Alpha
Rolie.

R12 version 5 brings the battery, chassis and lighting work into the completed saved assembly. The frame now fits the complete 6S6P pack, with braced cheeks, inward returns and a supported tray. Both fenders carry integrated front and rear COB cassettes with captured lenses.

Low outboard controller enclosures keep their sloped inner shoulders and removable caps. The armor follows the fender profile, while rounded hip transitions retain the existing chassis mounts. Four chin lamps now aim 30° downward, with outward yaw of 10° for the inner pair and 15° for the outer pair.

The front/rear strip targets are 100/75 mm at 45° downward aim; actual delivered tape still needs verification. Modeled robot mass is 16.95 kg, excluding optional sensor reserve and unmodeled main-body power-interface/distribution hardware. It is not a weighed result.

Board removal retains conditional engineering acceptance with named manual handling and numerical assumptions. Tested motor-plug paths collided, and straight inward removal of the wired inner cover failed. The accepted board-service sequence leaves that cover installed. The body PCB still has seven unfinished connections.

Watch the new three-view edit
Battery and receiving-chassis details
Lights, boards and open service work

Build records: completed October 9 at 12:04 pm Central; R12 v5 final manifest and handoff, actual native/archive verification, source-bound captures and owner recordings.

These new views come from the current fender CAD candidate: front and rear accent bars, a lower enclosure, and controller packaging beneath its cover. The images are source-bound work-in-progress renders. They show design intent; final native adoption, service access and physical qualification are still being checked.

The fender controller now has an R3 engineering handoff. All seven connectors use side entry, supporting my request for a lower enclosure. The 80 × 60 mm board outline, mounts and nonconnector geometry stay the same. Its recorded native electrical, routing and connectivity checks pass; the earlier fender remains preserved.

The mating housings and contacts are selected, and the input harness has a concrete material proposal: an uncompressed Ø8 mm sleeved bundle, R25 mm bends and Ø9 mm clear passages. Those are engineering reservations. Actual terminations, source protection, actuator pin mapping, complete CAN topology and finished dimensions still need qualification.

I resumed the main CAD work after the travel checkpoint. The receiving chassis and low-profile fenders are being integrated. PCB-removal analysis has progressed within stated numerical and handling assumptions; tested motor-plug and wired-cover removal paths collided and remain unqualified for physical service.

Explore the R3 fender and harness work
Follow receiving-chassis integration

Build records: October 9 immutable R3 fender engineering handoff; later harness and transition research; explicit builder continuation and active CAD progress.

The battery is now a completed isolated mechanical design: 36 owned EVE cells in a 6S6P pack, with the owned 40 A JK BMS. Its core is 146 mm wide, with a documented interface for the chassis. Source checks and a fresh rebuild are recorded; physical fit, welds, protection and thermal/current behavior remain ahead.

The dedicated PCBA work is developing the body and fender controllers in native KiCad. The fender board now has a checked static-design handoff and mechanical reference. The body board stopped at the overnight cutoff with seven connections remaining; factory DFM and physical acceptance are still ahead. The new board outlines require their own mechanical accommodation.

The receiving-chassis and body work later stopped at a preserved incomplete travel checkpoint. R12 version 4 remains the last completed whole-robot CAD save.

Explore Battery & Power
Explore Lighting & PCBA

Build records: isolated battery handoff and delivery verification; dedicated electronics work order and October 8 status snapshot; active receiving-chassis progress.

I pushed for 25 mm of nose shortening, closer inner covers, and protected cable passages through the tibia and its matching rear adapter. The finished design brings those changes into the connected assembly while keeping the original orange motor wiring.

Looking closely at the bearing support exposed a gap. We worked through recessed outside-in screws, pressed threaded inserts in the inner retainer, and continuous printed bosses. The removable electronics lids and raised outer panels now carry the dark gray accent; the outer panel also provides access to the local lamp wires.

The compact design was delivered as R12 version 3. After recording the can-reference pose, I asked Codex to return Rolie to normal standing height. Version 4 saves that pose with the body level, both tires on the ground, and the can hidden. It changes the presentation, with the compact geometry retained.

Watch the two-pose edit below →

Build records: October 7 completed compact-design handoff, native/archive/cloud and selected-mesh receipts, plus the subsequent completed standing-pose save. One cable’s Fusion volume was unavailable; its complete native geometry comparison passed.

Reviewing the first R12 sent the design back through another round. I wanted the sculpted shroud and vertical lamps, a straight leading lip with angled corners, and an inner mounting cover that conceals the cable entry. Version 2 brings that combination into the same Fusion document.

The fenders now have flush controller lids, reversible mounts to the existing tibia, and vertical lamp pairs on the outside. The original wheel and tire geometry, four body lamps, finishes, and can stance are retained.

I also asked for the long added red lighting routes to go. They are removed from the legs and chassis, including hidden pose alternatives. Each fender now has a short local feed from its unused stationary wheel-actuator socket, carrying two power wires and two CAN wires, followed by local lamp leads. The original orange motor wiring stays intact.

Build records: completed R12 sculpted-fender/local-lighting report and hash-bound evidence. This is CAD delivery for builder review; local archive reopening and cloud completion do not include an independent cloud download/reopen.

Before heading to bed, I asked Codex to keep working on the same revision. The brief grew into compact faceted fenders, four lights below the grille, two in each fender, and a more believable can stance. The completed R12 brings those additions together in the native Fusion assembly.

The printed fenders have flat roofs, angled faces, and narrow outer supports around the inherited bearing envelopes. The original tire and wheel geometry stays intact. The chin lights follow the requested crossed layout: outer lamps aim inward and slightly down, while the inner pair aims down and outward. All eight lenses have a pure-white emissive CAD appearance.

The selected Sensor brackets eyes sit in a full charcoal brow and camera surround, matched to the grille. Twenty-two modeled aluminum parts have a matte-gray visual finish. The saved can pose uses a 10° body bank and 13° outward lean at the raised hip, applied through the connected assembly.

Three compact controller cases and dedicated lighting power/CAN routes are integrated, with localized lead passages in the removable intake pan. The electronics remain a design concept: actual LED and bearing parts, circuit design, fuse takeoff, connector ratings, and CAN interfaces still need qualification.

Build records: completed R12 delivery report and machine-readable evidence; native receiving, archive, cloud, combined offline audit, and mass-study receipts. The local archive was reopened; independent cloud download/reopen is outside the reported scope.

R11 turns my wiring feedback into matching open guards on the femur and tibia. Small local zip-tie slots let the cable lay into the guards, with no printed tube to thread it through. Compact knee rails bolt directly to the actuator holes, with close plug hoods replacing the projecting carriers and spacers.

Two recessed red displays flank the preserved camera. Their delivered static CAD appearance uses mirrored, segmented black visor graphics. A detailed 12-ounce Diet Coke can gives the assembly a familiar size reference: one tire meets the rim, while the other stays on the ground. The pose uses the connected joints.

The next direction: I chose the open, level “Sensor brackets” eye shapes and asked Codex to aim close to normal ride height in the can pose. The raised leg can lean outward a little if needed. That direction led into R12; the images here document the delivered R11 state.

Build records: R11 final delivery and verification report, completed October 6; subsequent builder feedback. The can is a CAD presentation prop. Pose-specific cable displays are static and do not deform automatically.

After the R9 cleanup, I wanted the cable supports to have real attachments and removable covers. The femur and tibia now use bolted, openable carriers. Knee half covers protect the plugs on the stationary motor housing while leaving the moving cable spans free.

The wheel conductors follow the actual metal bore and intact bushing. The planned assembly order feeds the conductors through first, closes the channels, and then installs the terminated connector. Orange insulation makes the installed neutral cable routes easy to follow.

The centered enclosed OAK-D Pro W and restrained front/side details are included as visual previews. Camera attachment, heat transfer, and optical field of view still need review and qualification.

Physical material/process strength, fit, practical tooling, wire life, electrical integration, cooling, and commissioning remain open. The six flexible cable spans are pose-specific CAD displays.

Build records: R10 final delivery and verification report. Completed October 6, 5:46 pm Eastern.

My review was specific: bulky stop hardware and exposed wire loops distracted from the robot and made the design harder to build. R9 brings six stop assemblies into the joints and organizes routing at 14 connector positions. Downward louvers screen the fan while keeping open angled channels.

The handoff checked 30 native motion workflows, the installed geometry of 286 selected parts, and a reopened editable Fusion archive. Cloud saving finished separately. That gives us a durable CAD baseline to review and build from.

Build record: R9 completed CAD revision and verification. Completed late October 5, Eastern time.

R7 brings the battery holders, structural frame, shell, stops, and cable routes into one CAD review package. Battery service space and fit coupons are documented alongside the actual components. Joint travel gets explicit limits.

Wires have to survive articulation, so selected poses are checked with matching cable geometry. The CAD cable display is pose-specific; it does not automatically flex through every manual joint edit.

R8 is a saved intermediate checkpoint for smaller stops and improved routing. It retains unresolved connector fan-outs and pigtail approaches. Those open items lead into R9 rather than being called a finished release.

Build records: R7 delivery package; R8 resume checkpoint.

R6 integrates the face and compact wire exits. A travel checkpoint records which changed-part checks finished, which cable sweeps are partial, and what has actually been saved. That makes the next session a continuation of the work.

A parallel SteveCAD experiment carries an R6 source representation into another workflow. Linked part builds and an actual grille edit-and-restore work, but full-assembly generation reaches the platform's 300-second budget. The project remains in Fusion while that limitation is documented.

Build records: R6 travel checkpoint; October 5 CAD workflow migration report.

Bulky tie blocks give way to formed sheet parts. Leg plates become thicker and wider after load and deflection screening. Battery restraints, intake, fan transition, internal baffle, and rear exhaust become part of the same assembly.

Laser cutting, press-brake forming, machining, and engineering-filament printing shape the choices. The shop's tooling and bend information are inputs to the design, along with the robot's appearance.

Flat-pattern and print-review artifacts are prepared, while bend coupons, real airflow, temperatures, and assembled fit remain future checks.

Build records: R5 refinement handoff; load and thermal reviews.

A continuous quarter-inch aluminum web, deeper cheeks, and ribbed connections tie the hip supports into a shared chassis. The change addresses the load path instead of only improving the exterior.

It also adds weight. Codex compares estimated mass, joint loading, and idealized stresses so the stronger-looking structure can be assessed against its cost to the rest of the robot. Physical strength qualification remains ahead.

Build records: R4 backbone handoff and engineering review.

The leg baseline changes to six Damiao DM-J6248P-2EC actuators. Two 24 V DM-J4310P-2EC drives remain at the wheels. Knee outputs face inward so the upper and lower legs can pass during a deep crouch.

The selection follows research into holding duty, load-dependent speed, motor mass, interfaces, and available packaging. The earlier Unitree and larger Damiao combinations stay in the record as alternatives considered.

A local supported MuJoCo demonstration makes speed saturation visible at normal playback. A deliberately excessive wheel request reaches about 183 rpm; an excessive knee request reaches about 57 rpm under the model's declared limits. The yellow support guide constrains body roll and pitch, and self-collision is disabled for this isolated study.

Build records: R3 overnight review; final actuator-speed readout.

I asked for slimmer legs and a compact form closer to the original Rolie. Codex revised the geometry toward formed 5052-H32 aluminum plates, shifted the hips rearward, and lowered the neutral hip height to 320 mm.

Appearance feedback becomes mechanical direction. The revised layout is checked with actual joint ownership and selected clearance poses, keeping the original source intact.

Build record: R2 final CAD verification.

The starting point is an unfinished SolidWorks wheeled-biped concept. Its familiar body and the builder's rover wheels guide the new assembly. The intended behaviors include smooth crouched driving, wheel-held stepping, and eventually curbs and small stairs.

R1 uses six Unitree GO-M8010-6 leg motors and two Damiao wheel drives. Eight native revolute joints make the mechanism inspectable. Static wheel raises and sampled poses check the layout; later actuator research changes the leg baseline.

Build records: mini-TRON1 development plan; September 27 overnight CAD handoff.