Balance and Stability
A chair sitting still on a flat floor is stable. It will not fall over unless you push it hard. A pencil balanced on its tip is also still — momentarily — but touch it the slightest bit and it topples immediately. Both objects are following the same laws of physics, yet their stability is completely different. Robots face this same spectrum of stability, and the engineering choices a designer makes about balance determine what terrain the robot can cross, how fast it can move, and how complex its control system must be.
Center of Mass and the Support Polygon
Every object has a center of mass — the single point where you could balance the entire object on a pin. Gravity acts as if it pulls the entire mass of the object downward through that one point. For a stable resting object, the vertical line from the center of mass down to the ground — called the gravity line — must fall within the support polygon. The support polygon is the convex shape formed by connecting all the contact points between the robot and the ground. A four-legged robot with all four feet on the ground has a support polygon roughly the size of the rectangle between its feet. The robot is stable as long as its gravity line stays inside that polygon. If it leans so far that the gravity line exits the polygon, it will tip over. A six-legged insect robot has an even larger support polygon. A one-legged pogo robot has a support polygon that is just the tiny foot contact area — much easier to exit with any perturbation.
The support polygon is the area enclosed by the outermost ground contact points. If the vertical projection of the center of mass falls inside this polygon, the robot is statically stable. If it falls outside, the robot will topple in the direction of the exit.
Static Stability: Standing Still Safely
A robot is statically stable if it remains balanced even when completely stopped at any point during its motion. Six-legged robots (hexapods) achieve this by always keeping at least three feet on the ground in a triangle that contains the center of mass. Even if all motion freezes mid-step, the robot will not fall. This conservative approach simplifies control dramatically: the robot does not need to worry about falling, so it can focus on planning and sensing. The trade-off is speed — moving carefully with many feet on the ground at all times limits how fast you can walk. Many four-legged robots walking slowly are also statically stable — their feet form a polygon large enough that the center of mass stays inside during a careful walk gait. At higher speeds, four-legged robots necessarily enter dynamic stability territory.
Dynamic Stability: Controlled Falling
Human walking is not statically stable. At the moment one foot is in the air mid-stride, the gravity line of your body extends outside your tiny single-foot support polygon. You would topple — except that your other foot is already swinging forward to catch you. Walking is a continuous cycle of controlled falling: you lean forward past stability, then your foot arrives just in time to create a new support point and stop the fall. This is dynamic stability — maintaining balance through continuous motion rather than through a large, always-safe support polygon. It requires real-time sensing of body tilt and limb positions, and a control system fast enough to move a catching foot before the fall progresses too far. Bipedal robots that walk dynamically use Inertial Measurement Units (IMUs) — sensors measuring tilt, rotation, and acceleration at hundreds of Hz — and fast controllers to replicate this controlled-falling strategy. Boston Dynamics Atlas is one of the most advanced dynamically stable bipeds, capable of backflips, parkour, and recovering from large pushes.
An IMU combines accelerometers (measuring linear acceleration) and gyroscopes (measuring rotational rate) into one sensor package. It is the robot's inner ear — measuring how it is tilting and spinning so the balance controller can react.
Zero Moment Point: Engineering Dynamic Balance
Engineers use a concept called the Zero Moment Point (ZMP) to design and verify dynamic walking controllers. The ZMP is the point on the ground where, if you placed a pivot, the robot would be momentarily balanced without falling. If the ZMP is inside the support polygon, the robot is stable at that instant. If it exits the polygon, the robot will rotate about the edge of the polygon and fall. A walking controller for a biped tries to plan the robot's motion so that the ZMP stays inside the support polygon at every instant of the step cycle — even though the support polygon itself changes as feet lift and land. This is a challenging optimization problem that requires modeling the robot's mass distribution and predicting where the ZMP will be several steps ahead.
Match each balance concept to its correct description.
Terms
Definitions
Drag terms onto their definitions, or click a term then click a definition to match.
A hexapod robot always keeps three legs in a triangle on the ground before lifting any other leg. Which type of stability is this strategy designed to maintain?
During a single human walking step, the body briefly has its gravity line outside the single-foot support polygon. What prevents a fall?
Support Polygon Investigation
- Step 1: Stand on both feet, shoulder-width apart, arms at your sides. Draw (or imagine) the support polygon beneath you. Is your center of mass inside it? Try gently swaying — how far can you sway before you feel you must step?
- Step 2: Stand on one foot. Your support polygon just shrank to one foot. How much harder is it to stay inside that polygon with any arm movement or wind?
- Step 3: Get on your hands and knees (four contact points). Mark the four contact points and draw the rectangle they form. Which position — two feet, one foot, or four points — has the largest support polygon relative to your center of mass?
- Step 4: Observe a walking video (or just walk yourself slowly) and identify the exact moment when the gravity line exits the single-foot polygon during a step.
- Step 5: Write two sentences connecting your observations to why bipedal robots are harder to build stable than quadrupedal robots.