# Can an accelerometer function at a constant velocity?

I am designing an anti-theft device that works as such: an accelerometer detects if the device is experiencing a percent change in acceleration greater than a to-be-determined value. If so, Bluetooth will engage and check if the device is increasing in distance from the user. If so, the anti-theft system will trigger. If a thief were to move away at a constant velocity, would the accelerometer fail to detect a change in acceleration (or any acceleration at all, for that matter) and as such would the device fail? If so, is there any way to get past that limitation?

As clarification, this is a small anti-theft device meant for use with laptops.

• Constant velocity is no acceleration. Your thief could slowly accelerate while staying under your acceleration limit. – JRE Jul 18 '20 at 21:22
• But you yould integrate over the acceleration and trigger the system with some threshold velocity, rather than acceleration. – jusaca Jul 18 '20 at 21:41
• Apollo, how would you discriminate between an accidental bump of the device and a thief who suddenly accelerates the device to a desired removal velocity and then holds the device at this velocity afterwards? – jonk Jul 18 '20 at 21:53
• I would use the accelerometer to detect engine vibrations or something instead when there shouldn't be any (or just detect that the engine is on directly). But that doesn't stop them from towing your car away. – DKNguyen Jul 18 '20 at 22:07
• JRE, that is my worry. Jusaca, theoretically that is true but from what I understand, in practice the associated noise would render my integrated value meaningless. – Apollo Jul 18 '20 at 23:12

Under normal conditions, an accelerometer will measure 1g acceleration from gravity moving at constant velocity (more-or-less) along the surface of the earth. It feels like it is accelerating upwards, and that is indistinguishable from actually accelerating upwards from the pov of the accelerometer.

Acceleration, by definition, is the rate of change of velocity (the derivative wrt time).

Integrating acceleration can give some indication of velocity, but accelerometers drift and unless the acceleration is fairly quick you won't really be able to distinguish that drift from actual acceleration.

You can, with an accelerometer, detect jarring and quick motions such as the initial acceleration of a car from a parked position. Of course there may be innocent shocks and so on (tapping on glass, earthquakes etc.) that could give false positives.

You could use a magnetometer to detect changes in orientation wrt the earth's magnetic field (may not be reliable in some locations). You could use GPS to detect change in position (may not be possible if the signal cannot be received).

• So this device is actually small, meant for use with laptops and not cars. As such it would be relatively easy for a thief to slowly pick a laptop up and walk away with it. Size and power draw are major concerns, which is why we nixed the idea of GPS (though we haven't looked into magnetometers, thank you for that suggestion). If we were to set the accelerometer threshold low enough, could it register virtually any motion and bypass the potential vulnerability of not detecting a low acceleration? – Apollo Jul 18 '20 at 23:23
• If it's set too low, bump the table or walk across a wooden floor and it detects the acceleration. So, maybe it will work, you'd have to do some research. – Spehro Pefhany Jul 18 '20 at 23:24

Realistically, some version of what you propose will work. You may have to dig in to the datasheet for the accelerometer and tune things a bit to get it to work reliably. But it is pretty much impossible in real life for a person to pick up an object and carry it away without triggering the motion detection of the accelerometer. In fact, there can be false detection if people stomp their feet or whatever, so tuning and testing is critically important (make sure to schedule time for it).