TECHNICAL GUIDE

GNSS + IMU Dead Reckoning: When It Helps, When It Won't

Dead reckoning carries a navigation state through short GNSS gaps, but its error grows. Start with the sky condition, then choose between standard RTK, integrated RTK + DR, and tightly coupled GNSS/INS.

Measured across 58.8 km of highway, residential streets, urban canyon, and tunnel — with no wheel-speed input.

RTK Fixed, DR Coasting, then RTK Fixed A state rail shows a trusted RTK Fixed solution, error growing during a dead-reckoning gap, and a trusted RTK Fixed observation constraining drift again. RTK FIXED DR COASTING RTK FIXED Trustworthy Fixed Error grows during the gap Fixed constrains drift
  • CONTINUITY, NOT FROZEN ACCURACY
  • SKY CONDITION CHANGES THE ANSWER
  • RTK CLOSES THE LOOP

START WITH THE OPERATING CONDITION

Find the System Fit for Your Outage

Choose the condition that dominates the route. GNSS dead reckoning can preserve a continuous state through a short gap, but the right system depends on whether satellite observations are absent, degraded, or still trustworthy.

System fit

Integrated RTK + DR for a short, moving outage.

Why

The IMU keeps position, velocity, and attitude propagating while GNSS is unavailable. Error grows with distance until a trustworthy RTK Fixed observation returns.

Verify

Longest blackout distance · acceptable exit error · where RTK Fixed reliably returns

System fit

Improve antenna and RF first; consider tightly coupled GNSS/INS when measurements remain usable.

Why

An IMU added after a finished, multipath-biased GNSS position cannot recover a correct absolute solution or turn Float into Fixed.

Verify

Time spent in Float · multipath pattern · access to raw or quality-aware measurements

System fit

Validate standard RTK before adding a DR variant.

Why

If the receiver stays RTK Fixed, there is no continuity gap for dead reckoning to bridge. An IMU may still help with attitude or output rate.

Verify

Fixed availability · actual position error · whether attitude or a faster output is required

System fit

This is an output and attitude requirement, not an outage problem.

Why

IMU propagation fills the interval between GNSS updates and carries short-term angular motion, but it does not create drift-free absolute heading by itself.

Verify

GNSS solution rate · fused navigation output rate · required heading reference

KALMIX FIELD TEST

Field Evidence: Measured Dead Reckoning Accuracy and Its Limits

Three measured conditions lead to three different decisions: carry continuity through a complete blackout, do not expect an IMU to repair sustained Float, and do not add DR where RTK stays Fixed.

2× SPEED

TUNNEL · COMPLETE BLACKOUT

DR COASTING

2.447 km

Complete-blackout tunnel pass.

CEP50 1.9 m · R95 3.727 m

DR replaced a loss of continuity with a propagated navigation state; it did not preserve RTK accuracy.

For a downstream controller, a gradually growing error is usually easier to handle than a sudden loss or jump in position. The acceptance question is exit error and how soon trustworthy RTK Fixed returns—not the availability percentage.

GT-02 CONFIGURATION COMPARISON

WITH DR · DR COASTINGThe evaluated configurations maintained a continuous tunnel track.

WITHOUT DR · NO CONTINUOUS TRACKThe evaluated configurations returned only sparse tunnel points.

ACROSS THE COMPARED ROUTE · 100% vs 95.83–96.03% Valid-Epoch Completeness

10× SPEED

OVERHEAD OBSTRUCTION · DEGRADED SKY

RTK FLOAT

97.85% RTK Float across 9,981 epochs

7.481 km test stretch · CEP50 2.86 m · R95 7.42 m

The IMU did not turn Float into Fixed.

10× SPEED

RESIDENTIAL · RTK CONTROL

RTK FIXED

100% RTK Fixed across all evaluated residential epochs

2.83 km residential comparison segment

No coasting was required in this segment.

We use these runs to separate continuity from accuracy: a continuous state can be useful through a blackout without preserving RTK accuracy, and an IMU does not turn sustained Float into Fixed.

Review the Ground Truth evidence

THREE SYSTEM ROLES, NOT ONE IMU LADDER

GNSS/INS Fusion Depth: Match the System to the Navigation Problem

GNSS/IMU sensor fusion covers systems built for different jobs. Fusion depth, aiding, motion constraints, and inertial class change together—not in a single ranked order.

System role How it fuses Best fit Boundary Kalmix path
Integrated RTK + DR COMPLETE BLACKOUT Industrial-grade 6-axis MEMS propagating a finished GNSS solution. Short moving outages; higher-rate fused output Cannot correct a multipath-biased finished GNSS position SCOUT SERIES AVAILABLE NOW
Tightly Coupled GNSS/INS SUSTAINED DEGRADED SKY Automotive-grade MEMS with raw-measurement GNSS/INS fusion. Partial sky where some observations still carry information Still needs usable satellite or external constraints EXPLORER COMING SOON
FOG GNSS/INS Reference REFERENCE & EVALUATION Fibre-optic GNSS/INS for reference trajectories and uncertainty evaluation. Reference trajectories and system evaluation Not a retail upgrade path for an integrated receiver FORERUNNER COMING SOON

If the sky stays healthy and RTK remains Fixed, standard RTK is the positioning baseline; add an IMU only when attitude or a higher-rate fused output is itself required.

USEFUL OUTPUTS, CLEAR INTEGRATION CONDITIONS

What GNSS + IMU Adds—and What It Still Needs

A fused state is only as usable as its measurements, mounting geometry, calibration, and motion assumptions.

Trajectory continuity

Carries a continuous position and velocity state through a short outage; error grows until a trusted update returns.

Attitude and angular motion

Adds roll, pitch, and short-term yaw propagation; it does not create drift-free absolute heading.

Higher-rate fused output

Fills intervals between GNSS updates with a fused state; it does not increase the GNSS observation rate.

Three different heading sources

Course over ground Single-antenna direction of travel while moving; not the orientation of a stationary body.
Baseline heading Dual-antenna orientation while GNSS is available; its meaning depends on antenna geometry and mounting.
IMU yaw propagation Short-term change from gyro integration; it needs motion or an absolute reference to limit drift.

What the integration must supply

  1. Rigid mounting and known geometryMount rigidly; record axes, mounting angle, lever arm, and output point. Revalidate stored geometry when deploying or remounting.
  2. Calibration and observable initializationCalibrate sensor and installation terms, then begin with trusted GNSS and observable motion. Calibration reduces repeatable error; it does not remove drift.
  3. Platform-specific aidingWheel speed, maps, or other aids must match the platform; validate timing, scale, direction, and failure cases.

The IMU carries the state. Trusted observations close the loop.

DECISION BOUNDARIES

When Dead Reckoning Is Not the Answer

Static operation with an absolute-heading requirement

A single antenna plus a six-axis IMU cannot supply drift-free stationary yaw. Use dual-antenna GNSS or another absolute heading reference.

Aircraft, handheld, or pedestrian motion

Full-6DoF aircraft motion and changing device or body orientation invalidate ground-vehicle assumptions. Use platform-specific GNSS/INS or PDR carrying-mode logic.

Extended GNSS loss

Without an external constraint, inertial error continues accumulating. Consider vision, UWB, LiDAR, beacons, maps, or odometry.

Unexpected accuracy loss in healthy sky

The IMU is usually not the primary limit. Check corrections, antenna placement, multipath, convergence, and receiver configuration.

FREQUENTLY ASKED QUESTIONS

GNSS, IMU, and Dead Reckoning FAQ

What does dead reckoning add to a GNSS receiver?

Dead reckoning propagates position, velocity, and attitude through a short complete blackout. It preserves a continuous state, not RTK accuracy; error grows until RTK Fixed or another reliable observation constrains it again.

What does the tunnel test show?

In one KALMIX FIELD TEST, a 2.447 km tunnel segment reported 85.463% DR availability, CEP50 1.9 m, and R95 3.727 m. This is one measured pass, not a universal product specification; mounting, calibration, motion, temperature, and outage distance all matter.

Is GNSS + IMU the same as a GNSS/INS system?

Not always. GNSS + IMU describes the sensors present; GNSS/INS describes a system estimating a shared position, velocity, attitude, and error state. An integrated DR receiver and a tightly coupled estimator use those sensors differently.

Will dead reckoning fix a prolonged partially blocked sky?

Not always. Under sustained partial blockage, the receiver may remain in degraded Float rather than enter pure coasting, so an IMU may not recover a trustworthy absolute position. Test the GNSS state and blockage pattern, not only the scene name.

Does dead reckoning improve RTK accuracy in open sky?

Usually not by itself. With healthy sky, RTK already supplies the absolute position; the IMU mainly adds continuity, attitude, or a higher-rate fused state.

Do I need wheel speed for GNSS dead reckoning?

No. The Kalmix evidence on this page was measured without wheel-speed input. Wheel speed can reduce vehicle-model drift, but its timestamp, direction, scale, tyre condition, and slip must be validated.

Is GNSS heading the same as IMU yaw?

No. A single GNSS antenna normally reports course over ground while moving; dual antennas can observe a baseline orientation. An IMU propagates yaw from angular motion, but yaw drifts without motion or another absolute reference.

How does RTK constrain IMU dead reckoning?

RTK/IMU fusion depends on short-term inertial propagation and trustworthy absolute updates. Because inertial error accumulates, RTK Fixed position and GNSS velocity constrain drift; Float or multipath-biased positions should not be treated as equal resets.

Does a higher-rate output require dead reckoning?

Not necessarily. An IMU can propagate a smooth fused state between GNSS updates, but first compare the GNSS solution rate and the fused navigation output rate. If continuity and attitude are unnecessary, native GNSS output may be simpler.

DEFINE THE OUTAGE FIRST

Bring the Route and Error Budget

If you are evaluating a receiver or module with dead reckoning, share the outage type, acceptable exit error, RTK Fixed return point, vehicle or platform, mounting geometry, and any wheel-speed or external aids. Kalmix will recommend the simplest architecture that fits.