Self supervised training enables inertial odometry without ground truth
Self-Supervised Inertial Odometry
Robotics
Summary
Tracking movement accurately usually needs labeled data showing exact paths, which can be hard to get. The authors demonstrate that a model can learn to estimate motion from just short bursts of sensor data without needing exact paths. By learning from a simple method that integrates data over short time windows, the model ends up predicting velocities more accurately than the simple integration method itself. This approach works across various robots and walking people, matching the accuracy of models trained with ground-truth data.
What this means in practice
- •For robotics engineers: Train inertial odometry models for different robots without collecting external ground-truth trajectories, reducing setup time and cost.
- •For mobile device developers: Enable accurate motion tracking on new smartphones or wearables by training models solely from sensor data during normal use.
Authors
Yuheng Qiu, Yutian Chen, Carlton Cort, Weison Huang, Shibo Zhao, Pieter Abbeel, Rocky Duan, Koushil Sreenath, C. Karen Liu, Carmelo Sferrazza, Junyi Geng, Sebastian Scherer, Guanya Shi
Abstract
Learning inertial odometry for a new platform usually requires ground-truth trajectories. We show that an inertial odometry model can instead be trained from scratch by a myopic teacher that integrates the IMU over each short window independently, a scheme we call integration self-supervision.Interestingly, although supervised only by short-window velocity changes, the student learns to directly predict velocity, and does so more accurately than its integration teacher. Pose graph distillation then refines the model with sequence-consistent targets from a pose graph over many windows. Training on a new platform therefore no longer requires collecting ground-truth trajectories, but only calibrating the IMU and recording data during normal operation. Across six datasets spanning pedestrians, drones, quadrupeds, and a humanoid, the same recipe with no velocity or position labels trains models that come within $0.07$~m/s average velocity error of a same-architecture supervised baseline.