Overlapping Sensor Fields of View for Redundant Perception
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Autonomous vehicles face challenges in achieving true redundancy for perception systems due to different fields of view from multiple sensors, leading to potential single points of failure and reduced reliability.
Innovation Solution
A perception system utilizing two image sensors with overlapping fields of view, where each sensor captures calibration data to identify common features, and ECUs reduce the image data to overlapping regions for redundant object detection, ensuring identical outputs from separate ECUs to meet automotive safety integrity levels.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Area of stationary object
If multiple sensors with different fields of view are used, then the perception system can cover more environmental areas, but true redundancy cannot be achieved and reliability is reduced
Solution Approach 1:
The patent divides the environmental monitoring task into separate segments handled by different sensors (first sensor for first area, second sensor for second area), then uses ECUs to process each segment independently while maintaining redundancy through overlapping regions. This segmentation allows each sensor to specialize in specific areas while still providing comprehensive coverage and redundancy where needed.
Solution Approach 2:
The patent introduces a spatial dimension solution by positioning sensors with overlapping fields of view and using ECUs to align and process images from different spatial perspectives. The overlapping regions create a third dimension of redundancy that resolves the contradiction between coverage area and reliability.
2Reliability
If sensors with overlapping fields of view are used, then redundancy can be achieved, but the complexity of coordinating and processing data from multiple sensors increases
Solution Approach 1:
The patent segments the data processing task by assigning specific sensors to specific areas and designating particular ECUs to process specific sensor data. This segmentation reduces coordination complexity by localizing processing responsibilities while maintaining overall system redundancy.
Solution Approach 2:
The patent introduces ECUs as intermediary components that mediate between sensors and the central processing system. Each ECU handles specific sensor data independently, reducing the complexity of direct sensor coordination while enabling reliable redundant processing through standardized intermediary interfaces.
3Reliability
If image data is reduced to overlapping regions only, then redundant processing can be performed, but the amount of useful information processed is reduced
Solution Approach 1:
The patent segments image processing into two parallel paths: one processing overlapping regions for redundant verification and another processing non-overlapping regions for exclusive information gathering. This segmentation ensures that redundancy is achieved without discarding valuable unique information from each sensor's specialized viewing angle.
Solution Approach 2:
The patent applies partial action by processing only the overlapping regions for redundant verification purposes, while accepting that non-overlapping regions provide complementary information that is processed separately. This partial redundant processing achieves reliability where needed without unnecessarily processing all data through all paths.
Data Source
AI summary
A perception system includes a first electronic control unit (ECU) coupled to a first image sensor and a second ECU coupled to a second image sensor. The first ECU and the second ECU are configured to perform feature detection for calibration using first calibration image data captured by the first image sensor of a first field of view and second calibration image data captured by the second image sensor of a second field of view. The first and second ECUs are further configured to (i) identify a set of pixels in a respective field of view having common features with another set of pixels in another field of view; (ii) receive image data from respective image sensor; (iii) reduce the image data to only a set of pixels in the respective field of view; and (iv) perform object detection on the respective image data consisting of the set of pixels.