Neural Networks Versus Conventional Filters for Inertial-Sensor-based Attitude Estimation
This work addresses the need for more accurate attitude estimation in moving objects, but it is incremental as it builds on existing methods with specific optimizations.
The paper tackled the problem of attitude estimation using inertial sensors by comparing neural networks to conventional filters, finding that neural networks outperform filters across all motions only with domain-specific optimizations, achieving improved accuracy in real-time estimation.
Inertial measurement units are commonly used to estimate the attitude of moving objects. Numerous nonlinear filter approaches have been proposed for solving the inherent sensor fusion problem. However, when a large range of different dynamic and static rotational and translational motions is considered, the attainable accuracy is limited by the need for situation-dependent adjustment of accelerometer and gyroscope fusion weights. We investigate to what extent these limitations can be overcome by means of artificial neural networks and how much domain-specific optimization of the neural network model is required to outperform the conventional filter solution. A diverse set of motion recordings with a marker-based optical ground truth is used for performance evaluation and comparison. The proposed neural networks are found to outperform the conventional filter across all motions only if domain-specific optimizations are introduced. We conclude that they are a promising tool for inertial-sensor-based real-time attitude estimation, but both expert knowledge and rich datasets are required to achieve top performance.