I don’t think the iPhone does naive dead reckoning like that, as the sensitivity of the needed inputs is so high for it be reasonable in a consumer handheld. The cost of components would be way to high, then there’s the issues of power consumption, because you would need to poll all your sensors at a very high rate to get enough accurate data to perform naive dead reckoning accurately.
Instead I think the iPhone cheat a bunch, and use a mixture of step counting, magneto and gyro inputs to perform dead reckoning. The iPhone uses periods when you GPS is in use anyway to calibrate your stride length at different speeds, which when mixed with magneto data to estimate a course direction, would allow an iPhone semi-accurately perform dead reckoning, without needing high resolution, wide scale, accelerometer input.
In short, the iPhone cheats. It’s dead reckoning systems take advantage human bio-mechanics to simplify the problem of dead reckoning, at the cost of building a dead reckoning system that only works effectively in a handheld device that semi-permanently lives on your person. A perfectly reasonable assumption for a phone, but a terrible solution for a more generic dead reckoning system.
Instead I think the iPhone cheat a bunch, and use a mixture of step counting, magneto and gyro inputs to perform dead reckoning. The iPhone uses periods when you GPS is in use anyway to calibrate your stride length at different speeds, which when mixed with magneto data to estimate a course direction, would allow an iPhone semi-accurately perform dead reckoning, without needing high resolution, wide scale, accelerometer input.
In short, the iPhone cheats. It’s dead reckoning systems take advantage human bio-mechanics to simplify the problem of dead reckoning, at the cost of building a dead reckoning system that only works effectively in a handheld device that semi-permanently lives on your person. A perfectly reasonable assumption for a phone, but a terrible solution for a more generic dead reckoning system.