Adaptive Dispensing System
A parametric, gravity-fed dispenser designed to separate and count mixed products out of a bin. It is the case for dispensing over robotic picking, worked out in CAD.
- Role
- Solo design
- Year
- 2026 to present
- Tools
- SolidWorks
Nothing has been built. This is a design in progress and the page is a live log of it; geometry, prototypes and failures will land here as they happen rather than appearing finished after the fact.
October 2026
The system has a name now, and so does the company. Each individual unit is a Gravploy Cell, which is gravity plus deploy. A full installation of them is a Gravploy Hub. The company is Dropline.
The equation work is far enough along that I am into rough geometry: the first real shapes a cell can take, sized against actual product envelopes instead of placeholder numbers.
I am also looking for discovery calls. If you have worked in fulfillment operations, current or former Amazon especially, and you have been around dispensing or picking equipment in practice, I would like half an hour of your time. You can reach me through charlieduffell.com.
Product sits in a bin. The system singulates it, counts out a specified number of items, and releases them. The whole geometry is parametric, driven by global variables in SolidWorks, so adapting it to a different product envelope means changing a handful of numbers rather than remodeling the assembly.
The point is to do a job that fulfillment centers currently throw robotic arms at, using gravity and geometry instead.
Nothing physical has been made. The foundation is the parametric skeleton: the equations that hold every dimension in relationship to the others, so that when one product dimension changes the rest of the geometry follows without breaking.
I started there on purpose. It would be faster to shape a chute by eye until one product falls through it correctly, but that only ever works for that one product, and the next package shape sends you back to the beginning. The whole idea is that the geometry can be recalculated rather than redrawn, so the equations had to come first.
It came out of a piece I wrote comparing robotic-arm picking, A-frame dispensing, and cube storage. The argument was that dispensing scales more cheaply than picking, and that once you do that, replenishment becomes the next bottleneck. Designing this is how I am working through the first half of that argument.