Event
SSIO Self-Supervised Inertial Odometry Paper Released
1 reports · 1 sources
Yuheng Qiu and others released a paper on arXiv proposing a self-supervised inertial odometry training scheme called SSIO: a "myopic teacher" independently integrates the IMU within each short window to supervise the student model, and the student is then further optimized through pose graph distillation. The authors claim that on six datasets including pedestrians, drones, quadrupeds, and humanoid robots, models trained without velocity or position labels have an average velocity error within 0.07 m/s of a supervised baseline with the same architecture; training on a new platform requires only IMU calibration and recording of routine operational data, with no need to collect ground-truth trajectories.
Report timeline
- SSIO: A Self-Supervised Inertial Odometry Method Without Ground-Truth Trajectories
The paper proposes a self-supervised inertial odometry training scheme called SSIO, in which a myopic teacher independently integrates the IMU within each short window to supervise the student model, an approach called integration self-supervision. Although the student is supervised only by short-window velocity changes, it learns to predict velocity directly and does so more accurately than its integration teacher. Pose graph distillation is then used to further refine the model with a cross-window consistency objective.
arXiv cs.RO ·