Behavioral Cloning Filter for Autonomous Vehicle Trajectory Planning
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Autonomous vehicles face safety challenges due to high speeds, complex environments, and false-positive perceptions from imperfect sensor data, leading to potentially catastrophic failures or overly conservative behavior.
Innovation Solution
A behavioral cloning approach that generates a bird's-eye-view grid with scalar confidence values to ignore vehicles with low confidence and select trajectories based on expert vehicle behavior policies, effectively addressing false-positives and improving safety.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If conventional modular behavioral cloning techniques are used for autonomous vehicles, then the system can process perception data, but false-positive observations in the input space lead to catastrophic failures or overly conservative behavior
Solution Approach 1:
The system performs preliminary filtering of perception observations by checking confidence values against a threshold before processing. Low-confidence observations are discarded in advance, preventing false-positives from propagating through the behavioral cloning pipeline and causing catastrophic failures.
Solution Approach 2:
The system applies different quality standards to different observations based on their confidence values. High-confidence observations are processed normally, while low-confidence observations are filtered out, creating a quality-based segmentation of the input data stream.
2Reliability
If the system ignores vehicles with low confidence values to reduce false-positives, then safety improves, but potentially valid vehicles may be missed
Solution Approach 1:
The system creates a filtered copy of the perception data that excludes only the low-confidence false-positives. This allows the behavioral cloning model to operate on clean data while preserving all high-confidence valid detections, effectively separating the filtering function from the original perception pipeline.
3Loss of information
If the system processes all perceived vehicles including false-positives, then no valid vehicles are missed, but catastrophic failures may occur
Solution Approach 1:
The system extracts and removes harmful false-positive observations from the input data stream by identifying them through confidence value thresholds. This extraction process isolates the harmful elements (low-confidence false-positives) from the useful data (high-confidence valid detections) before processing.
Solution Approach 2:
The system applies a preliminary counter-action to false-positives by filtering them out before they can cause harmful effects. The confidence threshold check serves as a preventive measure that stops false-positives from entering the behavioral cloning decision-making process.
Data Source
AI summary
A method for behavior cloned vehicle trajectory planning is described. The method includes perceiving vehicles proximate an ego vehicle in a driving environment, including a scalar confidence value of each perceived vehicle. The method also includes generating a bird's-eye-view (BEV) grid showing the ego vehicle and each perceived vehicle based on each of the scalar confidence values. The method further includes ignoring at least one of the perceived vehicles when the scalar confidence value of the at least one of the perceived vehicles is less than a predetermined value. The method also includes selecting an ego vehicle trajectory based on a cloned expert vehicle behavior policy according to remaining perceived vehicles.