Humanoid robots balance by constantly measuring how their bodies are tilted, where their weight is, and how the ground pushes back on their feet. This matters because a two-legged robot has a narrow base of support compared with a wheeled robot. If its center of mass moves too far outside its feet, gravity creates a tipping effect.
Robots such as Atlas, Honda ASIMO, and Tesla Optimus use fast sensing and control to stay upright while standing, walking, or recovering from a push.
A balance controller acts like a rapid feedback loop. Sensors such as an inertial measurement unit detect tilt and angular motion, while foot sensors estimate ground reaction forces. The robot then adjusts hip, knee, and ankle joint angles to move its center of mass back over its support area.
Many humanoid robots use zero moment point control, which keeps the effective tipping point inside the foot contact region.
Understanding How Humanoid Robots Balance
A robot does not get a single perfect reading of its posture. An accelerometer senses acceleration, including the effect of gravity. When the robot is nearly still, this gives a useful estimate of which way is down.
A gyroscope measures how fast the body is rotating, but small errors build up over time. Software combines both signals to make a steadier estimate. This is called sensor fusion.
Joint encoders add another important source of information. They tell the controller the angles of the hips, knees, ankles, and upper body. From those angles, the robot estimates the position and speed of each body part.
Walking makes balance harder because the support area changes every step. During double support, both feet touch the ground and the usable area includes the space between them. During single support, only one foot carries the body.
The controller must shift the body weight before lifting the other foot. A common simplified model treats the robot as an inverted pendulum. Its mass is represented near the body, above a pivot at the foot.
This model helps calculate how the body should move forward without falling. Real robots are more complex because their arms, torso, legs, batteries, and carried objects all move separately.
Zero moment point control is especially useful when the feet remain flat on the floor. The controller predicts where the combined effect of weight and motion reaches the ground. It then plans joint motions that keep this point within the contact region of the foot or feet.
The point does not have to sit at the exact middle of a foot. It may move toward the heel during braking or toward the toes during forward motion. If it reaches an edge, the foot may start to rotate.
The robot then needs another strategy, such as taking a step, lowering its body, swinging its arms, or using the ankle to recover. This is why push recovery is not just about being strong. It depends on recognising quickly whether ankle motion is enough or whether a new foot placement is needed.
Real floors create problems that simple models leave out. Carpet compresses, tiles can be slippery, and a small stone can change the angle of a foot. Force sensors in the feet help detect whether contact is firm, uneven, or missing.
Controllers must work within limits. Motors cannot change joint angles instantly, and a powerful correction can make the robot overcorrect in the other direction. Engineers tune the controller to be fast without becoming unstable.
Students can notice the same ideas in their own movement. When standing on one leg, people make tiny ankle corrections.
When a backpack shifts or someone gives a gentle push, people widen their stance or step. Those actions change the support area and redirect the body before a fall develops.
Key Facts
- Balance is stable when the center of mass projection stays inside the support polygon formed by the feet.
- Torque from gravity can cause tipping: τ = rF sin θ.
- An IMU measures motion using accelerometers and gyroscopes.
- Ground reaction force is the force from the floor that pushes upward on the robot's feet.
- Newton's second law guides motion control: F = ma.
- Zero moment point control tries to keep the net tipping moment at the ground equal to zero within the foot area.
Vocabulary
- Center of mass
- The average location of an object's mass, where gravity can be treated as acting on the whole object.
- Support polygon
- The area on the ground enclosed by the robot's feet or contact points.
- Inertial measurement unit
- A sensor package that measures acceleration and rotation to estimate tilt and motion.
- Ground reaction force
- The force that the ground applies to a robot's foot in response to the robot pushing on the ground.
- Zero moment point
- The point on the ground where the net tipping moment from gravity and motion is zero.
Common Mistakes to Avoid
- Thinking balance only depends on standing still is wrong because walking robots must also control acceleration, foot placement, and momentum.
- Confusing center of mass with the robot's geometric center is wrong because heavy parts like motors and batteries can shift the center of mass.
- Ignoring the feet is wrong because foot size, contact points, and ground reaction forces determine the support polygon and tipping risk.
- Assuming sensors fix balance by themselves is wrong because sensor data must be used by a controller to command joint torques and angle corrections.
Practice Questions
- 1 A robot has its center of mass 0.90 m above the ground. If its center of mass projection is 0.08 m from the edge of its foot support area, how much farther sideways can it shift before it reaches the tipping boundary?
- 2 A 60 kg humanoid robot stands still on both feet. What is the total upward ground reaction force if g = 9.8 m/s²? If the weight is shared equally, what force is on each foot?
- 3 A robot is pushed gently from the side and begins to lean left. Explain how its IMU, foot force sensors, and hip, knee, and ankle joints work together to prevent a fall.