Sensor Calibration for Sticky Bot

Stroke Sensors

The stroke sensors measure (roughly) the angular displacement of the upperlimb from the commanded stroke position, and from this data we are trying to determine shear load on the leg. The first step in calibration is to normalize across the differences in the sensor manufacturing and our mounting.

If the stroke range is +-30 degrees from perpendicular to body we get ~15% error in our readings at the extreme (30 deg), which drops to ~7% error at 20 deg. Is this enough to warrant a second step in the calibration where we correct for this trig relationship?

With the hall effect sensors (see HallEffectSensorDetails) we can get very close to a linear output with position of the sensor. Therefore each sensor needs only a zero point and a slope in order to normalize all the readings. Instead of having a look-up table for each sensor (which would become space costly), the normalization will be computed on the fly. If we decide that the second step is needed, once the sensors have been normalized they can be multplied by a factor based on the position of their associated servo (with the factor set to 1 for perpendicular to the body). Since this is a trig relationship, we would use a look-up table instead of computing on the fly.

-- BarrettHeyneman - 07 Aug 2006