The physical problem is to convert the acceleration data into the position data. As we know its formula is given like this:
Assuming the acceleration is constant over time the formula reduces to:
where 
are initial position and initial velocity, respectively.
Defining the Requirements Related with the Problem
The expected error is about 5 cm in 30 seconds. Using the formula above and assuming the initial velocity is 0, the error in the acceleration can be found as:
This means our resolution should be less than 
in order to reach 5cm error in 30 sec. Since for most of the sensors the resolution values are not available we should look at the noise densities. Noise determines the minimum resolution of the sensor and noise floor can be lowered by restricting bandwidth if the noise is Gaussian. [1]
So the required Noise density can be calculated like this:
You can see below the parking time vs. noise density for different error values.
The other very important parameters are range, frequency response, and non-linearity. Of course values can change from application to application but you can see the accepted values for navigation below.
The available sensors can be found on the web pages of producer like ADI, ST…
[1] Analog Devices FAQ 20.09.2007
