Week 4 Summary
Pedestrian Tracking
Tracking a person is a challenging problem especially in outdoor environments that have to previous preparation. This capability would prove very useful for things like search and rescue, as well as tracking emergency response workers (like firefighters). The NavShoe is a concept attempting to address this problem by using a small motion sensor that fits in the laces and requires minimal battery.
Typically, using motion produces position errors that expand cubically in time. The NavShoe attempts to reduce this problem by taking measurements during stance to eliminate the error and reduce it to a linear problem (rather than a cubic one). There are a number of details about specific techniques used to resolve the heading drift problem in the sensor device.
In the implementation section, it describes how this process was achieved. I was slightly disappointed to hear that the GPS is accurate to 6 meters. This made me think that even though this technology is cool for accurate tracking – in most of the applications I can think of, a GPS is more than enough accuracy. There may, however, be some sort of medical applications requiring much more precision – but then again, a medical room is a controlled environment (indoors) and GPS can be replaced with much more accurate local positioning systems. There is quite a bit of rather boring Kalman Filter tracking math in this section that demonstrates the process used to implement the sensor. Interestingly – the iron in the shoe was assumed to mess with the registration, and so this was also accounted for in the filter. There is an exciting diagram showing the NavShoe’s path as registered by the system, and it is cool to see how accurate it seems to be.
The results of the outdoor experiment showed that the NavShoe seemed reasonably accurate within a few kilometers. With an available GPS, a lot of magnetic error can also be reduced, which further improves the system. Ultimately, the NavShoe excels in that a user can walk into a building, and out of GPS range, and for at least a few kilometers, the user can be accurate about continuing to track himself. This becomes a nice supplement to a MR system, which relies on accurate user tracking.
I wonder after having read this article – how reasonable is it to incorporate this device directly into a mobile phone?
Comments are closed.