The first CrustBot now has a saved assembly package: integrated CAD, dimensioned workshop drawings, a four-page frame bench packet, eight P1S print projects and a staged assembly guide. The next physical step is to measure the stock, cut the deck and four ladder members, and dry-fit the unpowered frame. We have not recorded a completed frame or a powered field trial.
What this first vehicle includes
Remote driving on level pavement under direct operator supervision; the established Pi 5 / IMX219 perception stack; a powered load-bearing jaw; and a supported, removable bin. The frame uses plywood and timber with catalog drivetrain parts and selected printed mounts. It does not use a donor vehicle.
Active litter pickup and transfer into the bin are deferred. An opening jaw and a storage bin do not, by themselves, establish a complete trash-collection mechanism. Future upgrades have documented interfaces, without buying expansion hardware today.
The material estimate
The accepted planning material estimate is $604.29, including $11 in unresolved allowances: $2 for battery fittings, $6 for mouth fittings and $3 for the service pin. Existing materials are counted at their recorded cost. Freight, tax, tools and labor are separate; this is not a final checkout quote or a repeat-unit manufacturing cost.
The earlier $400 target and $500 ceiling remain historical comparisons. The current estimate was accepted for this initial assembly effort. The old aluminum re-buy is no longer the frame design we are preparing to build.
Checked on the desktop; still to check on the robot
The CAD package checks modeled fit and service positions. The software checks cover remote command validation, control timeouts and the read-only perception preview. Those checks do not demonstrate physical stopping, payload capacity or field detection accuracy.
The remaining work includes identifying the delivered battery’s exact terminals and electrical limits; fitting the actual hinges and actuator clevis; obtaining the service-pin quote; and commissioning the assembled wiring, controls and camera. Loaded travel, jaw holding, bin and battery changeouts, heat and endurance still need measured results.
Six useful hours per charge, 10 kg in a 30 L bin, a separate 0.5 kg jaw load and operation at a provisional 40°C ambient with sun are targets, not tested ratings. A battery swap or an idle soak does not prove six hours of useful operation. Robot field pickups remain 0.
Where simulation stands
The V1 simulation lab now uses the current CAD with articulated drive wheels, swivel casters and powered jaw. Newton1.7 / MuJoCo3.15 CPU runs provide recorded playback; official MuJoCo WASM supplies optional browser interaction. Empty,5kg and10kg bin cases explore assumptions.
Hurkamp Park and Fredericksburg Dog Park use georeferenced city imagery/GIS and coarse USGS elevation. Their surfaces and travel traces are map priors, not surveyed access or cleaned coverage. Registered site photos, measured scale and obstacle clearance remain open requirements.
Mass distribution, motor curves, traction and jaw holding remain uncalibrated. Synthetic camera views do not establish perception accuracy; estimated energy does not qualify endurance or heat. The sandbox control gate is separate from the complete Pico firmware. The older Rapier pickup demo remains historical. The assembly viewer still provides bin/battery service-position review.
Earlier perception benchmarks remain dated measurements of their named models and devices. They do not become qualification of this assembled vehicle when the CAD changes.