Vehicle Fusion: Parameters and Recommended Settings
FusionHub provides three fusion nodes for vehicle positioning. Each combines the same sensors in a different way, so the right choice - and the right settings - depends on what data you have and what you need from the output. This page explains what each node does, what its key parameters control, and the recommended values to start from.
All nodes ship with sensible defaults. Most of the values below are already applied to a freshly added node; a few are recommendations you may want to enable for your setup (noted where they apply). You usually only need to adjust a handful of them for your vehicle and use case.
Which fusion node to use
| Node | Inputs | Use it when |
|---|---|---|
| GNSS-IMU Fusion | GNSS (RTK) + IMU | You have a good GNSS receiver and want accurate outdoor positioning. The general-purpose choice for most vehicles. |
| Odometry-IMU Fusion | Wheel speed + IMU (+ GNSS) | You want robust dead reckoning from wheel odometry, especially to coast through GNSS gaps. |
| Full Vehicle Fusion | Wheel speed + IMU + GNSS | You want GNSS accuracy and odometry-grade coasting through long GNSS outages (tunnels, parking structures, urban canyons). |
A few terms used throughout:
- Measurement noise (error) parameters tell the filter how much to trust a sensor. Lower = trust that sensor more. Lower the value of a noise parameter to lean on that sensor; raise it to lean on the others.
- RTK Fixed is the centimetre-accurate GNSS quality level. Most settings assume RTK is available.
- The IMU-to-vehicle rotation and GPS antenna offset describe your physical install. They are vehicle-specific and are not part of the tuning recommendations below - set them once for your vehicle.
GNSS-IMU Fusion
Fuses RTK GNSS with the IMU. The GNSS provides absolute position and heading; the IMU fills in smooth, high-rate motion between fixes and carries the solution through brief GNSS dropouts.
Sensor trust
| Parameter | What it does | Recommended |
|---|---|---|
| Acceleration Error | How much to trust the accelerometer. Lower trusts the IMU more; higher leans on GNSS position. | 0.2 |
| Angular Velocity Error | How much to trust the gyroscope for orientation. Lower trusts the gyro more; higher lets GNSS heading dominate. | 0.04 |
| GNSS Measurement Error | How much to trust GNSS position fixes (metres). Lower trusts GNSS more; higher lets IMU dead reckoning carry more weight. | 0.07 |
| Orientation from GNSS Error | How strongly GNSS heading corrects the fused orientation. Lower gives GNSS heading more authority. | 0.006 |
GNSS aiding
These switch on the extra corrections the receiver can provide. Leave them on when the receiver supplies that data; they are harmless when it does not.
| Parameter | What it does | Recommended |
|---|---|---|
| Use GNSS Orientation Measurement | Continuously correct heading from the GNSS (dual-antenna) heading. | On |
| Use GNSS Pitch Measurement | Correct pitch from dual-antenna or Doppler-derived pitch. | On |
| Use GNSS Velocity Measurement | Use the receiver Doppler velocity as a direct 3D velocity update. Improves recovery after dropouts. | On |
| Use Gravity Leveling Measurement | Keep pitch and roll observable from the accelerometer, even during sustained turns. | On |
Behaviour
| Parameter | What it does | Recommended |
|---|---|---|
| Smoothing | Removes the small visible position step at each GNSS update, at the cost of a little latency. | On |
| Require RTK Fix | Accept only RTK-Fixed (centimetre) GNSS. Turn off only if you must operate on lower-quality fixes. | On |
| Innovation Gate (sigma) | Rejects/de-weights GNSS measurements that disagree too strongly with the prediction (multipath protection). | 6.0 |
The remaining defaults (bias random-walk, ZUPT, RTK-recovery timing) rarely need changing and are tuned for typical automotive operation.
Odometry-IMU Fusion
Dead-reckons the vehicle from wheel speed and the IMU turn rate, corrected by GNSS position. Because wheel odometry is far more accurate than IMU integration for distance travelled, this node coasts very well through GNSS gaps.
Sensor trust
| Parameter | What it does | Recommended |
|---|---|---|
| Velocity Error | How much to trust the wheel-speed input. | 0.5 |
| Angular Velocity Error | How much to trust the turn-rate input. Lower gives smoother heading, slower turn response. | 0.05 |
| Measurement Error | How much to trust the GNSS position update (metres). | 0.5 |
Heading source
The vehicle heading comes from the IMU gyroscope. These settings control that source and what happens if the IMU stops.
| Parameter | What it does | Recommended |
|---|---|---|
| Use IMU Turn Rate | Take the turn rate from the gyroscope rather than from differential wheel speeds. Far less sensitive to wheel slip and track-width error. | On |
| Use GNSS Yaw Measurement | Use the GNSS (dual-antenna) heading as a direct heading correction while moving. Recommended when a dual-antenna receiver is available. | On |
| Wheel Turn-Rate Fallback | If the IMU stream is lost, fall back to a wheel-derived turn rate. Off by default: most vehicles do not supply a wheel turn rate, so a lost IMU leaves no heading reference and output is stopped rather than continued on a stale value. Enable only with a trustworthy wheel angular rate. | Off |
| IMU Timeout (s) | How long without an IMU sample before the IMU is treated as lost. | 0.5 |
Online calibration
These estimate slow error sources automatically and improve long-range dead reckoning. Keep them on.
| Parameter | What it does | Recommended |
|---|---|---|
| Estimate Wheel Scale | Learn a wheel-speed scale factor (tyre wear/pressure) from RTK-fixed travel. | On |
| Estimate Gyro Bias | Estimate and remove the gyro yaw-rate bias while stopped - the dominant heading-drift source during outages. | On |
| Detect Wheel Slip | Reduce trust in the inputs on ticks where IMU and wheel turn rates disagree. | On |
Full Vehicle Fusion
A two-stage chain: the odometry-IMU stage dead-reckons the vehicle, and its output feeds the GNSS-IMU stage. This gives you GNSS-grade accuracy while RTK is available and odometry-grade coasting when it is not - ideal for routes with tunnels or long GNSS occlusions.
Its settings mirror the two nodes above, grouped under an Odometry stage and a GNSS stage, plus a small number of chain settings. Use the Odometry-IMU recommendations for the odometry stage. The GNSS stage tracks the odometry-bridged position rather than raw GNSS, so its defaults are tuned for that role - leave them as shipped.
| Parameter | What it does | Recommended |
|---|---|---|
| Use Direct GNSS | When off, the odometry stage bridges GNSS outages and feeds a continuous position to the GNSS stage (best coasting through tunnels). When on, real GNSS feeds the GNSS stage directly. | Off for maximum outage robustness |
| Odometry stage | Dead-reckons the vehicle - this is where outage performance comes from. | See Odometry-IMU recommendations |
| GNSS stage | Tracks the odometry-bridged position; its defaults are tuned for that role and should be left as shipped. | Defaults |
Because the chain’s coasting quality lives in the odometry stage, focus any tuning there (wheel-scale and gyro-bias estimation, turn-rate trust).
Single-antenna operation
Full Vehicle Fusion also works with a single-antenna receiver - heading then comes from course over ground and the gyroscope instead of the dual-antenna baseline, and GNSS pitch aiding is unavailable. Change these settings:
| Parameter | Setting | Why |
|---|---|---|
| Dual GPS (GNSS source node) | Off | The receiver supplies no heading or pitch; the source reports course over ground instead. |
| Init from GNSS Orientation (odometry stage) | Off | A single-antenna course is only meaningful while driving. With this off, heading initialises through the speed-gated course-over-ground path, cross-validated against wheel speed. |
| Use GNSS Yaw Measurement (odometry stage) | Off | Course over ground is not a usable absolute heading at low speed or in reverse - leave this off. |
| GNSS stage | Defaults | The GNSS stage takes its heading from the odometry stage, so its orientation settings need no change. Pitch aiding switches off automatically; pitch and roll are then maintained by gravity leveling and zero-velocity updates - keep both on (default). |
| IMU-to-Vehicle Rotation | Calibrate | With no GNSS pitch reference, an uncompensated mounting tilt appears directly in the output pitch. Set a measured rotation or enable Auto IMU-to-Car Rotation. |
What to expect compared to a dual-antenna setup:
- Heading initialises only once the vehicle drives faster than the initialisation velocity threshold (3 m/s by default); there is no standstill heading initialisation.
- Through GNSS outages the heading is carried by the gyroscope alone and drifts faster than with dual-antenna heading aiding. Keep gyro-bias estimation on.
If Use Direct GNSS is on, the GNSS stage sees the raw receiver data: in single-antenna operation additionally turn off the GNSS stage’s Init from GNSS Orientation and Use GNSS Orientation Measurement.
Adjusting for your situation
- Want smoother output / brief GNSS gaps handled gracefully? Trust the IMU a little more - lower Acceleration Error and Angular Velocity Error.
- Want the solution to follow GNSS as tightly as possible? Trust GNSS more - lower GNSS Measurement Error and Orientation from GNSS Error.
- Heading not pointing exactly along travel on a straight road? Make sure the IMU-to-vehicle rotation is calibrated (or enable Auto IMU-to-Car Rotation and start parked on level ground), and keep GNSS heading aiding on.
- Long GNSS outages (tunnels)? Prefer Odometry-IMU or Full Vehicle Fusion and keep wheel-scale and gyro-bias estimation enabled.
- No GNSS heading source (single antenna)? Heading is initialised from course over ground once the vehicle is moving above the velocity threshold. See Single-antenna operation under Full Vehicle Fusion for the settings.