Research map / 3D vision / Depth and stereo
IONet: Learning to Cure the Curse of Drift in Inertial Odometry
Changhao Chen, Xiaoxuan Lu, Andrew Markham, Niki Trigoni
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| Replace continuous inertial integration with independent-window polar displacement prediction using a 2-layer bidirectional LSTM | Inertial odometry on multiple users and attachments · Multiple users and attachments | position error: 30%- 40% improvement | Improved |
| Replace continuous inertial integration with independent-window polar displacement prediction using a 2-layer bidirectional LSTM | Inertial odometry · Large-scale indoor localization | absolute position error | Improved |
| Use inertial-only IONet instead of visual-inertial Tango | Trolley tracking · Trolley tracking | trajectory accuracy | No clear effect |
| Use a 2-layer bidirectional LSTM instead of a vanilla RNN (ablation) | IONet polar-vector regression · Polar-vector regression validation | validation loss | Improved |
| Use a 2-layer bidirectional LSTM instead of a CNN (ablation) | IONet polar-vector regression · Polar-vector regression validation | validation loss | Improved |
| Use a 2-layer bidirectional LSTM instead of a 1-layer LSTM (ablation) | IONet polar-vector regression · Polar-vector regression validation | validation loss | Improved |
| Use bidirectional rather than unidirectional 2-layer LSTM (ablation) | IONet polar-vector regression · Polar-vector regression validation | validation loss | Improved |
| Train separate models for each phone attachment rather than training jointly across attachments (ablation) | IONet · Attachment-specific prediction | prediction performance | Improved |
| Increase input window length from 50 to 100, 200, or 400 frames (ablation) | IONet · Window-size selection | accumulative location error and predicted loss | Mixed |
| Train and test with measurements affected by high sensor bias rather than matched measurement domains | IONet · High-bias IMU measurements | odometry performance | Hurt |
Closest papers by meaning
- EqNIO: Subequivariant Neural Inertial Odometry (ICLR 2025 · similarity 0.88)
- Tartan IMU: A Light Foundation Model for Inertial Positioning in Robotics (CVPR 2025 · similarity 0.86)
- Adaptive VIO: Deep Visual-Inertial Odometry with Online Continual Learning (CVPR 2024 · similarity 0.85)
- iMoT: Inertial Motion Transformer for Inertial Navigation (AAAI 2025 · similarity 0.84)
- From Variance to Veracity: Unbundling and Mitigating Gradient Variance in Differentiable Bundle Adjustment Layers (CVPR 2024 · similarity 0.83)
- LEAP-VO: Long-term Effective Any Point Tracking for Visual Odometry (CVPR 2024 · similarity 0.83)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
