Vehicle Perception Sensor Pose Verification Before Autonomous Operation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
The accuracy of perception systems in autonomous vehicles can be compromised due to misalignment or damage of perception sensors during maintenance, storage, or transport, necessitating a method to verify their accuracy before activation.
Innovation Solution
A method involving the vehicle traversing a path around a fixed target with a known pose, allowing each perception sensor to capture images of the target and comparing the detected pose to the known pose to ensure alignment, with additional verification through image sequences and cross-checks between sensors.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Manufacturing precision
If perception sensors are initially calibrated during manufacturing or installation, then manufacturing precision is improved, but reliability deteriorates due to potential misalignment or damage during maintenance, storage, and transport
Solution Approach 1:
The patent performs calibration verification before the perception system is activated for navigation. This preliminary check ensures that any misalignment or damage occurring during maintenance, storage, or transport is detected before the system is put into service, preventing unreliable operation
Solution Approach 2:
The system compares the detected pose of calibration targets with their known poses to provide feedback on sensor accuracy. This feedback mechanism allows the system to identify calibration drift or damage and trigger appropriate corrective actions
2Measurement precision
If multiple perception sensors are used to improve detection accuracy, then measurement precision is improved, but device complexity increases
Solution Approach 1:
The patent uses a single mobile calibration target that serves multiple functions: it can be detected by different sensor types (cameras, LiDAR, radar), it can verify calibration for multiple sensors, and it can be positioned at various locations to test different sensor fields of view. This multi-functional approach reduces the need for multiple specialized calibration devices
Solution Approach 2:
The system uses images captured by perception sensors as digital copies of the physical calibration targets. These image copies are then processed to determine target pose and compare with known poses, eliminating the need for direct physical measurement and simplifying the verification process
3Reliability
If the vehicle traverses a path to bring targets into sensor FOV for verification, then reliability is improved through comprehensive checking, but loss of time increases due to the verification process
Solution Approach 1:
The patent performs calibration verification on a subset of calibration targets and sensors rather than requiring complete verification of all sensors in all conditions. This partial verification approach provides sufficient confidence in system reliability while significantly reducing the time required compared to exhaustive verification
Solution Approach 2:
The perception system performs its own calibration verification using its existing sensors and processing capabilities. The system autonomously captures images, determines target poses, compares with known poses, and identifies calibration issues without requiring external verification equipment or personnel, thereby reducing verification time
Data Source
Figure 1
Figure 2A~2B
Figure 3
AI summary
A method of verifying accuracy of a perception system of a vehicle includes causing the vehicle to traverse a path around a target that is fixed in an environment. The target has a known pose. The path is configured so that the target comes into a respective field of view (FOV) of each respective perception sensor of one or more perception sensors of the perception system along the path. The method further includes, for each respective perception sensor of the one or more perception sensors, while the target is within the respective FOV of the respective perception sensor, acquiring a respective image of the target using the respective perception sensor; at the perception system, determining a respective pose of the target based on the respective image; and at a computer system communicatively coupled with the perception system, determining whether the respective pose matches the known pose of the target.