Category Archives: curve-fitting

clothoid weekend update

For context, see past posts in the curve-fitting category that I just created. To recap:

The curves I’ve been drawing are the paths made by a point moving at constant speed at an angle which is a piecewise quadratic function of path length. Curvature, the first derivative of angle, is continuous.

Such a path that hits a given sequence of dots is fully determined if it loops, but otherwise it has two degrees of freedom. For any angle and curvature at the starting dot, there is a quadratic coefficient that lets the path reach the next node, and likewise for the next.

My current code starts with an estimate for the length of each segment (between two dots) and the angle at its midpoint, and uses these basis functions to fit those angles: a constant, a linear function, and a family of “solitons”: piecewise quadratics, zero outside a sequence of four dots, discontinuous in the second derivative at each of those dots. For n segments, there are n-2 solitons, so the constant and linear functions are needed to consume the last two degrees of freedom.

Eventually I noticed a flaw in this scheme: the curvature of the resulting path is the same at both ends, namely the slope of the linear component, because the solitons contribute nothing to it. That’s appropriate for ā€˜Cā€™, but wrong for plenty of other strokes; in ā€˜Sā€™ the end curvatures ought to have opposite sign.

The next thing I’ll try is a least-squares quadratic fit to the whole sequence, then fit the residues with solitons as before. That should be an improvement but it’s not ideal; curvature is a local feature. Perhaps I’ll think of something better later.


Here, each curve hits alternate dots: first exactly (above), then with offsets pushing it toward the other curve. Below is the result of eight iterations.

With enough iterations, the top of ‘s’ eventually gets a more symmetrical arch, as the change in curvature is spread more evenly.

But many runs get stuck in the fitting phase, and I don’t yet know why. Attacking one likely flaw didn’t help.

Scribbles: The Ensmoothening, Part II

One thing I noticed in that last series of charts is that more than one degree of discontinuity doesn’t help: the best-looking curves are mostly on the diagonal, where only the last nonzero derivative is discontinuous. Here, therefore, are those curves all together.

In column zero, the tangent angle is piecewise constant; in column one, it is a piecewise linear function of path length, resulting in six circular arcs; in column two it is piecewise quadratic, resulting in six clothoid arcs with continuous curvature; and so on.

Of course the arcs are approximated by cubics; to improve the match, I put a knot wherever any derivative crosses zero, as well as at the discontinuities. (See the knots.)

ensmoothening scribbles

Presented for your consideration: the somewhat disappointing results of an experiment in using piecewise polynomial spirals, of varying degrees of continuity, to fit the Takana — disappointing in that few if any of the curves are as pretty as I hoped.

I treat here only those that can be drawn with a single stroke. (The others can be built by combining subsets of these strokes.) In each chart, the degree of continuity increases downward, and the degree of the polynomials increases to the right.

A polynomial spiral is a curve whose tangent angle is a polynomial function of arc length; it has the form integral(exp(i*f(t))). (I implement it as a Taylor series.) In principle, f could be any real-valued function. If f is constant, you get a straight line; if f is linear (leftmost column in these small charts), you get a circle; if f is quadratic, you get an Euler spiral or Cornu spiral or clothoid, which is much used in railroads and highways to avoid sudden changes in lateral acceleration.

Here f is a least-squares fit to the step function which is the direction of the squared stroke. The top row of the chart shows continuity of degree zero: the component arcs meet, but that’s all; f is discontinuous. Degree one: the tangent angle is a continuous function of arc length. Degree two: the first derivative of tangent angle with respect to arc length, i.e. the curvature, is continuous. Degree n: the (n-1)th derivative of tangent angle, i.e. the (n-2)th derivative of curvature, is continuous.

Click each chart to extend it.

Later: I have come to a couple of conclusions. In most of these charts, the best entry to my eye is where f is piecewise quadratic with one continuous derivative. More than one degree of polynomial above the continuous degree adds little fidelity and detracts from beauty.

Continue reading