Cable-Driven Neck Bracing Exosuit

Dropped head posture and poor head & neck motor control follow from ALS, muscular dystrophy, cerebral palsy, and stroke. The clinical default is a static collar, which holds the head upright but does not restore movement, and over extended wear brings pain, reduced range of motion, and difficulty swallowing and breathing (Bales & Zhang, 2024). Patients with cervical motor impairments need repeated head-neck training over long periods, not immobilization. The lab's platform delivers that training well: cables can only pull, never push, so a cable-driven robot needs at least one more cable than the degrees of freedom it controls, and the platform uses seven around the headpiece with the center of rotation taken at the base of the neck near the C7 vertebra.

The limitation is not the control. It is that the workspace is a chair with a rigid frame around it. Rehabilitation that has to happen in a lab session is rehabilitation that happens rarely. A wearable version could be used through the day, during ordinary activity, which is where the therapeutic value actually accumulates. The platform itself established that the assistance works: in a target-reaching study, neck muscle activation rose 19% when the robot resisted head movement and fell 28 to 43% when it assisted (Bales & Zhang, 2024).

A prior cable-driven exosuit out of the same research line proved the concept: three components, headband, vest, and belt. Against a state of the art rigid link neck exoskeleton it allowed more head rotation, a much larger absolute range of motion, and less compensatory torso movement across three everyday tasks, looking for traffic, drinking from a bottle, and picking an object up off the floor. Simulation showed it balancing the gravitational moment of the head throughout its range. Its failure modes are documented and specific, which makes them a design brief.

Raising an arm lifted the device off the wearer's shoulders and dislodged the Bowden tubes from their end stops. The headband concentrated clamping pressure on a small area, which becomes painful over the session lengths the device is meant for. The geometry only fits users between 163 and 188 cm. And it has no onboard kinematic tracking, relying on an external Vicon system, which ties it back to a instrumented room.

Seven cables is more hardware than a wearable can carry. I ran the layout optimization to see how far it could come down: a six-cable arrangement holds full 6-DoF control, and a five-cable arrangement works if an IMU on the headpiece supplies the missing state directly rather than inferring it from cable lengths.

The kinematic model takes a fixed center of rotation at the base of the neck, between the C7 and T1 vertebrae, which is the simplification the cable-driven parallel robot literature converges on for head & neck work. The method reads motor angles, converts them to cable lengths, and solves forward kinematics from the system geometry to localize the end effector. Given a target force-moment profile at the head, quadratic programming distributes it into individual cable tensions. Insertion point orientation and spool radius are the variables. There is a tradeoff to respect: optimizing purely for transmission ratio produces a robot that is harder to backdrive, which is the wrong direction for a device a patient wears and moves in.

The platform's motors are what make it stationary. The exosuit replaces them with passive actuators built around the idea of constant-force spring retractors, one under each joint of the link assembly. A constant-force spring holds tension nearly flat across its extension, so the assist does not fall off as the head moves through its range, and there is no power budget or motor mass to carry on the shoulders.

An encoder on the actuator preserves the state feedback the forward kinematics needs. Cable length is still measured, just not commanded.

The structure that replaces the extrusion frame is a mirrored pair of link assemblies, one per side, that sit on the shoulders and carry the cable exit points out to where the layout optimization wants them. Each side stacks two links through a pivot, with a plate at the top for the head to side termination and a housing at the base for the actuator and its encoder board.

The row of holes down the long link is what makes the geometry adjustable: exit point position can be moved along it between test sessions, so the optimized layout can be checked against a physically built one rather than only in simulation.

Everything sits on the shoulders, with adjustable straps so one device fits a range of body sizes instead of a 25 cm height window. The head interface spreads load over a larger surface area and conforms to the skull rather than clamping a band around it, since peak suit force is ultimately limited by what skin and underlying tissue can bear.

How much force the shoulders can take is a documented number rather than a guess: measured comfort limits average roughly 620 N pulling down on the shoulders, but vary from about 250 to 1200 N between individuals and climb 20% or more as a wearer habituates over days (Yandell et al., 2020). That spread is an argument for adjustability rather than a single tuned tension. Material choice follows the soft exosuit literature: maximize suit to human series stiffness to maximize power transfer, which argues for high stiffness webbing and for anchoring where there is bone close under the skin rather than soft tissue. A commercially available garment base is the starting point for the fabric layer, with printed rigid parts carrying the cable paths and actuator mounts.

The prototype is being built now. The test plan compares it directly against the chair mounted platform on range of motion and assist force, then moves to trials of everyday activities performed while wearing it, which is the case the platform cannot be measured on at all. It has since been picked up and continued by a PHD student.