Sensor Data Drift Correction for Precise Speed Estimation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional methods for evaluating sensor data often result in significant drift due to sensor errors and incorrect initial values, leading to inaccurate measurements, especially when integrating acceleration data or determining speed.
Innovation Solution
A method that corrects raw and processed sensor data using a mathematical model, specifically by determining and removing drift through a probabilistic filter like the Kalman filter, and setting the speed gradient to zero when movement is constant, thereby improving measurement accuracy.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If sensor data is integrated over time to determine speed or position, then measurement capability is improved, but drift increases due to accumulating sensor errors
Solution Approach 1:
The patent implements a feedback mechanism by continuously monitoring the processed sensor data for drift characteristics and applying corrective transformations. The system uses the relationship between raw and processed data to detect drift and apply compensating transformations that feed back into the measurement pipeline, thereby maintaining accuracy over time.
Solution Approach 2:
The patent transforms the sensor data by applying drift compensation parameters that are calculated based on the relationship between raw and processed measurements. These parameter changes include correcting speed determinations and position calculations by applying transformations that counteract the accumulated drift effects.
2Measurement precision
If processed sensor data is used to improve measurement accuracy, then measurement precision is improved, but drift is magnified due to integration errors
Solution Approach 1:
The system establishes a feedback loop that compares processed sensor data against expected physical constraints and relationships. When drift is detected in processed data such as speed or position, the system applies corrective transformations that reference the original raw measurements, thereby preventing error magnification while preserving measurement precision.
Solution Approach 2:
The patent applies drift compensation transformations preliminarily to the processed sensor data before final measurements are taken. By pre-correcting the processed data using transformations derived from raw measurements, the system prevents integration errors from magnifying drift in subsequent calculations of speed and position.
3Measurement precision
If conventional correction methods are applied to sensor data, then some measurement accuracy is improved, but significant drift remains due to incorrect initial values
Solution Approach 1:
The patent fundamentally changes the correction parameters by using transformations that reference raw sensor measurements rather than relying on conventional correction methods that use fixed or estimated parameters. This parameter change allows the system to correct drift caused by incorrect initial values by continuously referencing the actual raw measurements throughout the correction process.
Solution Approach 2:
The system implements feedback by continuously monitoring corrected measurements for signs of drift and applying additional transformations based on the relationship between raw and processed data. This feedback mechanism ensures that even when conventional correction methods leave residual drift, the system can detect and correct it using the feedback loop.
Data Source
AI summary
A method for evaluating sensor data. In the method, firstly, raw sensor data and/or processed sensor data from at least one sensor are input and measurement data determined from the raw sensor data and/or the processed sensor data. The measurement data are then corrected on the basis of a mathematical model, wherein, on correction, drift of the raw sensor data and/or of the processed sensor data is determined and removed from the measurement data. The corrected measurement data are furthermore output.