A seat that meets you at eye level
NUN
Summary
I turned the tension a cabin imposes into three adjustable quantities. Eye level, outward view and enclosure became software variables, and the eye-level axis was carried all the way to a physical mechanism that moves on the sitter's weight alone, with no motor.
- 3 × 3
- 17
- 0W
Overview
It began with a shared-mobility interior project run with Kia DEX. The goal was to remove three tensions the cabin imposes on people – relational, spatial, bodily. Whether the eye-level axis could actually be built was the question left open. NUN is what came from following that question to the end. It runs from concept to a test of feasibility.
Verification stops at simulation and folded paper. No physical mock-up was made and no user testing was run.
Problem
At level 4 and above the role of "driver" disappears, and in shared mobility riding with strangers is the default rather than the exception. Three tensions are then imposed on the rider.
- Relational tension – eye levels that don't match.
- Spatial tension – no choice of what to look at.
- Bodily tension – a body that doesn't fit the space.
Who you ride with, your body and your reason for travelling are all decided again on every trip. A fixed interior gives all of them the same state.
Key decisions
Don't handle the feeling – handle the three quantities that make it
You can't adjust 'comfort'. The three axes are what survived one test each: grounded in research on a psychological effect, continuously adjustable, and independent of one another. Without independence there is no state space.
Eye level · outward view · enclosure, across three states (intended tension, prospect, refuge). Each axis is defined as a software variable.
Propose and be accepted, never change by itself
A system that changes things on its own is another imposition. Accepting and declining is what personalisation can be built from.
Personal axes – your own eye level, the enclosure on your side – are controlled independently. Shared axes, like the door-side view and the common ceiling, move by request → the other person's acceptance → transition. Without agreement the axis follows whoever wants more privacy, and the requester is offered a correction on their personal axes only.
Separate what moves from what carries the load
A seat has to change height and hold a body. Moving means giving; holding means stiffening. One member doing both does both badly.
Four concepts (POP-LOCK, TENSION BLOOM, BREATH POD, TENSEGRITY WEAVE) compared on eight criteria; a frame-plus-tension-membrane hybrid won.
Weight instead of motors
The first build used motors and tendons. Control was precise, but in public, shared settings power, maintenance and failure break before precision does. And the two people who need to meet at eye level are already pressing the structure with different weights.
Zero motors, zero watts at design and simulation level – a one-degree-of-freedom rigid folding linkage with a bistable snap. At extreme weight ratios it cannot level fully, and that limit is recorded.
Level the eyes with nothing but sitting down – NUN
A bench that gives everyone the same height looks fair, but with different bodies the eye lines don't meet.
One side sinks and the other rises under the sitter's weight alone, with no controls. A self-locking origami structure unfolding from 40mm to 700mm, verified in folded paper (design value; not measured on a built seat), then carried from a park bench to an autonomous PBV interior.
Process
Research
- Desk research from papers – height difference creating hierarchy; adaptive seating raising flexibility while lowering privacy; prospect–refuge theory. Citations stay at author and year. No figures were copied across: the paper behind the enclosure range has no confirmed author or title, so its numbers are not on this page.
- Structural research – six origami patterns (Kresling, Yoshimura, Miura tube, bellows, waterbomb, Resch) compared on vertical travel, load and manufacturability. Kresling stayed a core candidate: compression couples to twist, and it is bistable.
The published work behind separating motion from load
Three papers were used. A Sarrus self-locking linkage (Kim et al., Science Robotics 2018) combines one-degree-of-freedom linear motion with locking at an orthogonal folding singularity. Dual-morphing origami (Kim et al., Science Robotics 2019) unfolds and then inflates in a quasi-sequential transformation, which is the basis for a wrapping motion. Variable stiffness (Park et al., IEEE RA-L 2023) changes stiffness through facet buckling.
One design principle came out of it: one scalar input plus local geometric programming gives distributed, sequential motion with no actuators. Instead of a motor per part, the structure itself carries the signal and the force.
The eight criteria the four concepts were compared on
Load bearing · fit to the concept · wrapping · flatness at rest · the choreography of deployment · ease of fabrication · scalability to production · safety while deploying. All eight sat in one table, and the frame-plus-tension-membrane hybrid won. No single concept won on every criterion.
Design
- Three layers of interface – choose a state in the app before boarding; adjust with three sliders and three presets in the cabin; then let posture and pressure sensors plus the city's context (entering a tunnel, a riverside stretch, traffic) propose a transition the rider accepts or declines.
- Six scenarios – mismatched eye levels, a body that doesn't fit, a conversation that matters, a picnic in the city, a space of one's own, shelter on a rainy day.
- Each scenario got its measures before anything else – subjective (a comfort scale, hierarchy-perception items, intent to ride again), behavioural (number of transitions, acceptance rate, time held), physiological (GSR, HR, HRV). Nothing was measured. The design of the measurement is what exists.
Development
The goal here was to make the digital variables and the physical mechanism point at the same thing. A preset is data, not hardware, and data grows with a software update.
The three axes as variables, and the state space
The three axes are software variables: eyeLevel.offset (continuous) and eyeLevel.matchMode (auto|manual|off); view.aperture (0–100) and view.mode; enclosure.ratio and enclosure.curl.
The cabin's state is a three-dimensional vector, {eyeLevel, viewAperture, enclosure}. All three axes are continuous, so the state space is infinite. Three points in it are meaningful enough to name as presets (intended tension, prospect, refuge). A preset is data, not hardware, so adding one is just a software update.
Five layers and an API sketch
L0 structure (origami skeleton and locking, hardware), L1 actuation (the weight-driven mechanism, hardware), L2 control (turning the three axes' targets into a trajectory and safety limits, software), L3 experience (presets, scenes and auto-suggestion rules, software, over the air), L4 service (subscriptions, personalisation and partner content, software). The line where hardware ends and software begins was drawn first.
POST /cabin/state sends target values for the three axes. POST /cabin/preset calls up a preset. GET /cabin/capability returns which axes and ranges a given vehicle supports. When POST /cabin/suggest makes a proposal, the rider's acceptance or refusal is logged as a cabin.suggestion.accepted or .rejected event, which becomes training data. It is a sketch, not a running server.
Prototype
- 17 physics simulators (output), in four strands – a weight-driven seesaw group of 3 (rigid link, rolling rocker, parallelogram), a pure-origami group of 5 (folded apex seesaw, Kresling twist, accordion, back-to-back wave, ground-supported), a pneumatic group of 6, and a mechanism-study group of 3 (lever, accordion, Sarrus linear motion).
- Design → verify → make – rigid panel chain simulation (PBD), a modified GPU origami simulator, a Python fold-pattern generator, SVG crease-pattern export.
- New ideas were stacked as new files rather than overwriting the old ones – a discarded concept stays as evidence, because the decision history has to be traceable.
Results
- 17
- 4
- 0W
The results are outputs and a validation design. There is no user outcome.
- Level of verification – simulation and folded paper. No physical mock-up was built and no user testing was run.
- The limit found – at extreme weight ratios, weight alone cannot level the eyes fully.
- What was not measured – comfort, perceived hierarchy, released tension. The measures are designed; there are no values.
Reflection
What held up
- A good answer came from re-reading a constraint, not adding a feature. Taking the motor out made the seat's problem look different: the two people who need to meet at eye level are already pressing the structure with different weights.
- Defining the three axes as software variables. Once a preset became data rather than hardware, adding an experience became an update.
- Proposing instead of acting. A space that rearranges itself is one more imposition, and the record of acceptance and refusal is material for personalisation.
What failed
- I assumed better control was the answer. Motors and tendons did give precise control. But in public, shared settings, power and maintenance break before precision does.
- I assumed weight-driven levelling would work for any pair of bodies. At extreme weight ratios it doesn't. Pretending to have solved something unsolved throws off everything designed after it.
- I assumed a paper's figure could be used directly as a design parameter. One of the references has no confirmed author or title, so its number was taken off this page.
What I would test next
- User validation – nothing was measured about whether meeting at eye level actually reduces perceived hierarchy. The designed measures need their first run.
- Physical limits – repeated-use durability, noise and vibration, temperature, crash safety. Getting past simulation needs a build.
- The working range of weight-driven levelling – how far the weight ratio can go before it fails, measured on a real seat. 0W is a number to confirm on a build too.
- Cost and regulation – going into a shared vehicle interior means passing production cost and safety regulation.