Automated Driving Perception Error Detection via Map Comparison
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current automated driving systems face challenges in accurately recognizing and interpreting their surroundings, leading to potential safety issues due to recurrent errors in scene comprehension, which can result in unsafe operation of motor vehicles.
Innovation Solution
A method that utilizes a surroundings sensor system to record and process raw data, generates semantic surroundings data through object recognition, and compares it with predefined semantically annotated map data to identify discrepancies, triggering safety measures and improving the reliability of automated driving by using high-resolution map data as a ground truth for error recognition.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Extent of automation
If automated driving systems use sensor systems to record and process surroundings data, then the system can automatically operate the vehicle, but errors in scene comprehension occur leading to safety issues
Solution Approach 1:
The patent implements a feedback mechanism where recognition results from the automated driving system are compared with map data to detect discrepancies. This feedback loop enables the system to identify errors in scene comprehension by contrasting sensor-based recognition with pre-stored map information, thereby improving reliability while maintaining automation.
Solution Approach 2:
The patent applies preliminary action by pre-storing semantically annotated map data before the automated driving operation. This pre-prepared reference data enables real-time error detection during operation, allowing the system to compare current sensor recognition results with expected values from the map, thus ensuring safety without reducing automation extent.
2Reliability
If the system compares recognition results with map data to detect errors, then safety is improved, but the device complexity increases
Solution Approach 1:
The patent applies universality by using the map data storage unit and comparison unit for multiple functions: they serve both as reference data storage and as error detection mechanisms. This multi-functionality reduces the need for separate dedicated error detection hardware, thereby improving safety while limiting the increase in device complexity.
Solution Approach 2:
The patent introduces map data as an intermediary element that mediates between the sensor system and the automated driving control. This intermediary provides a reference framework for error detection without requiring direct complex interactions between sensors and control systems, thus improving safety while managing system complexity.
3Loss of information
If semantic object recognition is performed on surroundings data, then scene comprehension is enhanced, but errors in object recognition and classification occur
Solution Approach 1:
The patent uses feedback by comparing the results of semantic object recognition with pre-stored map data. This comparison provides a verification mechanism that identifies discrepancies between recognized objects and expected objects from the map, thereby detecting recognition errors while maintaining enhanced scene comprehension.
Solution Approach 2:
The patent replaces direct reliance on sensor-based object recognition with a verification mechanism using map data comparison. This substitution reduces dependence on potentially error-prone sensor interpretation by using pre-validated map information as a reference, thereby improving recognition accuracy while preserving scene comprehension capabilities.
Data Source
AI summary
A method and an assistance device assist automated driving operation of a motor vehicle. Surroundings raw data recorded by way of a surroundings sensor system of the motor vehicle are processed by the assistance device in order to generate semantic surroundings data. This is accomplished by carrying out semantic object recognition. Further, a comparison of predefined semantically annotated map data against the semantic surroundings data is performed. This involves static objects indicated in the map data being identified in the semantic surroundings data as far as possible. Discrepancies detected during the process are used to recognize perception errors of the assistance device. A recognized perception error prompts a predefined safety measure to be carried out.