Built for the world as it is.

Humanoids trade speed for reach. Wheeled robots trade reach for speed. Scalar Robotics is building the robot that refuses the trade: wheel-legged, fast on open ground, agile on stairs and rough terrain, and engineered for deployment in the real world.

The machine’s parts come together on its true joint layout: wheels, legs and hip actuators, then the deck with its visor. Two amber eyes light up, and one learned policy takes it up a flight of stairs, onto a block taller than its wheels and up a ramp, replayed from physics simulation.

Legs alone are too slow for real work.

Humanoid robots move at a person’s pace and pay for every stride in energy. Deliveries, warehouses and campuses run on speed and range, and legs alone deliver neither.

Our robot rolls on wheels at the ends of its legs and balances on them: the speed and efficiency of a wheeled machine, with the reach of a legged one. In physics simulation it already holds itself upright on two wheels and recovers from a hard push.

In profile, the machine stands on two wheels. A ring of light marks a shove from behind; it rolls forward with the push, pitches, recovers and comes to rest, its poses sampled every tenth of a second.

We are building our robot to find its own way through the real world: along roads and sidewalks, around people and through the clutter of everyday places, with lidar and cameras to see what is around it.

Its first job is the hardest stretch in logistics, the last one: from the street, over the curb and up the steps to the door.

In physics simulation it already drives itself to a goal through a course of posts, a doorway and clutter.

From behind and then from above, the machine drives through a course of posts, a doorway and clutter to an amber goal, leaving a trail of its earlier poses that never touches an obstacle.

Where wheels stop, it climbs.

Curbs, stairs and broken ground end the route for wheeled robots. Ours lifts its legs, climbs and rolls on, and on rough ground its legs work as suspension.

One learned control policy climbs stairs, blocks and ramps, and every one of these skills is validated in physics simulation.

The machine drives at a short flight of stairs and climbs it one wheel at a time, the leading leg alternating; amber marks show each touchdown on the treads.

Founders

Scalar Robotics is led by two founders who have built both halves of a robot: the intelligence that drives it and the machine itself.

  • Founder, Machine Learning & Robot Policy Engine

    Leads the learning behind the robot’s control policy. Fifteen years in machine learning, with a background in model training and fine-tuning for Amazon Alexa.

  • Co-founder, robot hardware

    A mechanical engineer by degree, with hands-on experience building humanoid robots at Tesla and at Humanoid.

Our founders’ experience*

*These logos represent our founders’ past experience. They do not indicate any authorised collaboration with, or endorsement by, these companies.

Connect to know more.

We’re working with investors, partners and early customers who need robots that work in the world as it is, not in a lab.

Email usrobot@scalarrobotics.xyz

Or send us a message

Thank you. Your message is on its way, and we’ll reply by email.

We couldn’t send your message from the site. Please email it to us instead; we’ve written it out for you: robot@scalarrobotics.xyz

Every path the machine was measured driving in simulation, drawn from one starting point.

The machine stands at the centre of every path it was measured driving in simulation, each drawn from the same starting point and rising where it climbed.