Hey friend
If there is one concept that separates robots that can walk with confidence from those that fall over every few steps it is the Zero Moment Point or ZMP.
I still remember watching early humanoid robots wobble and crash. Then I learned about ZMP. Suddenly it all made sense. ZMP is the most important idea in humanoid balance and stable walking. Today I will explain it in a way.
What Is the Zero Moment Point?
Imagine your robot standing or walking. Gravity is pulling it down. The floor is pushing back.
The Zero Moment Point (ZMP) is the point on the ground where the total moment around that point is zero. In terms it is the spot where the robot is “perfectly balanced” so it does not tip over.
Think of it like this:
When you stand normally the ZMP is in the middle of your feet, if you lean far forward the ZMP moves toward your toes, if it reaches the edge of your toes and keeps going you fall forward.
The golden rule of ZMP is simple:
For walking the ZMP must always stay inside the support polygon (the area under the feet that is touching the ground).
Why ZMP Matters Much for Humanoids
Robots cannot rely on instinct or reflexes like humans do. They need a way to plan every step so they do not fall.
This is where ZMP shines. Engineers use it to design walking trajectories. Smooth paths for the feet and body that keep the robot balanced.
Real-world examples:
- Tesla Optimus uses ZMP-based controllers to walk naturally while keeping the ZMP safely inside the foot area.
- Boston Dynamics Atlas constantly. Adjusts its ZMP during fast turns jumps and recoveries.
- Figure 01. Unitree G1 rely heavily on ZMP to generate stable walking patterns.
When the ZMP stays inside the support polygon the robot feels stable. When it approaches the edge the control system must adjust to bring it back.
The Simple Physics Behind ZMP
Here is the core idea:
The robot has a Center of Mass (CoM). As it walks the CoM moves forward. Side to side, the floor produces Ground Reaction Forces under the feet, the ZMP is the point where all these forces balance out.
A basic way to think about it: If the CoM is above the ZMP the robot is in perfect balance, if the CoM is in front of the ZMP the robot wants to fall, if the CoM is behind the ZMP the robot wants to fall.
During walking the ZMP moves from the heel toward the toes as the foot rolls forward.
How Robots Use ZMP, in Practice
Modern humanoid control systems do three things:
- Plan a desired ZMP trajectory that stays safely inside the feet.
- Calculate the required movements to achieve that trajectory.
- Use real-time feedback to correct any errors.
Some newer approaches combine ZMP with techniques allowing robots to recover from pushes or walk on uneven terrain.
My Personal Take
ZMP is not exciting. It is the foundation that makes everything else possible. Without an understanding of ZMP a robot will fall.
We have talked about Center of Mass, Base of Support toes and degrees of freedom. ZMP ties all of them together.
The coolest part? As computing power improves we are moving toward dynamic human-like balance.