DIY Quadruped Robot Using Low-Cost Gimbal Motors: First 2-DOF Leg Test Documented

Building a Quadruped Robot from Scratch: A Low-Cost Approach

Among robotics hobbyist communities, quadruped robots (commonly known as "robot dogs") have always been one of the most challenging DIY projects. Boston Dynamics' Spot is genuinely impressive, but commercial products costing tens of thousands of dollars put them well out of reach for most independent developers. Recently, a Reddit user shared their progress building a robot dog from scratch — the standout feature being the use of low-cost gimbal motors as the drive system, along with a first test video of the leg structure.

What makes this project worth paying attention to is that it explores an alternative path that sidesteps expensive frameless torque motors. Traditional quadruped robots typically rely on costly direct-drive or quasi-direct-drive actuators, while gimbal motors — originally designed for camera stabilization gimbals — are inexpensive and compact. If a well-designed reduction system can boost their torque output sufficiently, it could significantly lower the cost of the entire machine.

Technical Breakdown: Gear Ratio and Compact Design

According to the author's description, the current leg under test uses a 2 degrees of freedom (2 DOF) structure — this is the foundational validation stage for a single leg's motion. The core design philosophy behind the entire actuator system revolves around two competing objectives: torque amplification and spatial compactness.

A ~14:1 Compound Reduction System

The inherent weakness of gimbal motors is their limited torque; direct drive alone cannot support the weight of the chassis. This makes the reduction mechanism the critical design challenge. The author uses a total reduction ratio of approximately 14:1, achieved not through a single reduction stage but through a combination of a planetary gearbox and belt reduction.

This compound approach has its clever aspects: a planetary gearbox can deliver a substantial reduction ratio with coaxial output in a very small axial footprint, while belt reduction allows the motor and joint output shaft to be offset from each other — effectively "hiding" the motor in a more favorable position and reducing the bulk and weight concentrated at the joint. Together, they provide enough reduction to amplify torque while keeping the overall leg structure compact.

Backdrivability is a critical metric in quadruped actuator design — it refers to whether an external force applied at the output can drive the motor in reverse. High backdrivability means the leg can passively comply with terrain like a spring upon contact, rather than resisting rigidly. This is essential for stability and safety during dynamic gaits. In general, higher reduction ratios lead to lower mechanical efficiency and poorer backdrivability. A 14:1 ratio sits right at the boundary between "acceptable" and "noticeably compromised." High-performance designs like the MIT Mini Cheetah tend to use reduction ratios below 6:1 paired with high-torque motors, specifically to preserve good backdrivability so that simple torque control algorithms can achieve compliant locomotion. By choosing a higher reduction ratio to compensate for the gimbal motor's limited torque, the gimbal motor approach represents a deliberate design trade-off — one that will require the control software to compensate for this physical constraint through techniques like impedance control or torque estimation.

Engineering Trade-offs Toward the Final Configuration

The author has stated a clear end goal: the complete robot will feature 12 degrees of freedom (12 DOF) — three per leg across four legs — with a total weight of approximately 3 kg.

A 3 kg target weight is fairly aggressive. For quadruped robots, a lighter chassis means lower torque demands on the actuators, and dynamic gaits (such as trotting or jumping) become easier to achieve. This means the author faces significant trade-offs among material selection, structural strength, and weight. The current single-leg 2 DOF test is fundamentally a validation of whether the gimbal motor plus compound reduction approach can deliver adequate responsiveness and torque, laying the groundwork for eventual full-body 12 DOF integration.

Why the Gimbal Motor Approach Deserves Attention

The significance of this project lies not in how impressive the finished product is — it's only a single leg test at this stage — but in what it represents: a "cost reduction" exploration trend within the open-source robotics community.

In recent years, projects like MIT Mini Cheetah and various open-source quadruped platforms have been pushing toward more accessible high-performance actuators. But genuine frameless torque motors with high-resolution encoders still carry a substantial price tag. Gimbal motors, as consumer-grade mass-produced products, can cost a fraction of what professional actuators do. If a reduction mechanism can compensate for the torque deficit, and if control precision and backdrivability issues can be adequately addressed, the entry barrier for building a robot dog could drop by another significant step.

Of course, this approach faces real challenges. Belt reduction introduces a degree of elasticity and backlash that affects force control precision; gimbal motors have inherent limitations in thermal dissipation and continuous torque capability; and a 14:1 reduction ratio, while boosting torque, comes at the cost of backdrivability — placing higher demands on the dynamic gait control algorithms. These are all problems the author will need to confront when advancing to full 12 DOF whole-body testing.

Gimbal motors are brushless motors originally designed for camera anti-shake stabilizers. They are characterized by a high pole count (typically 12 or more pole pairs), low winding resistance, slow rotational speed, and smooth response — making them well-suited for low-speed, precise position control. Compared to frameless torque motors designed specifically for robotics, gimbal motors have lower peak torque density and thermal capacity, but can be 5 to 10 times cheaper. Professional frameless torque motors — such as T-Motor's U-series or Antigravity series — typically cost hundreds of dollars each, while comparable gimbal motors can often be purchased for tens of dollars. This price gap is precisely what has motivated the exploration of gimbal motor-based low-cost approaches. It is also worth noting that FOC (Field-Oriented Control) ESCs for driving these motors have become extremely affordable thanks to the economies of scale driven by the drone industry, further reducing the overall procurement cost of the solution.

From a Single Leg to a Full Robot: How Far Is There to Go?

From a single leg with 2 DOF to a complete 12 DOF, 3 kg robot dog, there are still a substantial number of engineering challenges to overcome: multi-joint coordination control, gait planning, full-body dynamics, power and compute integration, and more. Passing the single-leg test is just the first step of a very long journey, but it validates the most critical hardware foundation — whether the actuator approach is viable at all.

For hobbyists looking to replicate this themselves, the project offers a valuable reference: consumer-grade gimbal motors combined with a planetary gearbox and belt compound reduction can indeed produce functional robot dog joints. As the author continues to post updates on higher DOF tests, the true potential and limitations of this low-cost approach will gradually become clear.

Note: This article is based on project descriptions and the first test video shared publicly by the Reddit author. The project is still in early development, and specific performance data remains to be verified through further testing.