AV Driving Path Label Generation for Construction Zone Training Data
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing technologies require significant amounts of expensive and prone-to-error human-labeled construction zone data for training autonomous vehicle (AV) navigation models, leading to poor performance due to human errors and dataset noise.
Innovation Solution
Generate a final AV driving path using a path adjustment operation that minimizes human input, employing a roadgraph to identify candidate paths with cost terms, and refine them into smooth, continuous curves, reducing human errors and noise.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Quantity of substance
If human-labeled construction zone data is used for training AV navigation models, then training data availability is improved, but error rate increases and data quality deteriorates due to human errors and dataset noise
Solution Approach 1:
The system generates its own training data through automated path adjustment operations and simulation environments, eliminating dependence on human-labeled data. The autonomous vehicle system itself produces the training datasets by simulating construction zone scenarios and generating ground truth driving paths through computational algorithms rather than human annotation.
Solution Approach 2:
The patent replaces the manual human labeling process with automated computational algorithms. The path adjustment operation uses cost functions and optimization algorithms to automatically generate accurate driving paths, substituting the mechanical human annotation process with an automated computational system that produces error-free training data.
2Manufacturing precision
If automated path adjustment operations are performed to generate final AV driving paths, then manufacturing precision of training data is improved, but device complexity increases due to multiple processing steps
Solution Approach 1:
The path adjustment operation is divided into distinct computational stages: initial path generation, cost function evaluation, path optimization, and final path refinement. Each stage processes specific aspects of path generation independently, allowing the complex task to be managed through modular computational steps that can be executed systematically.
Solution Approach 2:
The system performs preliminary path adjustment operations during the training data generation phase, pre-computing optimal driving paths for various construction zone scenarios. These preliminary computations establish ground truth data that is then used to train the navigation model, avoiding the need for complex real-time path planning during actual autonomous operation.
Data Source
AI summary
A system identifies a set of input data including a roadgraph identifying an intermediate autonomous vehicle (AV) driving path related to a scene representing an environment proximate an AV. The intermediate AV driving path reflects a modification to an initial (AV) driving path to avoid one or more obstacles obstructing the initial AV driving path. The system performs, using the set of input data, a path adjustment operation that identifies one or more candidate AV driving paths based on the intermediate AV driving path and determines a cost value for each of the candidate AV driving paths. The system identifies, among the candidate AV driving paths, a final AV driving path having a cost value that satisfies an evaluation criterion. The final AV driving path is to be included in the set of training data as a target output paired with training input including scene data identifying the scene.