How long can you navigate without GNSS?

It is the first question every integrator asks, and the honest answer is a method rather than a figure. The number depends on three things, and you already know two of them.

The three things that set the answer

1. How long is the outage? Seconds through a short tunnel is a completely different problem from an hour of contested operation. Start with the worst case on your route or mission, not the average.

2. What else is aiding the solution? An inertial sensor running completely alone is the hardest case and the rarest. Almost every real platform has something else: wheel odometry, air data, a Doppler log, vision, a magnetometer, fixed beacons. Every one of those extends the answer, often dramatically.

3. How much error can you accept at the end? Five metres and fifty metres are different products. This is the number people skip, and it is the one that actually decides the specification.

Why there is no single number

Inertial navigation is dead reckoning. Every answer is the previous answer plus the newest measurement, so every error stays in the total and compounds.

Heading error is usually what gets you. A small constant gyro error becomes a growing heading error, and heading error turns into position error in proportion to how far you have travelled. Drift a degree off and every kilometre you drive puts you roughly 17 metres off track. This is why gyro bias stability is the specification people argue about.

Accelerometer error compounds faster but usually matters less. Because acceleration is integrated twice to get position, a constant accelerometer bias produces position error growing with the square of time. On most vehicles, distance is known well enough from odometry or air data that heading dominates in practice.

Unaided inertial error is never bounded. It accumulates. A good sensor accumulates it slowly and predictably enough that you can size the worst case, which is what your engineering actually needs.

Working out your own number

Step 1. Write down your worst outage. The longest tunnel, the deepest cutting, the longest stretch of contested airspace.

Step 2. Write down the error you can tolerate at the exit. Be specific about which error: lateral, along-track or total.

Step 3. Count your aiding. If you have a speed source, say so. It changes the problem from a hard one to a manageable one, because distance stops being an inertial estimate.

Step 4. Work back to a gyro grade. As a rough guide, in degrees per hour of bias stability: consumer above 10, industrial 1 to 10, tactical 0.01 to 1, navigation 0.001 to 0.01, strategic below 0.001. A consumer part is lost in seconds. A tactical unit with aiding is useful for minutes to hours. Navigation grade free-navigates far longer.

Step 5. Then ask for measured data at your outage length, not a bench figure. See [How to read an IMU datasheet].

One real data point

We drove 21 minutes through Montreal with GNSS positioning denied. Heading came from each unit’s gyro alone, with a 1 Hz speed input for distance and no EKF.

Both units ran identical aiding, so the 3.6x gap is set almost entirely by heading drift. Both publish the same 0.5 °/hr minimum in-run bias on paper.

That last point is the useful one. Two units whose datasheets look close behaved very differently once they were moving, which is the entire argument for asking about measured performance rather than headline specs. The full data is in [our case study].

What we will not tell you

We will not quote you a free-inertial drift figure for our units, because we have not published one. If a supplier gives you an endurance number without asking about your outage length, your aiding and your error budget, ask them which of the three they assumed.

Frequently asked questions

Can I navigate indefinitely without GNSS?

Not without some other correction. Inertial error accumulates without limit. What you can do is accumulate it slowly enough that the mission finishes inside your tolerance, or add an aiding source that periodically pulls the error back.

Does adding odometry really help that much?

Yes. It removes distance from the inertial estimate entirely and leaves heading as the dominant error. For ground vehicles it is usually the single highest-value aiding source available, and it is often already on the bus.

Is a better IMU or better aiding the bigger lever?

Aiding, when you can get it. Sensor grade, when you cannot. Underground, subsea or in contested airspace, aiding is exactly what is missing, which is why those applications specify the highest grade they can carry.