Road User Influence Prediction for Autonomous Trajectory Planning
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods for predicting vehicle trajectories and operating autonomous vehicles fail to accurately quantify the influence of one road user on another, necessitating improved interaction detection and prediction for safe and collision-free navigation.
Innovation Solution
A method using a trained artificial neural network to calculate a scalar influence metric based on trajectory deviations, employing map-free and map-based approaches to learn and predict the influence of one road user on others, enabling reliable vehicle operation through collision warnings and trajectory planning.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If traditional trajectory prediction methods are used, then computational complexity is reduced, but prediction accuracy of road user interactions is insufficient
Solution Approach 1:
The system segments the set of road users into relevant and irrelevant groups based on interaction metrics. The attention mechanism divides computational resources by focusing only on road users that have significant interaction with the target vehicle, rather than processing all detected road users equally. This segmentation enables the complex neural network to operate efficiently by reducing the effective input space.
Solution Approach 2:
The patent introduces an attention-based interaction algorithm as an intermediary layer between raw sensor data and trajectory prediction. This intermediary computes interaction metrics and selects relevant road users, serving as a bridge that filters and prepares data for the subsequent trajectory prediction network. The intermediary enables the system to handle complexity by introducing a intermediate processing stage that reduces information dimensionality.
2Productivity
If all detected road users are considered in trajectory prediction, then comprehensive coverage is achieved, but computational efficiency decreases
Solution Approach 1:
The system applies partial action by considering only a subset of road users that are deemed relevant based on interaction metrics. Rather than processing all detected road users, the attention mechanism selectively focuses on those with significant influence on the target vehicle's trajectory. This partial processing maintains prediction reliability for critical interactions while improving computational efficiency by excluding irrelevant road users.
Solution Approach 2:
The patent changes the parameter of road user selection from a static inclusion of all detected users to a dynamic selection based on computed interaction parameters. The attention mechanism calculates interaction metrics that change based on relative positions, velocities, and other contextual factors, allowing the system to adaptively adjust which road users are considered relevant. This parameter-based selection enables efficient processing while maintaining reliability.
3Reliability
If interaction influences between road users are not quantified, then system simplicity is maintained, but collision avoidance accuracy deteriorates
Solution Approach 1:
The patent replaces traditional mechanical or rule-based interaction models with a data-driven neural network approach. Instead of using predefined rules for determining road user interactions, the system employs a trained neural network that learns interaction patterns from data. The attention mechanism substitutes traditional physics-based interaction calculations with learned attention weights, enabling more accurate collision avoidance while managing system complexity through data-driven approaches.
Data Source
AI summary
An influence of one road user on at least one other road user is predicted by evaluating traffic scenarios by a trained artificial neural network. The neural network is trained by recorded traffic scenarios, the traffic scenarios include several road users and are labelled with score values that represent an influence of one road user by other road users. A respective score value for one road user with respect to another road user is calculated based on a deviation between two trajectories of the one road user. One of the two trajectories is a detected real trajectory that the one road user actually takes in a respective recorded traffic scenario, and the other of the two trajectories is a simulated trajectory determined in a simulation and representing a trajectory that the one road user would take in the same traffic scenario if the other road user were not present.