The angle that doubles

A trailer being towed forwards in a straight line settles into line behind the tractor and stays there. The same trailer reversed in a straight line does the opposite: whatever small misalignment it starts with grows, slowly at first and then not slowly, until the rig is folded and the driver has run out of options.

Every driver of anything with a trailer knows this and most explanations of it reach for something about weight, or about the trailer pushing, or about the pivot being behind the axle. None of that is needed. It is the sign of one exponent in a first-order equation with no force in it anywhere.

The equation, linearised #

The towed axle’s heading obeys

$$\dot\theta_1 = \frac{v}{d}\sin(\theta - \theta_1),$$

with $\theta$ the tractor’s heading, $\theta_1$ the trailer’s and $d$ the trailer’s length. Write $\psi = \theta - \theta_1$ for the hitch angle, hold the steering straight so that $\dot\theta = 0$, and measure progress by distance driven $s$ rather than by time — which removes the speed and leaves a statement about geometry:

$$\frac{d\psi}{ds} = -\frac{\psi}{d} \qquad \text{for small } \psi.$$

Driving forwards, $s$ increases and $\psi$ decays as $e^{-s/d}$. Reversing, $s$ decreases and the same equation gives $e^{+|s|/d}$.

The length scale is the trailer’s own length and nothing else. Not its weight, not its wheel size, not the tractor’s wheelbase: $d$, the distance from the hitch to the trailer’s axle.

Measured, and where it stops #

The equation is integrated with the full sine — no linearisation anywhere in the computation — starting at $\psi_0 = 0.02$ rad for a six-metre trailer:

The reversing column tracks $e^{s/d}$ to three figures for the first eight metres and then falls away from it, which is the sine saturating: $\sin\psi$ is smaller than $\psi$ and the growth slows. By 24 m the linearisation predicts 1.09 rad and the mechanism has reached 1.00.

That is worth being clear about because the saturation might sound like a rescue and is not. A hitch angle of 1 radian is 57°, which is a jackknifed rig against the back of the tractor cab on most vehicles. The exponential ends because the mechanism has run out of angle, not because anything has stabilised.

The forwards column is the same law with the other sign and it has the reassuring property: $0.02 \to 0.0004$ over 24 m, a factor of fifty in four trailer-lengths, and the misalignment is gone.

Doubling distance: $d\ln 2 = 4.16$ m for a six-metre trailer. Every four metres of reversing doubles whatever hitch angle there is, so an error that is invisible at the start of a manoeuvre is a problem twenty metres later and nothing in between gave any warning.

The two equilibria #

The sine has two zeros in a full turn, and they are the two configurations at which a straight tow changes nothing.

$\psi = 0$, in line. Stable driving forwards, unstable reversing.

$\psi = \pi$, folded back on itself. Unstable forwards, stable reversing.

The second one is the finding that makes the popular account wrong in an interesting way. A rig reversing in a straight line is not merely unstable at zero; it is attracted to the folded configuration. Reversing a trailer without steering does not drift somewhere arbitrary — it converges, at the same exponential rate, on being jackknifed.

And that is the honest explanation of why a jackknifed rig stays jackknifed while reversing. It is not stuck; it is at a stable equilibrium of the reversing dynamics, and staying there requires nothing of the driver.

The turn a rig cannot hold #

Reversing is the famous half. The forwards half has a result that is less known and is sharper.

Hold the steering at a fixed angle $\phi$ and drive forwards. The tractor turns at $v\tan\phi / L$; the trailer’s heading chases it at $(v/d)\sin\psi$; and the two balance when

$$\sin\psi^* = \frac{d}{L}\tan\phi.$$

So a steady turn has a steady hitch angle, and for a 2.7 m wheelbase towing a 6 m trailer the closed form gives 11.2108° of hitch at 5° of steering, 23.0689° at 10° and 53.9811° at 20°. Driving the rig two hundred metres at each of those steering angles and reading the settled hitch angle off gives 11.2108°, 23.0689°, 53.9811° — every digit, from a route with no algebra in it.

The interesting part is the right-hand side. $\sin\psi^*$ cannot exceed one, so when $\tan\phi > L/d$ there is no equilibrium at all. The trailer cannot turn fast enough to keep up with the tractor, the hitch angle grows without bound, and the rig jackknifes going forwards.

The critical steering angle is $\arctan(L/d) = 24.23°$ for this rig, and the turn radius it corresponds to is

$$\frac{L}{\tan(\arctan(L/d))} = d.$$

A rig cannot hold a steady turn tighter than its trailer is long. Not approximately: the critical radius is exactly $d$, and the tractor’s wheelbase cancels out of it entirely.

An ordinary car has about 33° of lock, which is a 4.16 m turning radius — well inside the 6 m floor. Drive that car forwards on full lock with a six-metre trailer behind it and the rig folds: after sixty metres the hitch angle has reached 146° and is still growing. That is why a long trailer cannot be taken round a tight corner in one movement, and it is a geometric floor rather than a matter of care.

What a jackknife is not #

The word invites a particular explanation, and this site’s habit is to check it.

The explanation is that a folded rig has reached a singular configuration — that something in the mechanism has degenerated, a rank has dropped, a set of motions has become unavailable. Every other field on this site has configurations like that: a four-bar at a toggle, a parallel platform at a singularity inside its own workspace, an arm where a direction of motion is lost. It would be an entirely reasonable place to expect another.

The measurement says no. The growth vector of a car and trailer, computed at five hitch angles:

Identical. Every configuration is reachable from every other; the mechanism at 180° is exactly as manoeuvrable, in the only sense this field measures manoeuvrability, as the mechanism in line. There is no singularity, no rank drop and no lost direction.

What it is instead #

If the geometry has not changed, something else has, and it is worth stating precisely.

Undoing a hitch angle requires driving forwards. The decay is $e^{-s/d}$ with $s$ the forward distance, so unwinding an angle of $\psi$ down to something manageable takes a few trailer-lengths of forward travel — twelve metres or so for a six-metre trailer. Steering helps, and steering is what an experienced driver uses, but the resource being spent is forward distance.

A jackknifed rig is one that has run out of forward room. It has folded because it was reversing, and it was reversing because there was nowhere to go forwards. That is why jackknives happen in yards and against loading bays and not on open roads: the configuration is recoverable and the space to recover it in is not there.

So the word names a situation rather than a configuration, and the situation is a configuration plus an absence of room. That is a distinction the growth vector cannot see, because the growth vector is computed in an empty plane, and it is one this field consistently declines to take on: what obstacles do to a mechanism’s reachable set is somebody else’s subject.

It also explains the standard advice, which is otherwise mysterious: pull forward and start again. Not because the fold is irreversible but because undoing it costs distance in a direction the driver has been avoiding.

Speed is not in the equation #

There is one variable conspicuously missing from everything above, and its absence is the most useful practical consequence this rung has.

The hitch angle’s law is written in $s$, the distance driven, and not in $t$. The angle doubles every 4.16 m of reversing for a six-metre trailer, and it does so whether those metres take two seconds or thirty. Speed does not appear, it does not appear because the kinematics have no time in them, and the constraint that produced the equation — a wheel rolls along its own heading — is a statement about the path rather than about the schedule.

That contradicts what almost every driver believes, and the belief is worth taking seriously because it is not stupid. Reversing a trailer slowly does work better. What is wrong is the account of why. Going slowly buys time per metre, which is time to see the angle growing and time to make a correction, and it buys nothing at all in the growth itself: the same number of doublings occurs over the same number of metres, and a rig reversed at walking pace arrives at the same hitch angle as one reversed at speed, at the same point on the ground.

The distinction has teeth. It says the quantity a driver is actually rationing is reversing distance, and that no amount of care changes the exchange rate. Two doublings is two doublings, so a manoeuvre needing twelve metres of reversing on this rig will multiply any initial misalignment by a factor of seven and a half regardless of how it is driven, and the only way to reduce that number is to need fewer metres.

Which is exactly the standard advice, arriving from the arithmetic rather than from the folklore. Get the approach right before starting to back, because the initial $\psi_0$ is multiplied by $e^{s/d}$ and a small one is worth a great deal — halving the starting misalignment is worth exactly one doubling distance of reversing, 4.16 m of the manoeuvre bought back for nothing. Pull forward and start again is the same statement with the sign reversed: forward metres divide the angle at the same rate reverse metres multiply it, so the correction is never worse than proportionate, and a rig that has run away is fixed by driving forwards about as far as it drove backwards.

It also explains why the difficulty scales so badly with the size of the yard rather than with the size of the rig. Two rigs of the same trailer length face the same exponent; the one manoeuvring in a space that demands twenty metres of reversing faces $e^{20/6}$, twenty-eight fold, and the one that needs eight faces threefold. The trailer’s length sets the rate and the site sets the exponent, and only one of those is on the vehicle.

The one thing speed genuinely changes sits outside this field, and it is worth naming so the claim is not over-read. Everything here is the kinematic path, and a rig also has inertia, tyre sidewalls that deflect and a hitch with slack in it — none of which is in the equation and all of which respond to how fast the manoeuvre is driven. So slow reversing is better for reasons that are real and are about forces, plus one reason that is about the driver rather than the vehicle. What it is not is a way of slowing the exponential down, and a driver who believes it is one will be surprised by the same angle arriving at the same place.

Two trailers, and why nobody reverses a road train #

Hitch a second trailer behind the first and the same equation applies again, with the first trailer’s axle as the second’s hitch. Each stage has its own length scale and its own exponent, and reversing multiplies them: an error at the back grows as $e^{s/d_2}$ while the error it induces at the front grows as $e^{s/d_1}$, so the rig has two unstable modes rather than one and they compound.

The practical consequence is categorical rather than quantitative. A single trailer’s instability is correctable, because the driver has one angle to watch and one steering input to correct it with — one control for one unstable mode. A double has two unstable modes and still one steering input, so no steering history can hold both straight, and reversing a two-trailer rig any distance is not a matter of skill.

That is why road trains are uncoupled to be reversed, why a two-trailer set in a yard is broken up and shunted one unit at a time, and why the rule is a rule rather than a caution. It is also the reason the growth vector’s length matters twice over: each towed unit adds a coordinate to be placed and a mode to be held, and the two difficulties arrive together while the number of controls stays at two.

Where the sine’s maximum is, and what it does #

There is a real geometric feature at $\psi = 90°$ and it is not a singularity either.

$\sin\psi$ is at its maximum there, so the trailer’s heading is changing as fast as it possibly can for a given speed: $v/d$ per unit time, or one radian per trailer-length driven. Past ninety degrees the sine falls again and the rate drops, even though the angle is still growing.

That gives the runaway its characteristic shape — slow, then fast, then slowing again as the rig approaches the folded equilibrium — and it is the reason a jackknife feels as though it happens suddenly. The fastest part is in the middle, between about 60° and 120°, which is precisely the range in which a driver is trying to correct.

What is not true is that anything about the mechanism changes at ninety degrees. It is the maximum of a smooth function, and a maximum of a rate is not a loss of anything.

What a driver actually does about it #

The equation says the angle grows while reversing with the steering held straight. What a driver does is not hold the steering straight, and it is worth saying what steering buys, because the answer is not stability in any general sense.

Steering changes the tractor’s heading, which changes $\psi$ directly. So a driver reversing a trailer is running a feedback loop: watch the hitch angle, steer to reduce it, and the rate available is $v\tan\phi/L$ against the instability’s $v\psi/d$. The loop can win — comfortably, for small angles, since $\tan\phi/L$ at even ten degrees of lock is 0.065 against an instability of $\psi/6$ — but two things bound it.

The lock runs out. Past the critical steering angle the tractor cannot turn fast enough for the trailer to keep up in either direction of travel, and beyond a hitch angle of about 40° on this rig there is no steering input that reduces it while reversing.

The correction is backwards. To reduce a hitch angle while reversing the driver must steer towards the side the trailer has gone, which is the opposite of what the same input does driving forwards. That reversal of sign is the whole of what has to be learnt, and it is a consequence of the sign of the exponent rather than a quirk of vehicles.

The trailer that is easier #

The equation says exactly what makes a trailer manageable and the answer is one number.

Length. A longer trailer has a longer length scale, so its angle grows more slowly per metre reversed: an eight-metre trailer doubles every 5.5 m against a two-metre trailer’s 1.4 m. That is why a small box trailer or a boat trailer is far harder to reverse than an articulated lorry, which contradicts everybody’s expectation and is what every driver of both reports.

And nothing else. Not the weight, not the load, not the tyre pressures, not the tractor. The tractor’s wheelbase appears in the rig’s other equations and not in this one.

The design consequence is that a trailer’s reversing behaviour is set the moment its axle position is chosen, and axle position is chosen for load distribution and for swept path rather than for this. Moving a trailer’s axle back lengthens $d$, which makes it easier to reverse and worse at cutting corners — a trade nobody states in those terms and everybody makes.