex scenarios involving multiple maneuvers and overlapping behaviors, leading to increased development and validation efforts and reduced reliability.
Innovation Solution
A method that uses digital maps to determine future vehicle paths and incorporates consistent driver behavior models for predicting multiple driving maneuvers, accounting for different driving styles and lane probabilities, to improve prediction accuracy and reliability.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If a fixed set of features and separate classifiers are used for each maneuver type, then the system can provide specific predictions for each maneuver, but the development and validation effort increases significantly and reliability decreases when classifier determination is unclear
Solution Approach 1:
The patent applies universality by using a single unified classifier that can handle multiple maneuver types (lane changes, turns, intersections) instead of separate classifiers for each maneuver. The classifier receives a fixed set of features and determines the most likely maneuver type and timing, providing multi-functional capability that reduces development complexity while maintaining prediction accuracy across different driving scenarios
Solution Approach 2:
The patent changes parameters by using a fixed set of standardized features (lateral position, longitudinal position, acceleration, steering angle, etc.) that can be applied across all maneuver types. This parameter standardization allows the same classifier to accurately predict different maneuvers by analyzing the same feature set, reducing the need for multiple specialized classifiers and their associated development efforts
2Adaptability or versatility
If multiple classifiers are trained for different maneuver types and feature combinations, then comprehensive coverage of driving scenarios is achieved, but the system complexity increases and prediction reliability decreases
Solution Approach 1:
The unified classifier provides universal applicability across all driving scenarios including lane changes, turns, and intersections. By using the same classifier structure and feature set for all maneuver types, the system achieves comprehensive scenario coverage while maintaining consistent prediction reliability, avoiding the uncertainty that arises from determining which of multiple specialized classifiers should be applied
3Measurement precision
If the system waits for more driver information (mirror glances, shoulder checks) before predicting intention, then prediction accuracy improves, but the reaction time of the driver assistance system is delayed
Solution Approach 1:
The system performs preliminary prediction of driver intention using available features before the driver completes additional verification actions like mirror glances or shoulder checks. By analyzing the fixed set of features (lateral position, acceleration, steering angle, etc.) in real-time, the system can predict the driver's intended maneuver and provide timely assistance, rather than waiting for all possible driver actions to confirm intention
Data Source
Figure 1~2
Figure 3~4
Figure 5~6
AI summary
The invention relates to a method for predicting driving paths of a vehicle, comprising the following steps: the actual position of the vehicle is used; a selection of future possible driving paths of a vehicle using a digital map, the actual position, and predetermined maximum width of each driving path, in particular a maximum length of each driving path, a maximum driving path time for each driving path or a maximum number of travel manoeuvres for each driving path, are determined; measurements of the predetermined type of route-relevant manoeuvres of the driver of the vehicle, in particular the speed or acceleration of the vehicle, the setting of an indicator, and/or the direction to which a driver looks to the surroundings of the vehicle, are used; for each driving path of the selection of driving paths, from one or more models, which respectively use a number of times the indicator is activated respectively for one or more types of manoeuvres of the driver, said number of times of activation providing a numerical value for the probability that the measurements of the driving-relevant manoeuvres when driving the vehicle take place on the respective driving path; the respective number of times the indicator is activated is determined for each driving path and for each model; for each driving path from the selection of driving paths, a number of executions, that is a numerical value for the probability that the corresponding driving path is driven along, is determined based on the respectively defined number of activations of the indicator for the driving path.