Data-Driven Based Cascading Orientation and Translation Estimation for Inertial Navigation

Recently, data-driven approaches have brought both opportunities and challenges for Inertial Navigation Systems. In this paper, we propose a novel data-driven method which is composed of cascading orientation and translation estimation with IMU-only measurements. For robust orientation estimation, we combine a CNN-based neural network with an EKF to eliminate orientation errors caused by sensor noises. We additionally propose a hybrid CNN-Transformer-based neural network which exploits both spatial and long-term temporal information to regress accurate translations. Specifically, we conduct detailed evaluations on datasets acquired by iPhone and Android devices. The result demonstrates that our method outperforms state-of-the-art methods in both orientation and translation errors.

Authors

• Xiangyu Deng
• Shenyue Wang
• Chunxiang Shan
• Jinjie Lu
• Ke Jin
• Jijunnan Li
• Yandong Guo

Keywords

• Neural networks
• Estimation
• Inertial navigation
• Robot sensing systems
• Intelligent robots
• Orientation Estimation
• Neural Network
• Data-driven Methods
• Translation Accuracy
• Extended Kalman Filter
• Sensor Noise
• Android Devices
• Orientation Error
• Spatial-temporal Information
• Root Mean Square Error
• Training Set
• Convolutional Neural Network
• Diagonal Matrix
• Long Short-term Memory
• Angular Velocity
• Attention Mechanism
• Error Propagation
• Position Of Point
• Preferred Direction
• Inertial Measurement Unit
• Inertial Measurement Unit Data
• Gyroscope
• Pose Estimation
• Magnetic Flux
• Dead Reckoning
• Physics-based Methods
• Absolute Orientation
• Visual-inertial Odometry
• Hybrid Network