Efficient Training Data Collection for Distance Sensor Arrays through Data Correction and Augmentation Approaches
Sogo AMAGAI, Shin'ichi Warisawa, Rui Fukui
AI summary
Problem
Reusing training data across different distance sensor arrays degrades machine learning accuracy due to sensor output characteristic variations, while collecting new data is highly time- and labor-intensive.
Approach
The authors propose two data conversion methods: a linear correction technique based on measured sensor characteristics, and a few-shot learning approach that uses a small batch of new data to train a mapping function for augmenting old data.
Key results
- OC-based correction reduces RMSE by up to 23% versus direct transfer
- Few-shot augmentation reduces RMSE by up to 58% versus direct transfer
- Accurate mapping functions can be built from only a small batch of target sensor data
- Practical guidelines are established for efficient minimal data collection
Why it matters
Enables rapid, low-cost deployment of machine learning-based sensor systems by bypassing time-consuming recalibration processes.
Abstract
Several machine learning (ML)-based measurement systems have been proposed to estimate difficult-to-measure quan- tities from the values of distance sensor arrays. However, variations in sensor output characteristics (OCs) can lead to degradation in the estimation accuracy when transferring training data acquired from the original acquisition sensors to new target sensors. More- over, acquiring training data from target sensors is time and labor intensive. We propose two methods to convert previously collected training data to reflect different OCs, enabling their repeated use. For evaluation, we use a device that estimates the relative position and orientation of vehicles based on the values of distance sensor arrays. The correction approach for the training data based on the OC data reduces the root-mean-square error (RMSE) by up to 23% compared with transferring training data. The augmenta- tion approach transforms the training data into data that include different OCs using a mapping function constructed from a small batch of training data. Furthermore, a method for collecting a small batchoftrainingdatatoachieveahigherOCconversionaccuracyis demonstrated. The RMSE is reduced by up to 58% by the proposed method compared with transferring training data. The results of this study demonstrate the feasibility of the practical applications of ML-based measurement systems using distance sensor arrays, which may facilitate the development of simple and fast calibration methods.