Raw MEMS output is not motion. A gyroscope drifts, an accelerometer cannot tell gravity from acceleration, and a magnetometer is pulled by the steel around it. We build the inertial sensing and sensor fusion that turns nine noisy axes into stable six-degrees-of-freedom pose — position and orientation — and fuse it with UWB, GNSS or vision when absolute truth is needed.
Faststream builds the full chain — sensor bring-up, calibration, the estimator and the real-time implementation — not a filter in isolation.
Six degrees of freedom is full rigid-body motion: three translational axes (position — x, y, z) plus three rotational axes (orientation — roll, pitch, yaw). It is the difference between knowing only which way a device is turned (3-DoF, like a phone's orientation) and knowing where it is and how it is turned at once — which is what AR/VR headsets and controllers, drones, robots and tracked tools all need.
Because the errors integrate too. A gyroscope has a small bias that, integrated over time, becomes an unbounded heading error; double-integrating accelerometer noise to get position diverges within seconds. The whole craft is fusion: the accelerometer's view of gravity and the magnetometer's view of north bound the orientation drift, and an absolute reference — UWB, GNSS or vision — bounds the position drift. Dead reckoning on inertial alone is only ever a short-term bridge between absolute fixes.
An attitude and heading reference system: orientation (roll, pitch, yaw) derived from a 9-axis IMU through a fusion filter — complementary, Mahony, Madgwick or a Kalman variant. It gives you a stable attitude cheaply, often on the sensor's own MCU, but it is orientation only. Adding position — the full 6-DoF pose — is where an absolute reference and a heavier estimator come in.
It is a trade of compute against accuracy and against which sensors you are fusing. Complementary and Madgwick/Mahony filters are light enough for a constrained MCU and excellent for orientation. An extended or error-state Kalman filter (EKF/ESKF) is the workhorse for full pose and for fusing inertial with UWB, GNSS or vision; an unscented filter (UKF) earns its cost only under strong non-linearity. We pick the lightest estimator that meets the accuracy and latency the application actually needs.
Position vs pose — UWB, AoA, TDoA.
TECHNOLOGYThe vision input to VIO / SLAM.
TECHNOLOGYMotion sensing inside ADAS.
TECHNOLOGYThe embedded layer it runs on.
SERVICESSensor & MEMS silicon we design.
INDUSTRYWearables and spatial devices.
If drift, calibration or a real-time fusion budget is on your critical path, tell us the sensors, the accuracy target and the platform.