Vehicle Sensor Fusion Diagnostics for Safe Control Handover
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing self-driving control systems face challenges in accurately diagnosing sensor fusion output data due to differences in data processing methods, leading to inconsistencies in detected objects, which complicates data validation and reliability assessment.
Innovation Solution
A vehicle control device configuration that includes multiple arithmetic blocks for sensor fusion processing, where one block performs raw data fusion, another block performs object data fusion, and a third block diagnoses the output results from the first two, ensuring accurate validation and reliability through comparison and majority decision-based diagnostics.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If raw data fusion and object data fusion are performed using different data processing methods, then sensor fusion processing capability is improved, but diagnostic accuracy deteriorates due to inconsistencies in detected objects
Solution Approach 1:
The system segments the sensor fusion process into two distinct processing paths: raw data fusion and object data fusion. Each path handles data differently - raw data fusion processes sensor outputs directly, while object data fusion processes after object detection. This segmentation allows each path to be optimized for its specific purpose while maintaining diagnostic capability through separate validation of each path's results.
Solution Approach 2:
The system applies different quality standards and validation criteria to each fusion path based on its specific characteristics. Raw data fusion is validated for data consistency and processing accuracy, while object data fusion is validated for object detection accuracy and trajectory consistency. This local quality approach allows diagnostic accuracy to be maintained for each path despite using different processing methods.
2Reliability
If multiple arithmetic blocks perform sensor fusion processing with different methods, then processing reliability is improved, but system complexity increases
Solution Approach 1:
The system divides the sensor fusion functionality into multiple arithmetic blocks, each responsible for a specific fusion method (raw data fusion and object data fusion). This segmentation enables parallel processing and cross-validation, improving reliability through redundancy while organizing complexity into manageable, specialized modules with clear division of responsibilities.
Solution Approach 2:
The system implements a diagnostic block that continuously monitors and compares outputs from multiple arithmetic blocks. This feedback mechanism validates processing results in real-time, identifying inconsistencies and ensuring reliability. The feedback loop manages system complexity by providing automated validation and error detection, reducing the need for complex manual monitoring systems.
3Ease of operation
If object-based fused data is compared with raw object data using error tolerance ranges, then data validation is simplified, but diagnostic capability deteriorates when detectable objects differ due to processing method differences
Solution Approach 1:
The system segments validation into path-specific checks: raw data fusion results are validated for data consistency and processing accuracy, while object data fusion results are validated for object detection accuracy. This segmentation maintains diagnostic capability by applying appropriate validation criteria to each path rather than using a single generic comparison method.
Solution Approach 2:
The system changes validation parameters based on the fusion path being evaluated. Different error tolerance ranges, comparison metrics, and validation thresholds are applied depending on whether validating raw data fusion or object data fusion results. This adaptive parameter approach maintains both validation simplicity and diagnostic accuracy by matching validation criteria to the specific characteristics of each fusion method.
Data Source
AI summary
There are realized a vehicle control device and an electronic control system with high reliability capable of safely shifting control even when an operation abnormality occurs in a sensor around a vehicle or an arithmetic block that processes sensor fusion. The vehicle control device includes a first arithmetic block which performs sensor fusion processing based on pieces of raw data output from a plurality of surrounding environment sensors, a second arithmetic block which performs sensor fusion processing based on pieces of object data generated by processing the pieces of raw data output from the plurality of surrounding environment sensors, and a third arithmetic block which diagnoses an output result of the first arithmetic block by using the output result of the first arithmetic block and an output result of the second arithmetic block.