Adaptive Attitude Estimation Using a Hybrid Model-Learning Approach

TL;DR

This paper introduces a hybrid deep learning and model-based method for smartphone attitude estimation, improving accuracy by adaptively tuning sensor weights with neural networks in a pedestrian context.

Contribution

It presents a novel hybrid approach combining classical models with neural networks to adaptively estimate attitude from inertial sensors.

Findings

Outperforms traditional model-based methods in accuracy.

Neural network effectively adapts accelerometer weights.

Demonstrates robustness in pedestrian walking scenarios.

Abstract

Attitude determination using the smartphone's inertial sensors poses a major challenge due to the sensor low-performance grade and variate nature of the walking pedestrian. In this paper, data-driven techniques are employed to address that challenge. To that end, a hybrid deep learning and model based solution for attitude estimation is proposed. Here, classical model based equations are applied to form an adaptive complementary filter structure. Instead of using constant or model based adaptive weights, the accelerometer weights in each axis are determined by a unique neural network. The performance of the proposed hybrid approach is evaluated relative to popular model based approaches using experimental data.