Online Radar Sensor Calibration Using Environmental Targets
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing radar sensor calibration methods are cumbersome and require laboratory settings, making it difficult to update sensor calibrations in vehicles, especially after parts replacement or damage, which affects the reliability of vehicle assistance systems.
Innovation Solution
A computer-implemented method that determines ground truth values and calibration update data using sensor data and vehicle operation parameters, allowing for on-line calibration updates without a dedicated laboratory, utilizing a trained machine learning model to reduce computational complexity and improve reliability.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If radar sensors are calibrated in the plant or manufacturing line, then initial calibration is achieved, but recalibration becomes difficult after vehicle sale or part replacement
Solution Approach 1:
The system enables the radar sensor to perform self-calibration using naturally occurring environmental targets (vehicles, pedestrians, infrastructure) as reference objects. The calibration process is automated through machine learning models that process sensor data and update calibration parameters without human intervention, allowing the sensor to maintain itself throughout its operational life.
Solution Approach 2:
The calibration parameters are made dynamically adjustable based on environmental conditions and sensor performance degradation over time. The system continuously monitors sensor data quality and automatically updates calibration parameters when deviations are detected, transforming calibration from a static factory-setting process to a dynamic operational process.
2Measurement precision
If traditional calibration methods are used, then accurate calibration can be achieved, but it requires extensive laboratory or workshop environments
Solution Approach 1:
Environmental targets serve as intermediaries between the radar sensor and the calibration process. Instead of requiring specialized calibration equipment and facilities, the system uses naturally occurring objects in the vehicle's operational environment as reference targets, simplifying the calibration system while maintaining accuracy.
Solution Approach 2:
The patent replaces complex mechanical calibration systems (physical calibration equipment, laboratory setups) with computational methods using machine learning models. The calibration process is transformed from a physical/mechanical procedure to an information-processing task that can be performed using sensor data and algorithms.
3Ease of operation
If on-line calibration is implemented, then recalibration becomes possible without laboratory, but computational complexity increases
Solution Approach 1:
Instead of performing full calibration computations on all sensor data, the system selectively processes only relevant portions of data that contain information about calibration deviations. The machine learning models are trained to identify and process only the necessary features from sensor data, reducing computational burden while maintaining calibration effectiveness.
Data Source
AI summary
A computer-implemented method for updating a calibration of one or more sensors in a vehicle. The computer-implemented method includes obtaining sensor data from the one or more sensors. The computer-implemented method includes determining a ground truth value based on at least a part of the sensor data and on one or more vehicle operation parameters. The computer-implemented method includes determining calibration update data based on the ground truth value. The computer-implemented method includes updating the calibration using the calibration update data.