Research Company
Defenix advances the state of the art in inertial navigation systems (INS), blending classical estimation with modern machine learning for robust positioning when GPS is unreliable or unavailable.
Sensors
IMU · LiDAR · GNSS · Wheel
Core
Kalman · Factor Graphs
ML
Deep Fusion · Bias Nets
Targets
Robotics · Vehicles · Wearables
A focused, research-first software company.
To deliver reliable navigation everywhere—under dense canopies, indoors, underground, and in urban canyons—by fusing inertial measurements with learning-based perception.
We pair proven probabilistic estimators with neural components that model sensor bias, drift, and non-linear dynamics, yielding accurate and explainable state estimation.
Inertial navigation systems (INS), dead reckoning, visual-inertial odometry (VIO), SLAM, and resilient sensor fusion for edge devices.
Prototype pipelines, evaluation datasets, and embeddable libraries optimized for low-power hardware.
Selected problem areas we explore.
Neural modules that estimate IMU bias, temperature drift, and non-Gaussian noise in real time, integrated with EKF/UKF or factor-graph back-ends.
Robust dead reckoning using foot-mounted IMUs, wheel odometry, barometer, magnetometer, and opportunistic signals (Wi‑Fi, UWB) for drift-bounded position.
Tightly-coupled VIO with online calibration and loop closure, designed for edge accelerators and low-light conditions.
Methods & tools we like.
We collaborate with teams building navigation‑critical products.