Autonomous Driving Map Reliability Evaluation From Continuous Road Recognition
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing map evaluation systems struggle to efficiently assess the accuracy of map data when a vehicle deviates from a guidance route, making it difficult to ensure safe autonomous driving.
Innovation Solution
A map evaluation apparatus that includes a microprocessor and memory for recognizing external situations, generating maps based on recognition results, calculating evaluation reference values, and determining the reliability of maps using confidence intervals to ensure accurate map evaluation.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If map evaluation is performed only when the subject vehicle departs from a guidance route, then the evaluation process is simple, but the evaluation efficiency and timeliness are insufficient
Solution Approach 1:
The patent performs map evaluation continuously during normal driving before any route deviation occurs. The evaluation unit calculates reliability values based on recognition information accumulated during regular travel, so that when deviation happens, the evaluation is already complete or near-complete. This transforms the evaluation from a reactive post-event analysis to a proactive continuous process, dramatically improving evaluation efficiency and timeliness without requiring complex additional hardware.
2Measurement precision
If continuous map evaluation is performed during normal driving, then the evaluation timeliness is improved, but the computational load and processing complexity increase
Solution Approach 1:
The patent extracts only the essential elements needed for evaluation: recognition information (road width, curvature, gradient) is stored in memory, and the evaluation unit selectively processes only this extracted data to calculate reliability. By separating the essential evaluation data from the complete set of recognition data and focusing processing only on the extracted relevant portions, the system achieves precise continuous evaluation without overwhelming computational complexity.
Solution Approach 2:
The evaluation is performed locally for specific road sections rather than globally for entire maps. The evaluation unit calculates reliability values for individual road segments based on local recognition information, and these local evaluations are accumulated to form the overall map reliability. This localized approach enables continuous precise evaluation while keeping computational complexity manageable by processing only relevant local data.
3Reliability
If map reliability is determined based on deviation detection only, then the evaluation criteria are simple, but the reliability determination is insufficient and delayed
Solution Approach 1:
The patent implements continuous map evaluation during normal driving operations. The evaluation unit continuously calculates reliability values based on recognition information accumulated throughout the driving process, maintaining an ongoing assessment rather than waiting for deviation events. This continuous useful action ensures that when route deviation occurs, the reliability determination is already available or can be completed immediately, eliminating time delays while providing accurate and timely reliability information.
Data Source
AI summary
An apparatus for evaluating the reliability of a map used in autonomous driving. The apparatus recognizes an external situation around a subject vehicle, stores recognition information including recognition results in the memory, and generates a map based on the recognition results. The apparatus calculates an evaluation reference value to be used for evaluation of the map based on the recognition information stored in the memory, and determines a reliability of the map based on a deviation between the recognition information used to calculate the evaluation reference value and the evaluation reference value.