99
Warm-up period
During the first 15 minutes of operation the system will not conform to specification.
During this period the Kalman Filter runs a more relaxed model for the sensors. By
running a more relaxed model the system is able to:
3. Make better estimates of the errors in the long term (if it does not get these correct
then they become more difficult to correct as time goes on).
4. Track the errors in the inertial sensor during their warm-up period (when their errors
change more quickly than normal).
During this period it is necessary to drive the vehicle or the errors will not be estimated
and the specification will not be reached. The NCOM output message (and CAN outputs)
includes status information that can be used to identify when the required specification
has been met. These are plotted in the example below.
The warm-up period is a concern to some customers but it is often very simple to
overcome. Below is an example of a good warm-up procedure that did not involve a lot
of work for the user. In this example the key features are:
• The RT3003 was configured well—the GNSS antenna position, Advanced slip
options and dual-antenna separation were measured accurately in advance.
• The RT3003 was turned on as soon as possible. In this case it took us 15 minutes to
get all the other equipment sorted out. The RT was stationary for most of this
period—which is not a problem.
• Although in this example the RT3003 was receiving corrections from a base station
while stationary, it is not necessary. The base station should be working before the
dynamic driving starts so the RT3003 can use the best information to self-calibrate
(if a base station is not being use this does not apply).
• There are 6 minutes during which the car was driven in figures of eight. From the
graphs you can see the RT3003 is accurate almost after the first figure of eight, after
that the improvement is very small.
The trick is to turn the RT on early, do not reconfigure it (which resets it) or cycle the
power.
Figure 47 shows the route driven and Figure 48 shows the accuracy estimated by the
Kalman filter for various output parameters during the first 25 minutes. The quality of
initialisation would have been the same if the stationary period was 10 minutes, followed
by 5 minutes of driving. The time on the graphs is the time from initialisation. In this
example the RT3003 was initialised 25 s after starting up; the quality of initialisation
would be the same if it had been not been initialised for the first 10 minutes, then
initialised and driven for 5 minutes.
To Next Page
Loading...
