Vehicle Trajectory Planner Active Learning for Edge Case Validation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Trajectory planner systems in autonomous vehicles struggle to perform optimally in edge case driving scenarios, leading to reduced safety, comfortability, and traffic law adherence compared to human-driven vehicles, due to insufficient data and lack of granular understanding of complex driving situations.
Innovation Solution
An active-learning computing process that identifies and flags edge case driving scenarios by comparing human-driven and compute-based trajectories, using machine-learning models and hard-coded rules to generate learned parameters, validate, and retrain the trajectory planner system, iteratively improving its performance over time.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If typical trajectory planner systems are used with large aggregates of driving data, then they can generate accurate planned trajectories for rudimentary driving scenarios, but they struggle to perform optimally in edge case driving scenarios compared to human drivers
Solution Approach 1:
The system performs preliminary actions by proactively identifying and flagging edge case scenarios before they become critical failures. The active learning system continuously monitors trajectory discrepancies between human drivers and the trajectory planner, pre-identifying scenarios where the planner underperforms so these edge cases can be prioritized for improvement through targeted data collection and model retraining.
Solution Approach 2:
The system implements a closed-loop feedback mechanism where trajectory planner outputs are continuously compared against human driver behavior in similar scenarios. When discrepancies are detected in edge case situations, this feedback triggers active learning processes to collect additional data and retrain the model, thereby improving future performance in these challenging scenarios while maintaining reliability in routine situations.
2Productivity
If the trajectory planner system is trained on large aggregates of driving data, then it performs well on common driving scenarios, but it lacks granular understanding of complex driving situations
Solution Approach 1:
The system segments the driving data and scenarios into distinct categories: common driving scenarios and edge case scenarios. By dividing the overall dataset and applying different analysis depths to different segments, the system maintains efficient processing for routine situations while dedicating specialized computational resources and active learning attention to granular analysis of complex edge case situations.
Solution Approach 2:
The system applies local quality by enhancing the analysis depth and data granularity specifically for edge case scenarios where it is most needed, while maintaining standard processing for common scenarios. The active learning mechanism dynamically adjusts the level of scrutiny and data collection intensity based on the specific scenario type, ensuring high measurement precision for complex situations without sacrificing overall processing productivity.
Data Source
AI summary
A method includes generating a parameter of a trajectory associated with a scenario using a path planner. The parameter is generated based on a training dataset. The method includes comparing the parameter of the trajectory against a validation parameter associated with a validation dataset. The validation parameter is based on human-based vehicle driving trajectory data associated with scenarios that satisfy a level of similarity with the scenario. The method further includes determining a level of similarity between the parameter associated with the scenario and the validation parameter associated with the scenarios, and, subsequent to determining that the level of similarity fails to satisfy a similarity threshold, the method concludes with providing training data associated with the scenario to the training dataset so that a subsequent parameter of a subsequent trajectory generated by the path planner and associated with the scenario satisfies the level of similarity against the validation parameter.