A motorcycle can already balance itself, but only inside a speed window: the standard bicycle model is self-stable between about 4.3 and 6.0 m/s, and below that it falls like an inverted pendulum. To balance forever, including at a standstill, a bike needs active control: steering motors that move the front contact patch under the falling mass, as Honda showed in 2017, or a control moment gyroscope. Holding a 250 kg bike against a 5° lean takes about 128 N·m, which a 10 kg flywheel spinning near 11,000 rpm can supply for a few seconds at a time.
Calculated in this post from standard gravity and these example inputs, which are my estimates and not measurements: a 250 kg bike and rider, a centre of mass 0.6 m high, a 5° lean, a gimbal rate of 1 rad/s and a 10 kg steel disc of 0.15 m radius.
- 4.3 to 6.0 m/s is the speed window in which the benchmark bicycle balances with no rider input (Meijaard, 2007).
- 0.25 s is the fall time constant of a stopped bike; a 1° lean becomes 10° in about 0.74 s.
- Gyroscopes are not required: a bike with no wheel spin and slightly negative trail was built and stayed upright above 2.3 m/s (Kooijman, 2011).
- 128 N·m holds a 250 kg bike at 5° of lean; the front wheel's own gyro gives a fraction of what is needed below 50 km/h.
Calculated in this post from standard gravity and these example inputs (250 kg, 0.6 m, 5°, 1 rad/s).
- Verdict: possible with changes. Forever-balance needs active steering or a gyroscope, and some cost in weight and rider skill.
This is a tier-2 what-if, so it gets worked properly: the real models of bicycle and motorcycle stability, the numbers for a 250 kg bike and rider, a sensitivity check and a verdict. It ends with what you lose if you never again have to hold a bike up.
On 5 January 2017, at CES in Las Vegas, Honda rolled out a motorcycle that stood up on its own with no rider, no kickstand and no hands. It called it the Honda Riding Assist. Its press release made a point of what the bike did not use: no gyroscopes, which it said would add a lot of weight and change how the bike feels to ride.
Instead, the bike borrowed control ideas from Honda's UNI-CUB personal mobility robot. A year later, Honda engineers explained the trick in a technical paper: at a standstill the fork geometry changes to give negative trail, and a motor makes tiny steering movements that keep the bike's weight over its tyres.
Honda Global: Honda Riding Assist motorcycle debuts at CES 2017What are the assumptions?
Two bodies of work anchor this post: the benchmark bicycle model and Honda's riding-assist paper. My own estimates are labelled.
| Symbol | Meaning | Value used |
|---|---|---|
| m | Bike plus rider | 250 kg (estimate) |
| h | Centre-of-mass height | 0.6 m (estimate) |
| g | Gravity | 9.81 m/s² |
| Iw | Front wheel and tyre spin inertia | 0.6 kg·m² (estimate) |
| r | Front wheel rolling radius | 0.3 m (estimate) |
| vw, vc | Weave and capsize speeds, benchmark bicycle | 4.3 and 6.0 m/s |
Why does a moving bike stay up at all?
Because it steers itself into its own fall. When a moving bike starts to lean right, its geometry and mass make the front wheel turn right, which drives the contact patches back under the centre of mass. Three effects create that steering: the trail of the front wheel, the gyroscopic torque of the spinning wheel, and the way the front assembly's mass sits ahead of the steering axis.
The standard description is the Whipple model: four rigid parts (rear frame with rider, front fork and handlebar, and two knife-edge wheels) joined by hinges. Meijaard and colleagues published the definitive linear version in 2007, checked by two independent hand derivations and two computer programs.
The paper gives the reference equations for an uncontrolled bicycle and a benchmark example with 25 geometry and mass parameters. Below about 0.5 m/s the bike falls like an inverted pendulum. At about 0.7 m/s an oscillating "weave" mode appears, which becomes stable at the weave speed of about 4.3 m/s. At about 6.0 m/s, the capsize speed, a slow lean-over mode turns mildly unstable. So this bike balances itself only between 4.3 and 6.0 m/s. The limit: it is linear, so it holds for small leans, and it ignores tyre slip and frame flex.
Read the paper (free)Is it the gyroscope that keeps a bike up?
No, not on its own, and sometimes not at all. The common belief is that spinning wheels act like gyroscopes and hold the bike upright. In 2011, a team from Delft, Twente, Wisconsin-Stout and Cornell built a bicycle that disproved it.
The team designed a "two-mass-skate" bicycle with tiny wheels, counter-spinning discs to cancel wheel angular momentum, and a slightly negative trail of −4 mm. By theory it should be self-stable above 2.3 m/s. Built and pushed, it rolled upright and recovered from sideways knocks, falling only as it slowed below about 2 m/s. The lesson: neither gyroscopic torque nor positive trail is necessary, and the mass of the front assembly matters just as much. The limit: it is a riderless bicycle on a smooth floor, and it showed a 20 Hz shimmy the simple model did not predict.
Read the paper (free)The authors add a point that matters for motorcycles: all known bicycle and motorcycle designs lose self-stability at high speed, because of the gyroscopic terms. At speed, the rider is always the final controller, even when it feels effortless.
How fast does a stopped bike fall?
Fast. A stopped bike is an inverted pendulum, and its lean grows exponentially. For small angles, gravity's tipping torque is \(mgh\theta\), and the bike's roll inertia is roughly \(mh^2\), so:
- Time constant. \( \tau = \sqrt{0.6/9.81} = 0.247 \) s.
- From 1° to 10°. \( \cosh(t/\tau) = 10 \) gives \( t/\tau = 2.99 \), so t = 0.74 s.
- Sensitivity. A tall adventure bike with h = 0.75 m has τ = 0.276 s and takes 0.83 s; a low cruiser at 0.5 m takes 0.68 s. Height buys you a little time, not much.
Calculated in this post from standard gravity and these example heights (0.75 m and 0.5 m), using t = 2.99 × √(h/g).
Three-quarters of a second is why riders put a foot down at every stop, and why low seat heights matter so much to shorter riders, as compared in the lowest seat-height bikes in India.

How big a gyroscope would hold it up?
A flywheel of about 10 kg spinning near 11,000 rpm. A control moment gyroscope (CMG) is a fast-spinning flywheel mounted in a frame that can be tilted. Tilting it at gimbal rate Ω produces a torque equal to its spin angular momentum H times Ω, from conservation of angular momentum.
Calculated in this post from standard gravity and these example inputs, which are my estimates and not measurements: a 250 kg bike and rider, a centre of mass 0.6 m high, a 5° lean, a gimbal rate of 1 rad/s and a 10 kg steel disc of 0.15 m radius.
Worked solution: sizing the flywheel
- Torque to hold 5° of lean. \( 250 \times 9.81 \times 0.6 \times \sin 5^\circ = 128 \) N·m.
- Angular momentum. With a gimbal rate of 1 rad/s (my choice), \( H = 128 \) kg·m²/s.
- Flywheel. A 10 kg steel disc of radius 0.15 m: \( I_f = \tfrac12 \times 10 \times 0.15^2 = 0.1125 \) kg·m².
Calculated in this post from the moment of inertia of a solid disc, ½mr², with these example inputs (10 kg, 0.15 m), and the torque worked out from standard gravity and the other example inputs above.
- Spin speed. \( \omega_f = 128 / 0.1125 = 1{,}140 \) rad/s, about 10,900 rpm.
Calculated in this post from the 128 kg·m²/s above, which rests on standard gravity and the example inputs, and the 0.1125 kg·m² inertia.
- Stored energy. \( \tfrac12 I_f \omega_f^2 = 73 \) kJ, about 20 Wh, roughly a large power bank.
There is a catch. A CMG only makes torque while its gimbal is moving, and the gimbal can only swing so far, perhaps 60° each way. At 1 rad/s that is about two seconds of full torque. A gyroscope cannot hold a bike leaning forever; it can only push the bike back upright and then rest. So a gyro-balanced bike must keep its mass centred, using the gyro to catch disturbances.
Why doesn't the front wheel's own gyroscope do the job?
Because at low speed it barely spins. The front wheel's angular momentum is \(H_w = I_w v / r\), which falls to almost nothing as you slow down, exactly when you need balance most.
See the data as a table
| Front wheel, H = I v / r (kg·m²/s) | CMG needed for 5° at 1 rad/s (kg·m²/s) | |
|---|---|---|
| 5 km/h | 2.8 | 128 |
| 10 km/h | 5.6 | 128 |
| 20 km/h | 11.1 | 128 |
| 30 km/h | 16.7 | 128 |
| 50 km/h | 27.8 | 128 |
| 80 km/h | 44.4 | 128 |
| 100 km/h | 55.6 | 128 |
Source: Calculated in this post with I = 0.6 kg·m², r = 0.3 m, m = 250 kg, h = 0.6 m · The wheel's gyroscopic torque mainly helps the bike steer into a lean; it was never strong enough to hold the bike up by itself.
Use this to check what you just read. Change the bike's mass, the lean you want to recover from and the flywheel, and see how fast it must spin.
Gyro-stabiliser sizing
How does Honda do it without a gyroscope?
By changing the fork geometry and steering under the fall. Trail is the distance between where the steering axis meets the road and where the tyre touches it. Normal bikes have positive trail, with the contact patch behind the axis, which helps them self-steer at speed. Honda's prototype switches to negative trail when stationary.
The authors modelled roll as a fixed mass above the centre of gravity and a movable mass below it, shifted by steering. With negative trail, steering moves both the low mass and the front contact point in directions that add up to a restoring roll moment, so steering alone can stand the bike up.
But negative trail destroys stability at medium and high speed, so the prototype switches back to positive trail above a set speed, with steering help fading out and a compensator to avoid fighting the rider. The limit: the abstract gives no switchover speed and it is a prototype, not a product.
Read the abstractOthers have taken different routes. Yamaha's AMSAS prototype on an R25 uses a six-axis IMU and two motors: one drives the front wheel at starts and stops, and a handlebar actuator takes over above roughly 5 km/h. In March 2026, OMOWAY said its OMO X, balanced by a control moment gyroscope, had entered mass production. The fork maths behind trail is in rake and trail explained.
| Approach | Example | Works at a standstill? | Main cost |
|---|---|---|---|
| Passive geometry | Every bike, benchmark model | No, only about 4.3–6.0 m/s | None |
| Steering motor plus negative trail | Honda Riding Assist (2017) | Yes | Actuators, steer-by-wire safety |
| Front-wheel and bar actuators | Yamaha AMSAS prototype | Yes, at walking pace | Actuators, software |
| Control moment gyroscope | OMOWAY OMO X, Lit Motors C1 | Yes | Flywheel mass, energy, gimbal limits |

What would riders lose?
The low-speed skill that makes riding feel like riding. Most of riding's craft lives below 20 km/h: slow U-turns, clutch control in traffic, feet-up balance at a signal. A bike that never falls takes that away, and with it some of the rider's awareness of the machine.
- Lost: the need for slow-speed balance, a skill that also trains clutch and throttle control.
- Lost: the feedback that tells you the bike is heavy, loaded or on a slope.
- Gained: no more dropped bikes in parking lots or on gravel stops, which matters for heavy adventure bikes and short riders.
- Unchanged: at speed, the rider still leans and steers by countersteering, because nothing above changes the high-speed physics.
A gyroscope cannot hold a bike leaning forever. It can only push it back upright and then rest.From the CMG working above
So, can a motorcycle balance itself forever?
- Passively: only between about 4.3 and 6.0 m/s for the benchmark design; outside that window something must steer.
- At a standstill: yes with active control, either Honda-style steering with negative trail or a gyroscope.
- With a gyro: about 128 N·m for 5° on a 250 kg bike, from a 10 kg flywheel near 10,900 rpm storing 73 kJ.
Calculated in this post from standard gravity and these example inputs (250 kg, 0.6 m, 5°, 1 rad/s).
- At speed: the rider remains the controller; no new hardware is needed.
No bike can balance itself forever passively: the benchmark model is stable only between about 4.3 and 6.0 m/s, and below 0.5 m/s it falls like an inverted pendulum, 1° to 10° in 0.74 s. With active control it becomes possible at every speed. Honda showed in 2017 that steering motors and negative trail can stand a bike up with no gyroscope.
The 4.3 to 6.0 m/s window is from Meijaard and colleagues (2007). The 0.74 s is calculated in this post from standard gravity and an example centre-of-mass height of 0.6 m, my estimate.
A gyroscope needs about 128 N·m for a 250 kg bike at 5°, from a 10 kg flywheel near 10,900 rpm, and can only push back for a couple of seconds before its gimbal must reset. Expect a weight cost of 10–20 kg (my estimate) and the loss of a skill riders spend years learning.
Calculated in this post from standard gravity and these example inputs, which are my estimates and not measurements: a 250 kg bike and rider, a centre of mass 0.6 m high, a 5° lean, a gimbal rate of 1 rad/s and a 10 kg steel disc of 0.15 m radius.
How the bike leans and turns at speed is explained in the physics of countersteering. Self-balance in one-sixth gravity, where it nearly happens at walking pace, is in what riding on the Moon would feel like, and the wider electronics picture is in the AI-native motorcycle of 2045. More engineering lives on Yantra.
What's new
What if next?
- What if a bike never needed fuel? A 20 Wh flywheel is fine. Moving the whole bike on sunlight is harder.
- What if an engine had no friction? Losses are why every machine eventually stops.
- How fast could a motorcycle go with no speed limit? Where the capsize mode and the wind both get serious.
Questions people ask
Why doesn't a moving bike fall over?
Because it steers itself into the direction it is falling, which moves the tyres back under the centre of mass. Trail, wheel gyroscopic effects and the front assembly's mass all contribute, and the benchmark model shows this works on its own between about 4.3 and 6.0 m/s.
Do gyroscopes keep a bicycle upright?
Not by themselves. Kooijman and colleagues built a bicycle with its wheel spin cancelled and slightly negative trail, and it still balanced above about 2.3 m/s. Gyroscopic effects help, but they are neither necessary nor sufficient.
Is there a self-balancing motorcycle you can buy?
OMOWAY said in March 2026 that its gyroscope-balanced OMO X electric motorcycle had entered mass production, with pre-orders planned in Indonesia. Honda and Yamaha have shown working prototypes but not production models.
How does Honda's self-balancing motorcycle work?
It changes the front fork to negative trail at a standstill and uses a steering motor to keep the bike's weight over the tyres, then returns to normal positive trail as speed rises. Honda said it deliberately avoided gyroscopes to save weight.
Sources
- Meijaard, J. P., Papadopoulos, J. M., Ruina, A. and Schwab, A. L.: Linearized dynamics equations for the balance and steer of a bicycle, Proc. R. Soc. A 463 (2007) (source for 0.5, 0.7, 4.3 and 6.0 m/s)
- Kooijman, J. D. G. et al.: A bicycle can be self-stable without gyroscopic or caster effects, Science 332 (2011) (source for −4 mm trail, 2.3 m/s, 20 Hz shimmy, high-speed loss of stability)
- Araki, M., Akimoto, K. and Takenaka, T.: Study of Riding Assist Control Enabling Self-standing in Stationary State, SAE 2018-01-0576 (2018) (source for negative trail and the switch to positive trail)
- Honda Global: Honda Riding Assist motorcycle at CES 2017 (2017) (source for 5 January 2017, no gyroscopes, UNI-CUB technology)
- Rushlane: Yamaha R25 self-balancing tech aims to reduce accidents (2023) (source for AMSAS, six-axis IMU, actuator handover near 5 km/h)
- Electrek: The world's first self-balancing electric motorcycle is now entering production (2026) (source for OMO X and its control moment gyroscope)
- Rushlane: BMW self-balancing enclosed bike patent (2026)
