Calib-Net: Calibrating the Low-Cost IMU via Deep Convolutional Neural Network

Abstract

The low-cost Inertial Measurement Unit (IMU) can provide orientation information and is widely used in our daily life. However, IMUs with bad calibration will provide inaccurate angular velocity and lead to rapid drift of integral orientation in a short time. In this paper, we present the Calib-Net which can achieve the accurate calibration of low-cost IMU via a simple deep convolutional neural network. Following a carefully designed mathematical calibration model, Calib-Net can output compensation components for gyroscope measurements dynamically. Dilation convolution is adopted in Calib-Net for spatio-temporal feature extraction of IMU measurements. We evaluate our proposed system on public datasets quantitively and qualitatively. The experimental results demonstrate that our Calib-Net achieves better calibration performance than other methods, what is more, and the estimated orientation with our Calib-Net is even comparable with the results from visual inertial odometry (VIO) systems.

MEMS Inertial Sensor Calibration Technology: Current Status and Future Trends

“…Ref. [105] proposed a lightweight and efficient deep convolutional neural network Calib-Net for low-cost IMU calibration. Calib-Net employs dilated convolutions for spatiotemporal feature extraction, learning to generate gyroscope measurement compensation dynamically, and introducing a mathematical calibration model to design the training and calibration framework.…”

Deep Learning for Inertial Positioning: A Survey

“…A similar approach is [28], who calibrates gyroscope using ConvNet, reporting good attitude estimation accuracy. Calib-Net [31] is another ConvNet framework that denoises gyroscope data by extracting effective spatio-temporal features from inertial data. Calib-Net is based on dilation ConvNet [67] to compensate the gyro noise, as illustrated in Figure 3.…”

Bayesian Calibration of MEMS Accelerometers

“…Error Model 1) Noiseless case: To understand the error model, we first ignore the stochastic noise. Several error models to calibrate the MEMS accelerometer sensors have been proposed in the past [1], [6], [8], [13]- [15]. All these linearly relate the threedimensional output vector a of the measured acceleration to the true acceleration g as follows a = Sg + b.…”

“…In this approximation, we consider the likelihood (8) and treat the parameters S −1 , b, and σ ′ as Bayesian variables. We choose the same priors for S, b, σ ′ as in III-A.…”

“…Common to all methods (autonomous or not) is the estimation of the parameters of a sensor response function [7]. The response function can span a wide range of complexity, from simple linear transformations [6] to complex neural networks [8]. Various estimation and optimization techniques have been developed to determine these parameters, including least squares approaches and statistically sound and unbiased maximum likelihood methods [5]- [7].…”