Environmental feedback with ballistic properties

In his personal introduction, Patrick Smith described his research interest:

Other examples include archery, skeet shooting, and navigation by dead reckoning.

Is this amenable to a similar approach as in the Exp. Brain Res. paper “Sensorimotor delays in tracking may be compensated by negative feedback control of motion-extrapolated position” as recently discussed in the Delay in control loops topic?

[quote=“bnhpct, post:1, topic:15717, full:true”]
In his personal introduction, Patrick Smith described his research interest:

BN: Other examples include archery, skeet shooting, and navigation by dead reckoning.

RM: Archery and skeet shooting, yes; like putting, both have a ballistic (uncontrolled) component of the control loop between output and its effect on the controlled variable. Not so, however, with navigation by dead reckoning, which involves continuous control of a variable (one’s angle relative to a point on shore), very much like controlling optical angles when catching a fly ball.

BN: Is this amenable to a similar approach as in the Exp. Brain Res. paper “Sensorimotor delays in tracking may be compensated by negative feedback control of motion-extrapolated position” as recently discussed in the Delay in control loops topic?

RM: I don’t think so. There is no ballistic component in the tracking task; there is continuous control of the controlled variable (T - C + dT/dt). The extrapolation component of the controlled variable, dT/dt, helps out to the extent that the change in target position at t+tau is highly correlated with the change at t, which it is with sinusoidal target movement. The best you can do with a putt is cognitively compute the likely trajectory of the ball after it’s been hit.

RM: You really have no control over the ball after it has been struck by the putter (or the arrow has been released or the trigger has been pulled). All the actual control that’s happening in putting (archery and skeet shooting) occurs before the ballistic component of the behavior is set in motion. And a lot of it occurs in imagination before the shot is made (such as calculating the predicted effect of the slope and wetness of the green on the ball’s trajectory). Once all those calculations are made there must be control of the angle and force of the putter on the ball.

Best Rick

It does not vitiate your point, but for the record dead reckoning is not so simple.

In navigation, dead reckoning is the process of calculating one’s current position by using a previously determined position, or fix, by using estimations of speed and course over elapsed time.
Wikipedia

I’m remembering that in the latter part of his sailing career Josh Slocum, a fellow Vineyarder at least some of the time, famously sailed around the world by dead reckoning.

Slocum navigated without a chronometer, instead relying on the traditional method of dead reckoning for longitude, which required only a cheap tin clock for approximate time, and noon-sun sights for latitude. On one long passage in the Pacific, Slocum also famously shot a lunar distance observation, decades after these observations had ceased to be commonly employed, which allowed him to check his longitude independently. However, Slocum’s primary method for finding longitude was still dead reckoning; he recorded only one lunar observation during the entire circumnavigation.
Wikipedia

Polynesian navigation uses an an astonishing array of perceptual inputs. The book Hawaiki Rising by Sam Low (another Vineyarder as well as Hawaiian) discloses many of these in the course of telling how they’ve been recovered.

To the point, however, just how people learn and get better by practice is the question Patrick is interested in.

Bowling, tennis, ping-pong, darts, shuffleboard, bocci, croquet … endless examples. In curling teammates intervene ahead of the trajectory.

Practice strengthens memory as the basis of imagination. You may recall the demonstrations that basketball players improved in foul shots just by imagining successful shots. (Here’s a 1989 thesis investigation.) But by practice the execution becomes more consistent and reliable as intended. That’s where Patrick’s interest in motor control applies. Looks like both are needed.

Hi All,

I think you could definitely model it as a ballistic process where the controller outputs an initial angle and initial velocity of the golf ball. Physics does the rest and then you take the Euclidean distance between the final position of the ball and hole as an error term and then reorganise the putter’s parameters to improve performance in a closed loop fashion over successive trials. I don’t think that would be difficult to do.

You could even have different terrain on the putting green and the virtual putter should learn how to get the ball in even without having an internal model or knowledge of the topology of the putting green. Quite a cool experiment actually…

I would be more than happy to collaborate on this and code up the model.