Vehicle Route Distribution Training for Maneuver-Aware Planning
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing technologies face challenges in efficiently generating vehicle driving routes that consider various maneuver modes while accounting for obstacles, requiring significant computational resources.
Innovation Solution
A method and apparatus that utilize a route search algorithm with maneuver mode-specific weights and a hybrid A* algorithm to generate trajectories, training a model to determine driving routes based on latent vectors and driving datasets, incorporating indicators for search time, reference velocity, lateral and longitudinal movements, and heading angles.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If a route search algorithm considers all possible maneuver modes and routes, then the completeness of route planning is improved, but the computational complexity increases significantly
Solution Approach 1:
The patent segments the route planning problem into multiple discrete maneuver modes (e.g., lane keeping, lane changing, stopping, swerving, following). By dividing the continuous decision space into distinct modes, the system can evaluate each mode separately using trajectory generation algorithms, reducing the overall computational burden while maintaining comprehensive route coverage.
Solution Approach 2:
The patent dynamically adjusts the evaluation process by generating trajectories only for relevant maneuver modes based on current driving situations and obstacles. Instead of statically evaluating all possible routes, the system adaptively selects and evaluates only the necessary maneuver modes, reducing computational complexity while maintaining route planning completeness.
2Measurement precision
If multiple indicators with different weights are applied to maneuver modes, then the precision of route determination is improved, but the complexity of parameter tuning increases
Solution Approach 1:
The patent changes the parameter representation by using a standardized set of indicators (search time, reference velocity, lateral movement, longitudinal movement, heading angle) with adjustable weights for each maneuver mode. This parameterization allows precise control over trajectory evaluation while providing a systematic framework for weight adjustment based on driving situations, reducing the arbitrary complexity of parameter tuning.
Data Source
AI summary
Provided is a training method for determining a driving route of a vehicle. The training method includes extracting a latent vector based on trajectories corresponding to maneuver modes that a vehicle is capable of selecting in a driving situation and training a model for generating route distributions to determine a driving route of the vehicle, based on a driving dataset and the latent vector, in which the trajectories are generated by applying, to a route search algorithm, weights of indicators for maneuvers differently according to the maneuver modes.