Wearable Runtime-Frame Velocity Estimation with Coordinate-Normalized ML
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Wearable computing devices face challenges in accurately estimating pose due to drift in inertial odometry, which increases over time, and developing machine learning models for each device configuration is costly and resource-intensive.
Innovation Solution
A wearable computing device uses an IMU to perform inertial odometry, combined with a machine learning model trained on transformed kinematic data to estimate pose, allowing for accurate pose estimation without requiring a model specific to the device's hardware configuration.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If inertial odometry is used for pose estimation in wearable devices, then pose estimation can be performed, but drift increases over time reducing accuracy
Solution Approach 1:
The system performs preliminary coordinate transformations during device configuration and model training phases. Transformation matrices are pre-computed based on IMU mounting locations, allowing the machine learning model to be trained once on reference coordinate frames and then applied to multiple device configurations without retraining, thus maintaining accuracy over time.
Solution Approach 2:
A coordinate transformation system acts as an intermediary between the IMU sensor data and the machine learning model. The transformation matrices convert kinematic data from device-specific coordinate frames to a standardized reference frame, enabling the ML model to process data from various device configurations uniformly and maintain estimation accuracy.
2Measurement precision
If machine learning models are trained for each device configuration, then estimation accuracy is improved, but development cost and resource consumption increase
Solution Approach 1:
The system creates a universal machine learning model that can process data from multiple device configurations. By using coordinate transformations to normalize input data from different IMU mounting locations to a reference coordinate frame, a single trained model serves multiple device variants, eliminating the need to develop separate models for each configuration.
Solution Approach 2:
The system changes the parameter representation of device configuration from model-specific training to transformation-matrix-based normalization. Instead of retraining models with different architectural parameters for each device, the approach transforms the input data parameters (coordinate frames) to a standardized form, allowing one model to handle multiple configurations.
Applied Scientific Principles
This section explains which scientific principles are used to turn an abstract innovation direction into a practical engineering solution.
Function Achieved in This Case
This approach reduces pose drift and enhances estimation accuracy by using a generalized machine learning model, reducing the need for costly retraining when device configurations change.
Implementation Method 1
an inertial measurement unit (IMU) including an accelerometer and a gyroscope
Implementation Method 2
an inertial measurement unit (IMU) including an accelerometer and a gyroscope
Data Source
AI summary
A wearable computing device, including a device body configured to be affixed to a body of a user. The wearable computing device may further include an inertial measurement unit (IMU) and a processor. The processor may receive kinematic data from the IMU while the device body is affixed to the body of the user. The processor may perform a first coordinate transformation on the kinematic data into a training coordinate frame of a training wearable computing device. At a first machine learning model trained using training data including training kinematic data collected at the training wearable computing device, the processor may compute a training-frame velocity estimate for the wearable computing device based on the transformed kinematic data. The processor may perform a second coordinate transformation on the training-frame velocity estimate to obtain a runtime-frame velocity estimate and may output the runtime-frame velocity estimate to a target program.