This month’s blog came about following a conversation with an engineer at a recent automotive show. They were wondering why, during the creation of a configuration file, we ask users to confirm the accuracy with which measurements are being made. More specifically, they wanted to know why the default accuracy was 10 cm, when most people can measure far more accurately? And they’re both very good questions.
We’ll start by clarifying what we’re talking about. Before one of our inertial navigation systems (INS) can be used, a configuration must be uploaded into it. A configuration is a file (technically it’s several files, but they’re saved and sent to the INS as one) that is created and uploaded to your INS by the NAVconfig program. Contained within the configuration file is information about:
- which features should be enabled and how the system should behave
- how you have installed it in the vehicle and where certain components are in relation to the IMU
And it’s the measurements relating to how the system is installed and where components like the antennas are located relative to the IMU measurement origin that require the user to specify their measuring tolerance.
The need for tolerance
So why do we use a default tolerance of 10 centimetres when it’s clear most people can measure accurately within 1—2 mm? Firstly, it can be difficult to get some of the required measurements when working around the car. The INS may be installed in the boot, for example, where it’s hard to see. This means it’s tricky to measure the vertical distance between the IMU origin and an antenna mounted at the front of the roof. Even so, it’s still reasonable to expect someone to be within one or two centimetres of the correct value. So, why the 10 cm default we use?
There’s another reason for the generous tolerance used when defining the position of the primary antenna, for example. This is that the INS isn’t really asking you for the precise location because it can work that out far more accurately than you. This is despite it being computationally intensive because there are so many places the antenna could be. Instead, the system asks you to help it by defining an imaginary cube within which it should limit its search. So when you set the antenna’s location within NAVconfig and use the default tolerance of ±10 cm, you’re really just limiting the area in which the INS needs to find its solution to a 20 x 20 x 20 cm cube. The exception to this is the distance between the primary and secondary antennas (where used). This must be measured as accurately as possible, ideally to within 3 mm.
The need for accuracy
So why bother being accurate with your measurements at all? Why not just leave them all at zero and set the tolerance to ±5 m; after all, that should be a large enough search area to include most cars and vans, shouldn’t it? The answer is, of course, you could! Just be prepared for a very long wait before the INS achieves anything like its stated accuracy. But in the same breath, be careful about being too tight with your tolerances or sloppy with your measurements.
For example, say you’re confident that you’ve measured the antenna position to within one millimetre. If you set the tolerance in NAVconfig to within 0.001 m, you’re assuring the INS that the centre of the antenna is absolutely, categorically somewhere within that specific eight cubic millimetre point in space. But suppose you’re out, just slightly. In that case, the system is never going to find a good solution for the antenna, because you were so confident it was in position x. In this case, you should again be prepared for a long wait.
Confidence in configuration
Armed with this information, we hope the next time you create a configuration you’ll feel confident to leave the tolerances at their default values, which are normally enough to allow for most measuring errors. Or, if you’re particularly unsure, and don’t mind the system taking slightly more time to achieve its stated output accuracy, you could always increase them very slightly.