BDRLABVANCOUVER · 49.28°N
Brief
CONNECTED — §00 SHEET HEAD

We build
the whole
machine.

Mechanical design, fabrication, electronics, firmware, controls and perception, on one bench in one room. We take a hard machine from CAD to a demo that runs unattended.

Catch rate · n=250
84%
Loop latency
41ms
Top speed
2.4m/s
Brief → demo
31d
FIG. 0.1 — BDR-26-04-AREV △4
TRASHCAN-CVBDR-26-04
A two-wheeled bin that watches the room, fits a ballistic trajectory to anything thrown toward it, and drives to the landing point before the object gets there.
Read the sheet
§01 · ONE BUILDING

One bench, one schedule, no handoffs.

The four blocks below are one machine, and every one of them was designed, cut, populated and flashed in the same room. Nothing was quoted out and nothing waited on a vendor, so a change to the mechanism and the change to the loop that compensates for it happen the same afternoon.

01 / Chassis & mechanism
Designed in Fusion, cut on our own machines.
412 parts · 38 machined here · 9.6 kg
02 / Actuation & power
Two BLDC hub drives with field-oriented control.
14:1 · 20 kHz FOC · 6S · 22 min
03 / Sensing & perception
Two global-shutter cameras, hard-synced, and a single-class detector.
120 fps · 1440×1080 · 180 mm baseline
04 / Compute & control
Vision and the motion loop deliberately split across two processors.
Orin Nano + STM32G4 · CAN 1 Mbit · 41 ms
§02 · SELECTED WORK
FIG. 2.1 · BDR-26-04-AMechanical design, Fabrication and finishing, Electronics and PCB, Firmware, Controls and motion, Perception and learned models

TRASHCAN-CV

A trash can that catches what you throw at it. 84% catch rate, 41 ms loop latency, 2.4 m/s top speed — n = 250 throws, nine people, indoor, mixed lighting.

31 d buildRead the sheet
FIG. 2.2 · BDR-26-03-AMechanical design, Fabrication and finishing, Firmware, Controls and motion

ORIGAMI-LINE

A six-station folding cell for a paper part nobody could fold twice the same way. 4.20 s/part cycle, 96.4% yield, 6 stations — n = 2000 parts, inspected against a go/no-go gauge.

9 d buildRead the sheet
§03 · CADENCE

9 days from brief to a working machine.

There are no handoffs here, so there is nothing to wait for. The person who authored the CAD is the person soldering the board and the person tuning the loop, which removes the two places a hardware programme normally loses a month: the wait on a vendor quote, and the week spent explaining a mechanism to whoever inherits it.

Short schedules come from a short org chart, not from weekends. Every duration on this site is the elapsed calendar time between the brief and a demo somebody watched.

  1. Brief · 0 d
  2. First drive test · 4 d
  3. Detector at 120 fps · 11 d
  4. First catch · 16 d
  5. Lid v2, torsion spring · 23 d
  6. Unattended 100-throw run · 31 d
BDR-26-04 — milestones as recorded, day offsets from the brief
§06 · REVISIONS

We publish revisions, not posts.

The three most recent revisions to published sheets.
DATEREVSHEETCHANGEBY
2026-04-19△4BDR-26-04Reprinted the camera mast in PETG after two PLA masts crept in sunlight.RM
2026-04-08△3BDR-26-04Ballistic fit window 120 ms → 90 ms. Catch rate 79% → 84%.RM
2026-03-29△2BDR-26-04Moved the vision loop off the G4 onto the Orin. Steering wobble gone.DZ

Every revision Bench notes

§04 · COMMISSION

Send us the thing that is not working.

The mechanism that binds, the loop that will not close, the prototype that worked once in October.
Send us the problem