Design lab
Which concept actually dispenses one pill?
Every concept was run through the same six test protocols on the same physics engine, thousands of dispensing cycles with a ground-truth pill count from the simulator, so a miss, a double and a damaged pill are counted exactly, not inferred. All charts have a table view; tooltips show the counts behind every number.
1 · Single-pill success
Protocol P1: 8 real pills (4 mm mini-tablets to 19 mm capsules) × 12 bottle fills, a dose of 5 each. Single-pill success = metering cycles that delivered exactly one pill. Bars carry 95 % Wilson intervals. Dose exact is the stricter, patient-facing number: the whole dose arrived, nothing spilled, nothing broken.
2 · What each concept does with each pill
Dose-exact rate per real-world product. Hatched = not run. A concept that is excellent on one pill family and blind on another is a very different product from one that is merely average everywhere.
3 · Realistic design range
Protocol P2: the machine is auto-tuned to each pill and the pill is swept in size (25 sizes, four families). Cell = dose-exact rate. The table under the map states the size range each concept covers at ≥ 90 %.
4 · Does the adjustment mechanism work?
Protocol P3: one adjustable setting swept per concept, everything else auto. A good adjustable mechanism has a wide, findable good region, not a knife edge. The concept statement asks exactly this of the adjustable pocket.
5 · Robustness: last pills, motor limits, pill safety
P4 asks for a dose of 3 from a bottle that is 100 %, 50 % and 20 % full, at 20 % that is often the last few pills, tucked in the corners. P5 sweeps the motor's force / torque limit on a correctly tuned machine: too low and the motor stalls on ordinary friction (a “jam”). P6 is the stress test: the clearance is set 0.5 mm too tight with rigid lips, so pills really do get squeezed, the motor limit is now the only thing between a pill and the crush strength.
6 · Run it yourself
The same trial code runs in your browser. Pick a concept and a pill; each trial is a fresh bottle and a dose of 5. Trials run one after another (about a second each on a laptop), leave the tab in the foreground.
| # | Seed | Result | Collected | Cycles | Sequence | Jams | Damaged | Sim time (s) |
|---|
Sequence: o one pill · d multiple · m miss.
How we measure
- Engine
- Matter.js 2-D rigid-body physics with a replacement sequential-impulse contact solver (Coulomb friction, restitution). 2 ms step, mm–ms–g units so forces are in newtons. The physics check page validates it against closed-form solutions.
- Actuators
- Motors are dynamic bodies with a force / torque limit. A pill caught in the mechanism stalls the motor, that is the jam signal, and the squeeze it sees is the applied force, checked against the pill's crush strength.
- Pills
- Mass from geometry × density with lot-to-lot variation per seed; crush strength scales with size (tablet 90 N at 9.5 mm, capsule 24 N, softgel 32 N); impact damage from a drop-height rating.
- Ground truth
- Every pill is a tracked body. Cycle outcomes are counted from bodies crossing the exit, not from the simulated sensor. How well a real sensor could confirm them is the sensing lab.
- Seeds
- Each seed changes the pile in the bottle and every pill's mass. Same seed → identical run.