End-to-End Learning Framework for IMU-Based 6-DOF Odometry

Abstract

This paper presents an end-to-end learning framework for performing 6-DOF odometry by using only inertial data obtained from a low-cost IMU. The proposed inertial odometry method allows leveraging inertial sensors that are widely available on mobile platforms for estimating their 3D trajectories. For this purpose, neural networks based on convolutional layers combined with a two-layer stacked bidirectional LSTM are explored from the following three aspects. First, two 6-DOF relative pose representations are investigated: one based on a vector in the spherical coordinate system, and the other based on both a translation vector and an unit quaternion. Second, the loss function in the network is designed with the combination of several 6-DOF pose distance metrics: mean squared error, translation mean absolute error, quaternion multiplicative error and quaternion inner product. Third, a multi-task learning framework is integrated to automatically balance the weights of multiple metrics. In the evaluation, qualitative and quantitative analyses were conducted with publicly-available inertial odometry datasets. The best combination of the relative pose representation and the loss function was the translation and quaternion together with the translation mean absolute error and quaternion multiplicative error, which obtained more accurate results with respect to state-of-the-art inertial odometry techniques.

Robust Stereo Visual Inertial Navigation System Based on Multi-Stage Outlier Removal in Dynamic Environments

“…wherez i , H x i , H f i , and n i [27] are block matrices formed by the elements in Equation (22). However, a severe problem in Equations (22) and (23) is that the primitive covariance matrices P x f , P f f , P n f related to the feature are unknown. Therefore, the linearized measurement in Equation (23) need to eliminate the error feature Gp f i .…”

“…The optimization-based strategy is commonly more reliable than filtering-based and is more robust to outliers. Most recently, deep learning based-VINS [20][21][22] have employed the Convolutional Neural Networks (CNNs) and Long Short Term Memory (LSTM) networks [23] architecture into an end-to-end learning process [24] to handle vision and inertial data simultaneously. The evaluations are sufficient but less precise compared to the model-based systems.…”

Localization of Biobotic Insects Using Low-Cost Inertial Measurement Units

“…Some authors have created learning models that combine both LSTM and CNN approaches (e.g., [54]) while others have favored using ensemble learning methods in lieu of neural networks [55,56]. The majority of the models in the literature involve position or velocity estimation; however, these are not the only quantities that can be estimated.…”

“…The majority of the models in the literature involve position or velocity estimation; however, these are not the only quantities that can be estimated. Orientation [ 50 , 54 ] and speed [ 38 ] can be estimated and the noise parameters for Kalman Filter frameworks can be learned as well [ 53 ].…”

“…The idea behind the end-to-end framework is that a learned model is sufficient to output the pose of an agent given its inertial signals. Systems utilizing the end-to-end paradigm tend to incorporate deep neural networks involving CNNs or LSTMs [ 50 , 51 , 54 , 57 ]. The primary benefit of an end-to-end framework is the ability to implicitly model the relationship between the agent’s egomotion (measured by the inertial signals) and its pose.…”