Building a Quadruped Robot Over Three Years, One Failure at a Time

This robot dog wasn’t built overnight.

It took many years.
Multiple failures.
Broken parts.
Wasted money.
And some very uncomfortable lessons in robotics.

In 2020–21, I decided to build a quadruped robot, inspired by Boston Dynamics—but with almost no budget and no industrial support.

Just curiosity, ambition, and a lot of wrong assumptions.

🎥 Project video:
https://www.youtube.com/watch?v=JWOIpflocaI

Version 1 (2020–21): Enthusiasm Without Understanding

The first version was… bad.

A tiny prototype that:

• couldn’t balance
• couldn’t walk
• couldn’t even stand reliably

There was no stability.
No proper dynamics.
No real control logic.

Just frustration.

At that point, I made the hardest decision in the project:

I stopped.

Not because the idea was bad.
But because I wasn’t ready.

That pause mattered.

The Gap Years: Learning What I Didn’t Know

Instead of forcing progress, I did something more important.

I learned.

I spent time understanding:

• robot dynamics and kinematics
• balance and center of mass
• joint design and torque requirements
• control systems
• mechanical tolerances
• 3D printing for functional parts, not aesthetics

I also saved money—because robotics is expensive, and shortcuts show immediately.

Only after that did I come back.

Version 2 (2022): A Robot That Finally Looked Right

The second version was a major step forward.

A fully 3D-printed robot dog:

• clean design
• better proportions
• improved structural integrity
• visually, it finally looked like a real robot

But there was a hard truth waiting.

Looks don’t walk.

Despite the improved design, motion was still unreliable.
The robot existed—but it wasn’t alive yet.

That hurt more than Version 1.

Version 3 (2023): Where Things Finally Changed

This is where the project turned a corner.

I focused on fundamentals:

• improved joint design
• better weight distribution
• more realistic expectations from motors
• smarter control logic
• incremental testing instead of big leaps

And for the first time…

The robot walked.

Not perfectly.
Not smoothly.
But on its own.

That moment erased three years of doubt.

Is It Perfect? No.

Let’s be honest.

• The dynamics aren’t ideal
• The gait isn’t optimized
• The control system can be improved
• The hardware has limitations

But that’s not the point.

After years of iteration, robot “Perceptor” is alive.

And in robotics, that matters more than polish.

What This Project Was Really About

This was never about copying Boston Dynamics.

It was about:

• building under constraints
• failing publicly
• learning mechanical humility
• respecting physics
• and proving something important

Independent engineers can build real walking robots
—with patience, discipline, and iteration.

No shortcuts.
No viral tricks.
Just engineering.

What This Robot Taught Me

More than any other project, this one taught me:

• Robotics punishes arrogance
• Control systems demand respect
• Mechanics matter more than code
• Iteration beats intelligence
• Progress is nonlinear—and slow

Most importantly:

You don’t become a roboticist by building a robot.
You become one by staying when it doesn’t work.

Closing Thought

This robot didn’t just learn to walk.

It taught me how to.

Slowly.
Imperfectly.
But forward.

And that lesson will outlast any version of hardware.