StarIO: A Lightweight Inertial Odometry for Nonlinear Motion
Shanshan Zhang, Siyue Wang, Qi Zhang, Liqin Wu, Tianshui Wen, Ziheng Zhou, Xuemin Hong, Ao Peng, Lingxiang Zheng, Yu Yang
AI summary
Problem
Existing data-driven inertial odometry methods suffer from significant drift during complex nonlinear motions because lightweight edge-deployed models lack the capacity to capture intricate nonlinear relationships in IMU signals.
Approach
StarIO employs a Star Operation to implicitly map IMU data into a high-dimensional nonlinear space, augmented by a dual-wing attention mechanism and multi-scale gated unit to efficiently capture temporal, channel, and local motion dynamics.
Key results
- Reduces Absolute Trajectory Error by 5.21% on RoNIN compared to R-ResNet
- Achieves state-of-the-art accuracy across six datasets with only 2.762M parameters
- Cuts FLOPs by 34.3% and peak memory by 10.4% relative to R-ResNet
- Effectively models complex nonlinear motion patterns and long-range contextual dependencies
Why it matters
Enables accurate, drift-resistant inertial odometry for resource-constrained edge devices and consumer-grade applications like AR and privacy-sensitive navigation.
Abstract
Inertial odometry (IO) is an attractive approach for consumer-grade localization. However, existing data-driven IO methods often suffer from significant drift under complex nonlinear motion patterns (e.g., turns), as they struggle to cap- ture the nonlinear relationships between Inertial Measurement Unit (IMU) signals and motion states. To address this issue, we propose a lightweight IO model, StarIO. Specifically, we first apply the Star Operation to project IMU signals into a high- dimensional implicit nonlinear feature space, enabling effective extraction of the complex nonlinear motion characteristics that typically cause drift. We then capture contextual dependencies across both the temporal and channel dimensions to enhance trajectory estimation over long sequences. In addition, we introduce a multi-scale gated unit that fuses fine-grained local motion dynamics with contextual information to achieve a com- prehensive representation of motion. Extensive experiments on six representative open-source datasets demonstrate that StarIO achieves a superior trade-off between model lightweightness and localization accuracy. For example, on the RoNIN dataset, our approach reduces the ATE by 5.21% compared to R-ResNet while using only 2.762M parameters.