Why give a robot legs at all?
Wheels are wonderful on smooth, continuous ground — a parking lot, a warehouse floor, a road. But the world is full of stairs, curbs, rubble, tree roots, and gaps, and a wheel needs an unbroken surface to roll along. The moment the ground disappears under it, a wheel is stuck. This is where legged locomotion earns its keep: a leg does not need continuous contact. It only needs a series of good places to put a foot.
Think of crossing a stream on stepping stones. A wheeled cart cannot do it, but you can — you pick discrete footholds and skip the water in between. That is the superpower of legs: they turn rough, broken terrain into a problem of choosing where to step. A legged robot can climb stairs, step over a log, and squeeze a foot into a narrow gap that no wheel could span.
Gaits: the rhythm of footfalls
A gait is the repeating pattern of how and when each foot lifts off and sets down — the choreography of walking. The same set of legs can move in several different patterns, just as a horse can walk, trot, canter, or gallop using the same four legs. Each pattern trades off speed, stability, and energy differently.
For a four-legged robot — the subject of quadrupedal locomotion — the classic slow gait is the walk: one foot moves at a time while the other three stay planted. Three feet on the ground form a wide, stable base, so a walk is cautious and steady. Speed up and the robot switches to a trot, moving diagonal pairs of legs together (front-left with rear-right, then front-right with rear-left). At full speed it may gallop, with moments where all four feet are off the ground at once.
A two-legged robot, the domain of bipedal walking, has a simpler-looking but harder cycle. Each leg alternates between a stance phase, where the foot is planted and carries the body's weight, and a swing phase, where the foot lifts and reaches forward for the next step. With only two feet, there is never more than one foot down during a normal step — and often, for an instant, none — which is exactly why two-legged balance is so demanding.
Two kinds of balance: pause-able and only-while-moving
The single most useful idea in legged robotics is the distinction between two kinds of balance, captured by static vs dynamic stability. The difference comes down to one test: if you froze the robot at any instant, would it stay up — or fall?
Static stability means you can pause at any moment and the robot stays standing. The rule is geometric: the center of mass — the average location of all the robot's weight — must stay directly above the support polygon, the shape you get by connecting the feet currently touching the ground. As long as the weight hangs over that footprint, gravity pulls the robot down onto its feet, not over an edge. A four-legged robot in a slow walk, keeping three feet down, is statically stable: it can stop mid-step and hold the pose, like someone carefully crossing ice.
Dynamic stability is balance that only exists in motion. Here the center of mass spends much of its time outside the support — the robot is, technically, always tipping over. It does not fall because it keeps catching itself: it throws a foot forward just in time to stop the fall, then tips again, catches again, step after step. This is exactly how a person jogs, or how you balance a broom on your palm by constantly nudging your hand. Stop moving and you drop. A trotting or galloping robot is dynamically stable — fast and fluid, but you cannot simply freeze it mid-stride.
Choosing a gait: matching rhythm to speed and ground
Gait choice is not arbitrary — it follows from what the robot needs right now. The two biggest factors are how fast it wants to go and how trustworthy the ground is. A robot picking its way across loose rocks wants a statically stable walk, keeping a wide base of feet down so that one bad foothold does not end in a fall. A robot sprinting across a flat field wants a dynamic trot or gallop, accepting the risk in exchange for speed and efficiency.
Real machines switch gaits on the fly, just as animals do — slowing into a careful walk where the footing is bad, breaking into a trot where it is good. Some designs even blend the two worlds entirely, putting wheels at the ends of legs in wheel-leg locomotion: they roll efficiently on smooth ground, then step like legs when the path breaks up. The legged-versus-wheeled choice from the start of this guide is not always either-or.
Underneath all of this sits a controller making the catch happen in time, on every step. Knowing where the weight is, where it is heading, and where to plant the next foot is the heart of dynamic balance — and the math that pins it down, including a precise stand-in for the support polygon while moving, is what we open up in the next guide.