locomotion | lsquarelabs

Here's what we do!

Our locomotion research explores the intersection of mechanical design, control theory, and machine learning. We're building legged robots from the ground up—designing custom mechanisms, developing simulation environments, and training reinforcement learning policies that transfer from simulation to reality.

Each project represents a different aspect of this journey: from single-leg prototypes that validate our kinematics, to fully indigenous quadrupeds built with open-source tools, to commercial platforms that let us test our AI controllers in the real world.

Our Walkers

Flea

Single Leg Mechanism

Rumi

Indigenous Quadruped

Lite3

RL Testing Platform

Flea

Single Leg Mechanism - A part of Rumi

Our leg mechanism was designed using CadQuery, a powerful open-source CAD framework that enables parametric 3D modeling through Python scripting.

We are currently 3D printing our prototype and are using Dynamixel MX-64 motors for actuation.

Photos

Videos

Sim-to-real test for single leg balancing

Rumi

Indigenous Quadruped Prototype

Our indigenous quadruped prototype, still under development. Rumi represents our commitment to building capable legged robots using open-source tools and locally sourced components.

Photos

Videos

Rumi flat terrain joystick control - forward motion

Rumi flat terrain joystick control - yaw rotation

Rumi quadruped standing up from prone position

DeepRobotics Lite3

RL Policy Testing Platform

The robot on which we test our reinforcement learning policies. The Lite3 serves as our validation platform for testing algorithms trained in simulation.