Bicycle stability

Why does a moving bicycle stay upright when the same bicycle standing still falls over?

Stand a bicycle up, let go, and it falls almost at once. Give the same bicycle a shove down a gentle slope with nobody on it and it will run a long way, wobbling, correcting, and staying up until it slows. Nothing was added to it. So the upright state is not a property of the machine but of the machine in motion — and the usual one-line answer, that the spinning wheels act as gyroscopes, is not the reason. What, then, is?

Start with what balancing requires of anything, bicycle or not. A body on a narrow base of support is unstable: tip it a little and gravity now acts to the side of the contact point, so the lean grows, and grows faster the further it goes. Only one recovery is available — you must move the base of support back under the falling mass.

Now the constraint that makes a bicycle interesting. Its base of support is the line between two contact patches, and it cannot be shifted sideways directly. The only way to move it is to steer, so the wheels track a curve and the line of support swings across underneath. Fall to the left, and the wheels must go left. So the question narrows sharply: what turns the handlebars toward the fall?

A rider is the obvious answer, and for a ridden bicycle it is the main one. But a riderless bicycle does it too, so the machine must contain a mechanism that converts a lean into a steer without anyone deciding anything.

The first candidate is the gyroscopic one. A leaning frame applies a torque to the spinning front wheel, and the wheel's response is at right angles to that torque — a steer toward the lean. Real, and pointing the right way; it is the same vector geometry as a spinning top's precession. But is it necessary? David Jones tested exactly that in 1970, fitting a bicycle with a second, counter-rotating wheel that cancelled the spin angular momentum. It remained rideable.

The second candidate is trail. The front wheel's contact patch sits some distance behind the point where the steering axis meets the ground, so the ground's push on the tyre acts on a lever arm and turns the fork like the caster on a trolley — again, toward the lean. Also real, and long treated as the true explanation once the gyroscope was demoted.

Then in 2011 Kooijman, Meijaard, Papadopoulos, Ruina and Schwab built a machine to break that story too. Their two-mass-skate bicycle had small counter-rotating wheels, cancelling the gyroscopic torque, and negative trail, reversing the caster effect. By the standard account it should have been unrideable. Rolled at the right speed, it self-stabilised anyway, recovering from a sideways knock as an ordinary bicycle does.

What was left to do the work? Mass distribution. If the front assembly is arranged so that, when the frame leans, the fork itself falls toward the inside of the lean, the steering turns for the same reason a hanging door swings — gravity acting on an off-axis mass. No spin, no caster.

So where does that leave us? Not with a fourth mechanism replacing the other three, but with self-stability as a property of the whole coupled system. The standard linearised bicycle model, benchmarked in 2007, takes about twenty-five geometric and mass parameters and yields modes of motion with names — weave, capsize, castering — and a bicycle is self-stable only in the band of speeds where all of them decay. Several different parameter combinations can open such a band, and no single one is required.

The gyroscopic claim is worth handling gently, because it is not simply false. The effect exists, points the right way, and contributes; it is sufficiency and necessity that fail. The collection's walk-through of gyroscopic precession sets out why the effect is real and why the bicycle is nonetheless a poor advertisement for it. What Jones and later Kooijman showed is that you can delete it and still have a self-stabilising bicycle — and, in the 2011 machine, delete trail as well.

It is equally worth resisting the tidy replacement. "It is really the trail" was the confident answer for forty years, and it did not survive contact with a bicycle built to test it. The honest statement today is that lean-to-steer coupling is the requirement, and that gyroscopic torque, trail, and front-assembly mass distribution are three routes to it a designer can trade against each other. Which dominates depends on the frame, and for a ridden bicycle at ordinary speed all of them are small beside the rider's own steering.

That is the uncomfortable part, and the reason this case is worth keeping. A bicycle is not exotic and has been ridden for a century and a half by hundreds of millions of people. It still has no single settled mechanism — only a set of contributing effects and a model that predicts the behaviour without singling out a cause. Familiarity is not the same as being explained.

• Why self-stability exists only between a lower and an upper speed, and what the weave and capsize modes actually look like.
• How countersteering — briefly steering away from the turn to initiate a lean — fits the same lean-to-steer picture from the rider's side.

Every term the collection defines is gathered in the glossary.