CONIX Publication

Poster: LocoMote: Ultra-Low-Power Hardware Framework for Deep Inertial Odometry

Authors: Sandeep Singh Sandha, Ankur Sarker, Luis Garcia, Mani Srivastava, Junha Park, Josh Geronimo, Urvi Shah, Suyog Vyawahare

Abstract:

Deep inertial odometry has shown superior trajectory estimation resolution over classical techniques in GPS-blocked environments for diverse applications. Algorithmic advances in neural inertial navigation require complementary development of sensor tags capable of running odometry models in the wild. Deployment of inertial odometry requires a long-term, low-power, durable and inexpensive hardware framework. Keeping the challenges in mind, we have designed LocoMote, a compact, ultra-low-power, environmentally resilient and modular sensor tag configuration that pushes the state-of-the-art in inertial odometry hardware. LocoMote integrates inertial measurement units, barometers, GPS, real-time clocks, ambient sensors, long-term storage, communication and energy harvesting systems, all within a 32 X 32 X 10 mm^3 tag. The tag is suitable for widespread deployment for land-air-sea-space applications without incurring noticeable size, weight or power overhead. We illustrate the performance of the tag against competing proposals and justify the design decisions.

Release Date: 10/12/2021
Uploaded File: View