Abakcus
← All articles

Geometry · Everyday Mathematics · Pythagoras

The Geometry of Perfect Parking

A mathematician took Pythagoras to the kerb and came back with a formula for the smallest space you can parallel park in. Here is the whole proof, with a car you can drive.

Aerial photo of a silver car reversing into a parallel parking space between a red car and a blue car, overlaid with white geometric lines, a 40 degree steering angle and a 4.5 metre space.
A parallel park seen from above, with the geometry drawn on the tarmac. Every line in it comes from circles and right angles.

You have found a space. The road is wide, the traffic behind you is patient for once, and the gap between the two parked cars looks as if it might be long enough. Might. This is the moment when most of us guess, and when Simon Blackburn, a mathematician at Royal Holloway, University of London, decided to stop guessing. In November 2009 he wrote a ten-page report, The Geometry of Perfect Parking, with a simple question in its abstract: how much length, beyond the length of your own car, do you need to parallel park?

The answer is a formula with two square roots in it. It did the rounds in the newspapers that winter, usually printed as a joke about mathematicians overcomplicating things. That reading gets it backwards. The formula is the short version. Behind it is one of the cleanest pieces of everyday geometry I know: three right triangles, one circle, and nothing harder than the theorem of Pythagoras, the one a future American president once proved with a trapezoid. In this piece we will rebuild the whole argument from the tarmac up, and you can drive the car yourself along the way.

1.43 mThe spare room Blackburn's example car needs on top of its own length to park in a single reverse sweep. Under that, the front corner meets the car ahead.

The rules

What counts as a perfect park?

Before any geometry, we need a definition. Blackburn’s is strict. No shuffling back and forth, no creeping a few centimetres closer with each try, and certainly no handbrake turns. A perfect park is one clean movement: reverse into the space with the steering wheel at full lock until the car sits parallel to the kerb, then roll straight forward to the middle of the gap. That is all.

The model around that definition is as plain as he could make it. The road is flat and wide, so the car can swing in at any angle it likes. The kerb is straight and low, low enough that the bodywork can pass over it. The cars on either side are neat rectangles parked tight against the kerb. Your own wheels sit at the edges of your own rectangle. None of this is exactly true of a real street, and none of it is far off either. A good model is a street with the distractions taken out.

Here is that manoeuvre running in the example car. It reverses on right lock to point its tail at the gap, backs up in a straight line, then winds on full left lock for the final swing. Watch the brick-red dot on the front corner. On the last arc it travels along a circle, and the whole problem comes down to whether that circle clears the amber corner of the car in front. Drag the slider below 1.45 m and see what happens.

Play

The red line is the path of the front kerb-side corner. The dashed arc is the circle it must follow on full lock, centred on the point X.

Two things are worth noticing before we go on. First, the tight moment is not at the back of the space. It is at the front, where your nose swings past the rear corner of the car ahead. Second, the earlier parts of the manoeuvre hardly matter. However you arrive, the last thing you do is ride a full-lock circle down to the kerb, and that final circle decides everything. So the question becomes: where is the centre of that circle, and how big is it?

The mechanism

How a car actually turns

A car is a rigid object on four wheels, and a wheel only likes to roll in the direction it points. If the car is to turn without scrubbing its tyres sideways, every wheel must be travelling along its own circle, and all four circles must share one centre. The rear wheels cannot steer, so that centre has to lie somewhere on the line through the rear axle. Remember that sentence. The entire proof hangs on it.

The front wheels then have a job to do. The inside wheel is on a smaller circle than the outside one, so it must be turned further. Carriage builders knew this in the age of horses. A Munich coachmaker, Georg Lankensperger, worked out a linkage that does it automatically, and his agent Rudolph Ackermann patented it in England in 1818. It has been called Ackermann steering ever since. In our example car at full lock, the inner front wheel is turned about 41° and the outer one only about 29°. Look at the photo at the top of this page again and you will find the same pair of numbers drawn on the road, as 40° and 27°.

Play

Lines drawn through the rear axle and through both front wheels meet at one point, X. The solid circle, traced by the outer front wheel, is the kerb-to-kerb turning circle that car makers quote.

Ease the steering off with the slider and the point X slides away along the rear axle line, and the circles grow. Wind it back to full lock and you get the smallest circle the car can make. Its radius, measured out to the outer front wheel, is the number rr. For an ordinary family car it is about five and a half metres, which is far larger than the car is wide. That matters too: it means every part of the car is turning the same way round X.

The rear axle always points at the centre of the turning circle. Everything else in the proof follows from that one fact.

The ingredients

Four numbers you can look up

Blackburn chose his inputs carefully. Every one of them is something you can find in a car brochure or measure with a tape in your driveway.

  • rr, the radius of the kerb-to-kerb turning circle. Brochures give the diameter, so halve it.
  • ℓ\ell, the wheelbase: the distance from the centre of a front wheel to the centre of the rear wheel behind it.
  • kk, the front overhang: from the centre of the front wheel to the very front of the car.
  • ww, the width of the parked car in front of the space. Not yours. Theirs.
Xrℓkw
rturning radiusℓwheelbasekfront overhangwwidth of the car in front

And here is what they produce. The space must be at least as long as your car, plus this much:

spare room  =  (r2−ℓ2)+(ℓ+k)2−(r2−ℓ2−w)2  −  ℓ−k\text{spare room} \;=\; \sqrt{(r^2-\ell^2)+(\ell+k)^2-\left(\sqrt{r^2-\ell^2}-w\right)^2}\;-\;\ell-k

It looks like something to be afraid of. It is not. Read from the inside out, it is just Pythagoras used three times in a row, and each use has a triangle you can point at.

The proof

Three triangles and a circle

Freeze the car at the instant it finishes reversing. It is parallel to the kerb with its tail against the car behind, about to roll forward. Call the front kerb-side corner A, the kerb-side wheels E and F, and the centre of the turning circle X. The parked car in front has its rear corner at G, directly above the point H on the kerb. We know four lengths, and we want a fifth: the gap from A to H. Step through it.

DFEAHGX

The picture

The car has just finished reversing and sits against the kerb. A is its front kerb-side corner, E and F are the kerb-side wheels, X is the centre of the turning circle, and G is the corner of the parked car that A had to miss.

∣EX∣=r,∣EF∣=ℓ,∣AE∣=k,∣GH∣=w|EX| = r,\quad |EF| = \ell,\quad |AE| = k,\quad |GH| = w

That is the complete argument. Triangle EFX gives the distance from the rear wheel to the centre. Triangle AFX gives the radius of the circle the front corner rides. The circle hands that radius over to G. Triangle GKX turns it back into a distance along the kerb, and the last step removes the part of that distance that is already your car. Substitute each line into the next and you get the formula above, square roots and all. If you enjoy this kind of chain, where a right angle keeps turning up exactly where you need it, the circle hiding inside the 3-4-5 triangle is built from the same move.

The numbers

Try it with your own car

Blackburn’s example uses a turning circle of 10.8 m, so r=5.4r = 5.4, a wheelbase of 2.6 m, a front overhang of 1.3 m, and a parked car 1.7 m wide. The answer is 1.43 m. A car four and a half metres long therefore needs a space of just under six metres to be parked in one sweep. Now put in your own numbers. The presets are round, representative figures for each kind of vehicle, not the specification of any particular model.

You need your car’s length plus

1.43 m

Swing in at about 21° to the kerb.

0.00.81.62.44 m5 m6 m7 m8 mspare room (m) against turning radius r
Try

The curve shows how the spare room changes with turning radius while the other three numbers stay where you set them.

Play with it for a minute and some surprising things come out.

Most cars need about the same. A city car, a hatchback, an SUV and a pickup all land between roughly 1.4 and 1.7 metres. Bigger vehicles have bigger turning circles, but they also have longer wheelbases, and the two effects pull against each other. The length of the car matters to the size of the space, of course. It matters much less to the extra.

Your own width is nowhere in the formula. Blackburn points this out himself, a little surprised, since wide cars feel harder to park. In this model the width simply cancels out: the corner that matters is on the kerb side, and the far side of the car is out over an empty road. Width comes back the moment the road is narrow, or the moment you start shuffling.

The car in front matters a lot. Slide ww up and down. Behind a motorbike 0.8 m wide the example car needs only about 0.8 m of spare room. Behind a two-metre van it needs nearly 1.6 m. The wider the vehicle ahead, the further out its corner sits, and the longer your nose takes to get around it. So the feeling that it is harder to park behind a van is not nerves. It is geometry.

A tight turning circle is worth real metres. The London black cab is famous for being able to turn in about 25 feet, a radius under four metres. Put those numbers in and the spare room drops to roughly 0.7 m, half of what an ordinary car needs. That is the same lesson as the curve in the calculator: as rr shrinks toward the wheelbase, the centre X moves in close and the nose hardly swings out at all.

“I never like to hear the scrape of alloy against concrete when I park.”

Simon R. Blackburn, The Geometry of Perfect Parking, 2009

The approach

What angle should you come in at?

Most of us were taught something like this: point the tail at the gap, reverse straight back, and put on full lock at the last moment. Blackburn notes that the same diagram tells you what that angle should be. You want to wind the wheel on exactly when corner A has just passed corner G. At that instant the car has to be on its final circle already, so its angle to the kerb is fixed. It is the angle between the radius XG and the radius XA, which a little trigonometry on our two triangles gives directly.

θ  =  arctan⁡∣GK∣∣KX∣  −  arctan⁡ℓ+k∣FX∣\theta \;=\; \arctan\frac{|GK|}{|KX|}\;-\;\arctan\frac{\ell+k}{|FX|}

For the example car that works out, by my arithmetic, to about 21°. That is shallower than most people expect, and it is the angle the car holds in the animation near the top of this page. If the arctangents look unfamiliar, the unit circle is the place to make friends with them. Come in steeper than this and you will need more room than the formula promises. Come in shallower and you will touch the car ahead before you can turn.

There is one more detail hiding in the picture. As the car straightens, its rear kerb-side corner sits behind the rear axle, so it swings on a slightly bigger circle than the wheel does. For a 4.5 m car that corner passes about four centimetres over the kerb just before the car comes parallel. This is exactly why the model asks for a low kerb, and why real kerbs collect paint.

The long way

What if you don't mind shuffling?

Suppose the space is shorter than the formula allows. Are you beaten? Not in theory. Blackburn shows that in his model you can get into any space longer than your car, however small the margin, as long as you are patient enough. Each forward or backward move is a little S-bend, two opposite arcs at full lock, and each one carries the car a little closer to the kerb.

How little? Follow the midpoint of the rear axle. At full lock it travels on a circle of radius ρ=r2−ℓ2−12n\rho = \sqrt{r^2-\ell^2} - \tfrac{1}{2}n, where nn is the distance between the rear wheels. If you have a length cc to play with, one S-bend moves you sideways by

gain per move  =  2ρ(1−1−c24ρ2)  ≈  c24ρ\text{gain per move} \;=\; 2\rho\left(1-\sqrt{1-\frac{c^2}{4\rho^2}}\right)\;\approx\;\frac{c^2}{4\rho}

The square is the cruel part. Halve the spare room and each move gains only a quarter as much, so you need four times as many. With half a metre to spare, the example car gains about 1.6 cm per move and would need roughly a hundred moves to creep sideways by the width of a car. With 25 cm it needs more than four hundred.

Play

The animation is sped up, and it counts only the rear-axle arcs. A real car also has to keep its swinging corners clear, so real life is slower still.

Blackburn adds a nice wrinkle here. On a curved forward move the nose of the car briefly pokes further toward the kerb than where it ends up, so you cannot finish exactly at the kerb going forwards. The last move has to be in reverse. Anyone who has ever parked a long car in a short street already knows this in their hands.

The other kind

Nose first, and a curve called the tractrix

The report ends with a different problem: driving nose first into a bay, the way you do in a multi-storey car park. The geometry changes completely. In the delicate part of the turn you steer so that the front corner of the car runs along a straight line, as close as you dare to the car in the next bay. The front goes where you send it. The back just follows, and the path it follows is not a circle.

It is a tractrix, the curve of a thing being dragged. The story usually told is that around 1670 the Parisian physician Claude Perrault laid his pocket watch on a table, pulled the end of its chain along the table’s edge, and asked what path the watch takes. Newton, Huygens and Leibniz all studied the answer, and Huygens gave it its name. The watch always moves straight toward the hand that pulls it. The rear of your car does the same thing with respect to the front.

The faint circles all have the car’s length as radius and their centres on the line. The tractrix crosses every one of them at a right angle.

That property of the circles is, in fact, one way to define the curve. It is also why long vehicles cut corners: the back wheels of a bus take a shorter path than the front wheels, and the people who design junctions and car parks have to leave room for it. If you like curves that come with a story, the brachistochrone, the curve of fastest descent drew in many of the same seventeenth-century names.

The point

Does anyone park by formula?

No, and Blackburn never said they should. The report was commissioned by Vauxhall Motors and comes with a cheerful warning to take sensible precautions before testing the theory on a real street. Nobody is going to work out two square roots with a queue of traffic behind them. A good driver has the formula in their eyes and hands already.

What the formula does is explain what those eyes and hands have learned. It says why the front corner is the one to watch, why a van ahead makes the space feel smaller, why taxis seem to fit anywhere, and why a space that is only slightly too short costs you a great many moves instead of a few. That is a fair answer to the old question of what mathematics is good for. Sometimes it sends rockets to the Moon. Sometimes it just tells you, calmly and for certain, that you were never going to fit.

And all of it rests on a statement about right triangles that is about two and a half thousand years old. It sits first on the list of the seventeen equations that changed the world, and it is still earning its place every time somebody backs into a gap outside a bakery.

Sources

  • Simon R. Blackburn, The Geometry of Perfect Parking, Department of Mathematics, Royal Holloway, University of London, 30 November 2009. Commissioned by Vauxhall Motors.
  • Ackermann steering geometry, Wikipedia.
  • John Bryant and Chris Sangwin, How Round Is Your Circle? Where Engineering and Mathematics Meet, Princeton University Press, 2008.
  • E. H. Lockwood, A Book of Curves, Cambridge University Press, 1961.
  • The 21° approach angle, the kerb overhang and the shuffle counts are computed here from Blackburn’s model for his example car, taken as 4.5 m long and 1.75 m wide. The simulations use the same numbers.