Path Planning Error Detection Using Human-Driven Reference Paths
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Identifying and capturing errors in predictive algorithms, such as neural networks, used in autonomous transportation vehicles is challenging, especially when these algorithms deviate from human-driven paths, leading to potential safety issues and limited data storage efficiency.
Innovation Solution
Implementing a system that tests predictive algorithms by comparing predicted paths with human-driven paths, identifying errors, and recording conditions surrounding these errors in semi-permanent memory, using a combination of sensor data and machine learning algorithms to detect deviations, temporal instability, and ensemble uncertainty.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If predictive algorithms are tested by comparing predicted paths with human-driven paths, then error detection capability is improved, but system complexity increases
Solution Approach 1:
The testing system is divided into separate functional modules: a path prediction module that generates predicted paths, a human-driven path recording module, a comparison module that identifies deviations, and an error buffering module. This segmentation allows each component to perform its specific function independently, improving error detection capability while managing system complexity through modular architecture.
Solution Approach 2:
A buffer memory acts as an intermediary component that stores error information and surrounding conditions temporarily. This buffer serves as a mediator between the error detection process and the semi-permanent storage, allowing the system to collect and organize error data before committing it to long-term storage, thereby simplifying the overall error handling process.
2Reliability
If error information and surrounding conditions are stored in semi-permanent memory, then data reliability is improved, but data storage requirements increase
Solution Approach 1:
The system extracts and stores only the essential error information and relevant surrounding conditions in semi-permanent memory, rather than storing complete raw data sets. This selective extraction of critical error data maintains reliability for analysis purposes while significantly reducing the overall data storage requirements compared to storing all operational data.
Solution Approach 2:
Different types of data are stored with different retention policies and storage qualities. Error-critical information is preserved in semi-permanent memory with high reliability, while less critical operational data may be stored temporarily or discarded. This local quality differentiation ensures reliable error data storage while optimizing overall storage resource utilization.
3Measurement precision
If comprehensive error data is collected and buffered, then analysis accuracy is improved, but processing time increases
Solution Approach 1:
Error information and surrounding conditions are buffered in temporary memory immediately when errors are detected, preparing the data for future analysis. This preliminary action of capturing and organizing error data in real-time allows comprehensive data collection without delaying the vehicle's operational processing, as the buffering occurs in parallel with normal operations.
Solution Approach 2:
The system preemptively buffers error data and surrounding conditions as they occur, preventing data loss or degradation that might happen if collection occurred later during analysis. This preliminary anti-action ensures that all necessary error information is captured complete and accurate before any potential data degradation, maintaining analysis accuracy while enabling efficient batch processing later.
Data Source
AI summary
In a predictive path planning functionality for a transportation vehicle data describing a roadway are acquired by one or more sensors disposed on the transportation vehicle and analyzed. One or more predicted paths are predicted with at least one neural network. Data describing a reference path taken by the transportation vehicle are acquired. The one or more predicted paths are evaluated to detect one or more errors in the predicting operation wherein the one or more errors include an indication that the one or more predicted paths of the path planning system deviate from the reference path.