Humanoid Neck Joint

Hover or tap labeled images for captions

The Creative Machines Lab is developing a robotic head, "Emo," with a uniquely expressive face. But every head needs a sturdy neck to stand on, and the existing design undermines the face's organic appearance with its blocky construction and jarring tendency to slump over when de-energized.

I overhauled the existing 3-degree of freedom (3 DOF) neck joint capable of tilt, shake, and nod, retrofitting it with an organic structure while designing a locking mechanism to support the head’s weight when dormant.

Group

Creative Machines Lab (Columbia University)

Role

Undergraduate Researcher

Skills

Data-Driven | Nonlinear FEA | Iterative Design

Year

2024

Survey of considered designs in CAD. L to R: original, cable driven, torsion spring, cam follower, primitive stopper, final stopper

Process

Nonlinear contact FEA of servo mount cradle. Servo hidden for clarity. Slight yielding at contact stress concentration expected (red), but acceptable converged FOS in surrounding structures

Tight space constraints and the head's high weight, paired with requirements for cheap mass manufacturability and aesthetic appeal, made the simple goal considerably more challenging. I cast a wide net, surveying ideas from the academic literature and initially attempting to adapt them to the existing joint. Torsion springs proved unwieldy. Cable-driven designs were overly complex. And the cam-follower design added too much height.

The solution? A ground-up rethink of the joint structure. I moved from a two-piece design to an organic unibody, creating a cavity in the middle. Here, a servo-driven stopper made of flexible TPU wedges between the tilt and nod joints, locking both in place with just one moving part. Stress hot spots in the servo cradle were identified with nonlinear contact FEA, and further validated with physical prototyping.

From here, I iterated on the stopper geometry, leveraging additive manufacturing to prototype. A data-driven test process backed each iteration, where I systematically measured each design's "breakaway torque" (i.e., the resistance to head topple). Aiming for maximum breakaway resistance while retaining enough compliance to be wedged into place with the available servo torque, I converged on a satisfactory design after four major design revisions and countless more incremental tweaks.

Top: evolution of linkage design (original -> organic unibody -> reinforced servo mount); Bottom: evolution of stopper design

Outcome

Rendering of final joint design on its base. The stopper servo is elegantly hidden away when viewed from the front

The result of all of that analysis and iteration? An elegant unibody joint morphology, firm enough to not just prevent head topple but to stay upright even when bumped. It's set-and-forget, too—once locked, all motors can be de-energized. And locking both joints is powered by just one discrete micro servo.

The compliance of the TPU has the added benefit of acting as a resettable mechanical fuse—if too much force is applied, the lock will give way before the joint or other expensive components are structurally harmed.

Renderings of final joint design (L) from the rear and (R) with linkage hidden

Other work