- Remember that “physical meters” differ from “ultrasound meters” before calibration. Thus, in some cases, there is no point in measuring ultra-precisely (mm-level) in “physical meters”, because the difference with “ultrasound meters” will level any further potentially expected improvements in accuracy.
- In many cases, it is better to let the system measure all distances except height. Mixing “physical meters” and “ultrasound meters” before calibration may lead to accuracy degradation
I have the following questions:
* How can I perform calibration, in order to perform a validation of the installed system where I take metric measurements?
I asked an earlier question on how to perform validation and the resources in the reply referred to metric calibration but did not address best practices on how to achieve it or program it into the system? https://marvelmind.com/forum/viewtopic.php?f=2&t=35812
* Does lack of calibration imply that the output of positional tracking is is by default in ultrasonic meters units, including the output of your ROS driver nodes?
Our system is designed to deliver the robot from point A to points B, C, and D and return to A. In this case, you get your +-2cm accuracy of coming back to any point. The accuracy doesn't depend on the physical meters because everything operates in the internal coordinate system.
However, when there are external coordinates involved, for example, "physical meters" or some other external distances entered manually, for example, in the table of distance or points A, B, C, D, etc., are not defined by pushing the robot around and recording the coordinates and then letting the robot driving to the same locations autonomously. You will have a significantly more complex system and shall care about many "different types of meters" and different types of coordinate systems.
Why does this happen?
- Ultrasound-based system relies on the speed of sound. Without calibration, you can expect 1-3% inaccuracy of the distance measurement, because the speed of sounds depends on several parameters, but, first of all, the temperature. The system assumes a default 23°C, but it can be different, i.e., the absolute distance can be different. For the internal coordinate system, it doesn't matter because all distances will be scaled proportionally, and the resulting coordinates will be calculated correctly anyway. But when mixed with the external physical meters, up to 1-3% of distance error can be expected before calibration.
- Typical random taken absolute distance measurement based on our testing returns around 0.5-1% of absolute distance error, depending on the distance, sample, etc.
- The easiest way to perform the calibration is to measure the distance between the centers of the beacon, compare it with the ultrasound meters, and then adjust the temperature in the modem's settings until the distance match. The typically recommended distance for such a calibration is 6-15m
We don't recommend calibrating on the short distance (2-3m) because there are distance measurement error elements that don't depend on the distance - a small fixed error that depends on the HW and SW - not on the distance or temperature. Neither do we recommend measuring at a too large distance, for example, 30m, because the fixed error distance measurements would be too diluted by the distance measurement error due to temperature. Recommended distance of 10m is somewhat in the middle and returns the best result.
