LeKiwi Base: Delivery, Defects & Wheel Motor Setup

Unboxing and parts inventory, the Pi 5 power-board decision, three defects found and a formal complaint filed, and the combined arm+base motor ID assignment

Delivery

Base kit (Order #13152, ordered Aug 10, 2026) arrived Aug 31, 2026 — three weeks earlier than the original mid-September estimate.

Parts Inventory

• Base structure with 3× 85mm omni wheels and shafts, clear acrylic plate
• 3× 3215-C018 12V drive servos
• Servo driver board (pre-wired to the 3 drive servos)
• PLA robotic-arm base mounting structure ×1
• 2MP/30fps camera with 3m cable ×1
• 12V/5600mAh battery + charger ×1, 0.5m USB-C cable ×1, 3m USB-C cable ×1
• Assorted mounting screws/washers/standoffs
• Onboard buck converter, GYEC-DC0503 (12V→5V, 3A) — see power board decision below

Pi 5 Power Board — decision

The kit's included GYEC-DC0503 buck converter is rated 3A, below the Pi 5's recommended 5V/5A. Usable as a stopgap for light loads, but a proper replacement was sourced and ordered same-day. Full comparison of three candidate boards, and the reasoning behind the final pick, now lives in Hardware Notes — short version: a generic AliExpress PD module was rejected (7–12V input ceiling, too close to the battery's 12.6V full-charge voltage), a Treedix UPS HAT was rejected (rated for Pi Zero/2B/3B/3B+/4B only, not Pi 5), and the Yahboom Raspberry Pi 5 Power Expansion Board (6–24V input, 5V/5A PD output, DC5.5×2.1 input matching the battery's existing cable) was ordered from the Yahboom Official Store on AliExpress — €18 vs. €32 on Amazon.de for the identical board, same delivery window (8–15 Sept). Physically, the board's mounting posts won't co-stack with the existing metal Pi 5 case, so it'll be wired in loose rather than stacked for now — see Hardware Notes for the options considered there too.

Defects Found & Complaint Filed — Resolution

Three issues surfaced during initial inspection, beyond the power board:

1. Wrong buck converter — 3A shipped where 5A is recommended for the Pi 5. Not an assembly issue — the included GYEC-DC0503 is genuinely under the Pi 5's official 5V/5A spec. Resolved by ordering a correctly-rated replacement independently (Yahboom Power Expansion Board, €18); OpenELAB approved reimbursing this cost after review, receipt sent Sept [pending confirmation of clearing].
2. Arm mounting structure "doesn't fit" the base plate — resolved as an installation-method issue, not a hardware defect. OpenELAB's technical team identified that assembly should use the base plate's smaller hole pattern, not the larger one initially tried. Followed their linked tutorial and reassembled correctly — mount fits and feels solid. Worth noting: the kit shipped with no assembly documentation at all, which is very likely why the wrong hole pattern was tried in the first place — flagged to OpenELAB as feedback, separate from any specific request.
3. Mounting screws "undersized" — also resolved, same root cause. OpenELAB confirmed the kit includes M3×20, M3×12, and M4×10 screws, which were in the parts bags all along but not identified for this specific step. Once matched to the tutorial's instructions, sizing was correct.

Complaint filed with OpenELAB the same day parts arrived (Order #13152), with photos and a measured hand-sketch of the (initially, incorrectly-assumed) hole-pattern mismatch. OpenELAB's response cycle: acknowledged → technical review → tutorial + screw-size guidance → mount/screws confirmed resolved on retry → converter reimbursement approved pending receipt. Full complaint and reply threads not reproduced here; contact OpenELAB support directly (Order #13152) if the exchange itself is needed.

Outcome: all three issues resolved. Two turned out to be a documentation gap rather than a hardware defect; the buck converter was a genuine spec shortfall, now covered by reimbursement.

Wheel Motor ID Assignment

Rather than a separate ID-setting pass for the wheels alone, the combined lekiwi robot type was used, which addresses arm and base motors together in one command and one physical connection sequence — swapping the same Waveshare adapter previously used for my_kiwi_arm_1 across all 9 motors, one at a time:

lerobot-setup-motors \
    --robot.type=lekiwi \
    --robot.port=/dev/tty.usbmodem5B140288161

Connect the controller board to the 'arm_gripper' motor only and press enter.
'arm_gripper' motor id set to 6
Connect the controller board to the 'arm_wrist_roll' motor only and press enter.
'arm_wrist_roll' motor id set to 5
Connect the controller board to the 'arm_wrist_flex' motor only and press enter.
'arm_wrist_flex' motor id set to 4
Connect the controller board to the 'arm_elbow_flex' motor only and press enter.
'arm_elbow_flex' motor id set to 3
Connect the controller board to the 'arm_shoulder_lift' motor only and press enter.
'arm_shoulder_lift' motor id set to 2
Connect the controller board to the 'arm_shoulder_pan' motor only and press enter.
'arm_shoulder_pan' motor id set to 1
Connect the controller board to the 'base_right_wheel' motor only and press enter.
'base_right_wheel' motor id set to 9
Connect the controller board to the 'base_back_wheel' motor only and press enter.
'base_back_wheel' motor id set to 8
Connect the controller board to the 'base_left_wheel' motor only and press enter.
'base_left_wheel' motor id set to 7

Clean run, no errors. my_kiwi_arm_1 confirmed as the arm joining the base — my_kiwi_arm_2 stays reserved, untouched, for the future XLeRobot second-arm upgrade, per the original plan. Note the lekiwi robot type's motor naming differs from the standalone so101_follower naming used earlier (arm_shoulder_pan vs. plain shoulder_pan) — same physical motors, same IDs, different schema now that arm and base are addressed as one combined robot.

Scope note: this pass only wrote motor IDs via the standalone Waveshare adapter, motor-by-motor — same low-stakes, laptop-only process as the original arm setup, requiring no battery or Pi involvement. It does not constitute the physical daisy-chain wiring into the base's own onboard driver board; that step is still ahead, and currently blocked by the arm-mount defect above.

Current State & Next Steps

• ✅ All 9 motor IDs set (arm 1–6, wheels 7–9), my_kiwi_arm_1 confirmed as the LeKiwi arm
• ✅ Pi 5 power board sourced, compared, and ordered
• ✅ Complaint fully resolved — mount and screws fixed via correct installation method (OpenELAB tutorial); converter reimbursement (€18) approved, receipt sent
• ✅ New cameras arrived and installed — 2× on the SO-101 rig, 1× on the LeKiwi wrist. Noticeably better sensitivity/angle than the originals; initial informal tests promising. Reference photos of the pre-swap framing were taken beforehand per the earlier open item.
• ☐ Physical daisy-chain assembly (driver board → wheels → arm) — no longer blocked; mount now fits correctly
• ☐ Power board arrival (ETA 8–15 Sept) before full electrical assembly and power-up
• ☐ ACT validation pass on the LeKiwi arm remains the next real-data milestone once physically assembled
• ☐ Multi-task dataset planning for SmolVLA (task list, prompt wording, episode counts per task) — deliberately not rushed, to be worked out properly before the weekend recording session
• ☐ ML desktop UI/PyTorch install still finishing — longer than expected but not erroring, ordinary first-setup friction on a fresh Linux+GPU box

Continued in Technical Deep Dive: LeKiwi Arm Assembly & Checkpoint Transfer. Part of Stage 2 of the roadmap.