Electronic training aids for archery: Miyauchi 2012

A teacher at a Japanese technical college describes the electronic devices he built for school archery clubs that have no coach, and the field test of two of them. The first is an aiming signal: a three-axis accelerometer and a digital compass sit on the bow and feed an Arduino board, which stores the tilt and the direction of the bow at the moment an arrow lands in the ten and then, on later ends, tells the archer when those same positions return within the tolerance set, one degree for example. The light version lights up an eight by eight LED matrix mounted where the sight goes; the sound version changes pitch as the bow approaches the reference angle.

Tested on a single beginner with one year of practice, 36 arrows with the device and 36 without, grouping improves clearly: the geometric mean distance between the three arrows of each end falls from 17.3 to 10.7 centimetres at 30 metres with the light signal, and from 11.9 to 5.3 centimetres at 18 metres with the sound. The swing from one end to the next also narrows, with the standard deviation going from 5.2 to 1.7 centimetres in the first case and from 3.6 to 1.3 in the second.

With a Kinect sensor, over 36 draws without a bow, an archer with three years of practice shows less scattered joint coordinates than a beginner with one year. The clearest differences are on the horizontal axis, for the hand and elbow of both the bow arm and the string arm and for the right foot, and in height for the left shoulder, the neck, the left hip and the string hand.

The aiming device uses a three-axis KXM52-1050 accelerometer for tilt and twist and an HMC6352 digital compass for direction, with an Arduino Pro Mini 328 board computing, on every shot, the difference from the stored angles. The test compares the same archer with and without the device: twelve ends of three arrows, 36 in all per condition, at 30 metres for the light version and 18 metres for the sound one. Grouping is not measured as distance from the centre of the target but as the geometric mean of the three sides of the triangle the arrows form, that is the cube root of the product of the three distances: the closer the figure is to zero, the closer the arrows sit to each other.

For the Kinect comparison the two archers repeated 36 full draws without a bow, so that the sensor could follow fifteen joints; for each joint the standard deviation of the coordinates was computed in all three directions. The article then describes three further devices with no numerical test: an automatic reader of impact point and score, with infrared distance sensors and servo motors at the four corners of an 80 centimetre target at 30 metres, which shows the position on an LED matrix in about three seconds for three arrows and which the author reports as correct between 60 and 100 per cent of the time; a recorder of bow-arm movement, worn on the wrist, which saves the three angles from anchor to the end of the release onto a memory card; and its extension to six joints, which plots the deviations of 36 draws from the reference position on one graph.

The tests have one participant at a time: one for the aiming device, two for the Kinect comparison. There are no statistics, there is no control group and the order of the two conditions is not stated, so the tighter grouping may come from the device or just as well from warming up or from getting used to the distance. The outcome measured is the distance between the arrows of one end, not the score, and no later check says whether the archer shoots better once the signal is taken away.

The Kinect comparison puts two individuals side by side who differ in years of practice but in everything else as well, and it covers draws made without a bow. The author himself writes that with a bow in hand the sensor loses the bow-arm elbow and places it at the top of the riser, which it mistakes for part of the body. The article is in Japanese with an English abstract, and its register is that of a teaching report: it presents home-built systems and a few example measurements, not a study designed to establish their effect.

This is a piece of engineering education more than of technique, and for someone who shoots it says mainly two things. The first is that an immediate external reference, even one as rough as a light or a note tuned to a shot that landed in the ten, was followed by tighter groups in a beginner left without a coach; how much of that belongs to the device rather than to simply having a repeatable reference, this work cannot say. The second is that the archer with three years of practice repeated his joint positions more closely than the beginner, which puts figures on something archers know: what separates a starter from someone further along is repeatability more than any single position.

For a club training on its own it remains a concrete example of what can be built with a few tens of euros of components and some programming. It proposes no training method, and the centimetres reported come from one person: they are not values against which to measure your own grouping.