Sensor Fusion Output Validation for ADS Perception Reliability
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
There is a need for methods and systems that ensure the reliability of Automated Driving Systems (ADS) perception systems, as the quality of sensor fusion output directly affects the performance and safety of ADS functions.
Innovation Solution
A computer-implemented method and system for monitoring the reliability of a sensor fusion system in vehicles by comparing perception output data with stored input data from multiple vehicle-mounted sensors to validate object detections and free-space area indications, and outputting a signal indicative of the sensor fusion system's status based on validated measures.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If sensor fusion systems are used to process multiple sensor outputs, then the completeness of perception output is improved, but the reliability of the output cannot be ensured
Solution Approach 1:
The patent implements a feedback mechanism where the monitoring system continuously compares sensor fusion output with raw sensor data and feeds back reliability status to the ADS. This closed-loop feedback enables real-time validation of perception outputs without requiring fundamental changes to the sensor fusion architecture, thereby improving reliability while maintaining system complexity at acceptable levels.
Solution Approach 2:
The patent introduces an intermediary monitoring system that acts as a mediator between the sensor fusion system and the ADS. This intermediary layer validates sensor fusion outputs by comparing them with raw sensor data, providing an additional verification step that enhances reliability without directly modifying the core sensor fusion processing.
2Measurement precision
If sensor fusion output quality is improved, then ADS function performance is improved, but system reliability monitoring is lacking
Solution Approach 1:
The monitoring system establishes a feedback loop that continuously assesses the quality and reliability of sensor fusion outputs by comparing them against raw sensor data. This feedback mechanism provides real-time reliability information to the ADS, enabling the system to respond to quality variations and maintain precise operation.
Solution Approach 2:
The system performs self-validation by automatically comparing sensor fusion outputs with raw sensor data without requiring external verification systems. This self-service approach enables the system to autonomously monitor its own reliability and precision, ensuring high-quality perception outputs while maintaining system simplicity.
3Reliability
If multiple sensors are used to monitor surrounding environment, then perception completeness is improved, but output reliability validation is insufficient
Solution Approach 1:
The patent introduces an intermediary monitoring layer that validates sensor fusion outputs by comparing them with raw sensor data from multiple sensors. This intermediary system provides a unified validation approach that leverages data from all sensors without requiring complex inter-sensor coordination, thereby improving output validation while keeping the monitoring system architecture relatively simple.
Solution Approach 2:
The monitoring system creates a copy of the raw sensor data and uses it for validation purposes alongside the sensor fusion processing. This copying approach enables independent verification of sensor fusion outputs without interfering with the primary sensing and processing functions, maintaining system complexity at acceptable levels while improving validation capability.
Data Source
Figure 1
Figure 2
Figure 3
AI summary
The present disclosure relates to methods and systems for monitoring a reliability of an output of a sensor fusion system of a vehicle, wherein the sensor fusion system is configured to receive input data and generate perception output data. To achieve this, input data of the sensor fusion system is stored, wherein the input data comprises sensor data, obtained over a time period, from each of a plurality of vehicle-mounted sensors configured to monitor a surrounding environment of the vehicle. Perception output data that is output from the sensor fusion system is obtained using the input data obtained over the time period, the obtained perception output data comprising one or more object detections in the surrounding environment of the vehicle and one or more free-space area indications in the surrounding environment of the vehicle. Furthermore, for each sensor or subset of sensors of the plurality of vehicle-mounted sensors: the obtained perception output data is compared with the stored input data and determined whether any object detections indicated in the obtained perception output data is/are indicated in the field-of-view of that sensor or subset of sensors in order to validate any object detections indicated in the obtained perception output data, and whether any free-space area indications in the obtained perception output data is/are indicated in the field-of-view of that sensor or subset of sensors in order to validate any free-space area indicated in the obtained perception output data. Moreover, a signal indicative of a status of the sensor fusion system is output based on a measure of validated objects and a measure of validated free-space areas.