Skip to content

armgit / moss

Public Imported from GitHub
0

MOSS: a small, mostly 3D-printed litter-picking rover. Cleaner streets, one maker at a time.

Python

docs: track V0.4 hull, ordered control electronics and CAD tour

metroxcd4d565ee0bc
diff --git a/README.md b/README.md
index aa872e8..2a689d4 100644
--- a/README.md
+++ b/README.md
@@ -4,9 +4,9 @@ Cleaner streets, one maker at a time.
MOSS is a small litter-picking rover you can print and build yourself: a tracked base, a bin, and an arm derived from the SO-101 with a parallel gripper. About $1,700 in parts as built, with a Jetson and a RealSense; a Raspberry Pi version around $820 is the next experiment. It is being built in public, from a living room in Hua Hin, Thailand. Nothing here is finished; everything here is real.
-![MOSS exploded view, turntable](media/images/moss_exploded_turntable.gif)
+![MOSS hardware tour: inside the hull, back together, a scripted pickup](media/images/moss_v04_hardware_tour.gif)
-![MOSS picking up a bottle and dropping it in the bin](media/images/moss_pickup_loop.gif)
+*40-second CAD animation of the V0.4 design: inside the hull, reassembly, and a scripted pickup. The physical prototype drives; autonomous pickup is still ahead. [Video version](media/video/MOSS_V04_Inside_and_Action_20260924.mp4).*
## Where it stands
@@ -18,6 +18,8 @@ All 113 printed parts are on the bench, with the shafts, bearings and screws to
![MOSS with its bin on, 22 September](media/images/build_bin_on_20260922.jpg)
+In parallel, **V0.4** is being drawn from the assembly feedback: a revised hull, screw-terminal control electronics, one dual motor driver instead of two. Not built yet. See the [V0.4 hardware update](docs/hardware_v04.md).
+
Every step is written down when it is real, good or bad. The [build log](docs/build_log.md) has the dates and the numbers, the [assembly feedback](docs/assembly_feedback.md) has what did not fit, the [roadmap](docs/roadmap.md) has what comes next, without dates.
## What does it cost?
@@ -46,7 +48,7 @@ Jev, TypeSafe AI's decision model, choosing every step of a pickup (approach, al
## Build one, or come and watch
-The design files are not out yet. They come once the first MOSS rolls and the parts have earned it. Until then, the door is the Discord: ask anything, tell us what we are doing wrong, it helps the next builder too. https://discord.com/invite/x2MeNmgveT
+The design files are not out yet. The first MOSS now rolls; we are incorporating its assembly feedback into V0.4 before releasing the parts and their matching BOM. Until then, the door is the Discord: ask anything, tell us what we are doing wrong, it helps the next builder too. https://discord.com/invite/x2MeNmgveT
The build, day by day, is on X: [@metrox_eth](https://x.com/metrox_eth).
diff --git a/docs/assembly_feedback.md b/docs/assembly_feedback.md
index 6525821..18ba014 100644
--- a/docs/assembly_feedback.md
+++ b/docs/assembly_feedback.md
@@ -1,12 +1,12 @@
# Assembly feedback — V0.3 prototype
-One line per finding: what was observed, which part and revision, what was done on the prototype, what is planned for the next revision, and what validation actually happened. Nothing here is a completed drive test.
+One line per finding: what was observed, which part and revision, what was done on the prototype, what is planned for the next revision, and what validation actually happened. The first physical drive happened on 22 September; each row below distinguishes that observation from unvalidated V0.4 changes.
| Finding | Part / revision | On the prototype | Next revision | Validation done |
|---|---|---|---|---|
| TPU track too stiff | Track printed with a 0.8 mm nozzle | Not used; the 0.4 mm profile (nozzle, extrusion width, walls, layer height, infill, material) is the reference | Record the full print profile with the part | 0.4 mm track: manual floor test, good flexibility and grip |
| Motor mounts | Corrected brackets Ø31 / M3 (files marked V2-01) | Check screw depth at assembly | — | Not yet |
-| Motor mount to hull joint | Through-hole planned on the prototype | Drill through | Real through-joint designed in | Not yet |
+| Motor mount to hull joint | Interior access is difficult | Head recesses / through access drilled on the prototype | Exterior-access recessed M4 heads and captive nuts in the motor support; standard M4 × 20 in the CAD study | Local CAD checks; complete printed joint and tool access still to validate |
| Front of the cover and offset arm plate | Cover, arm plate | Keep the unfinished interface as is | Complete-cover study, parked | — |
| Camera brackets | Two RealSense D455 brackets | Fit at assembly | — | Not yet |
| Plate gasket | R8 gasket listed with the R11 plate | Check the real fit at assembly | — | Not yet |
@@ -15,3 +15,21 @@ One line per finding: what was observed, which part and revision, what was done
| INA219 cannot sit in the motor line | INA219 module terminal vs main cable | Voltage monitoring only; current path bypasses the shunt | Proper current sensing if wanted | I²C and voltage read confirmed; current readings invalid |
Available M4 heat-set inserts are an option for the next revision, not a validated feature of this prototype.
+
+## V0.4 follow-up — 24 September
+
+| Finding / request | V0.4 response | Status |
+|---|---|---|
+| Jetson caddy and cables do not fit | Integrated seat for the original plastic base, retainers and USB-side cable route | Current hull CAD; printed fit pending |
+| Individual Dupont connections are inconvenient | QT Py ESP32-S3 with screw-terminal BFF and STEMMA QT INA connection | Two sets ordered; wiring and firmware migration pending |
+| Two MD10Cs take more space than needed | One Cytron MDD3A dual driver | Ordered earlier, selected for CAD; rover validation pending |
+| RealSense straight USB plug lacks clearance | Ordered right-angle USB3 cable; housing reserve | Cable dimensions and USB3 stability untested |
+| INA connector access | Existing module model and fixings rotated 90 degrees | Approximate existing-board CAD; new Adafruit footprint not yet reconciled |
+| Battery installation under arm | Current pack 67 × 74 × 27 mm; larger 75 × 82 × 32 mm candidate envelope | Envelope only; detachable lower platform deferred |
+| Bin retention | Four 40 × 10 × 5 mm magnets | Ordered; mounting design pending |
+| Wheel / roller screws protrude | Recess the screw heads | Requested; not claimed implemented |
+| Hard-to-source screw lengths | Standard readily available lengths, without custom cutting | Design rule; full fastener audit pending |
+| Tracks may be too long | Test maximum adjustment and candidate one-tooth shortening | Hypothesis; no validated new belt length |
+| Separate fuse mount takes space | Inline fuse holder on a cable | Selected arrangement; final full-robot rating not established |
+
+Keep the corrected flank geometry. Do not move the fourth fixing solely because of the initial wrong-rail tensioner placement. See [the complete V0.4 status](hardware_v04.md).
diff --git a/docs/bom-costs.md b/docs/bom-costs.md
index 64f9adb..ae4b830 100644
--- a/docs/bom-costs.md
+++ b/docs/bom-costs.md
@@ -112,3 +112,9 @@ Accessed for the 21 September comparison; prices and availability vary by locati
## Media
[45-second video](../media/video/MOSS_cost_comparison_20260921.mp4) · [Poster](../media/images/MOSS_cost_comparison_20260921.jpg) · [Media credits](../media/MOSS_cost_comparison_20260921_credits.md)
+
+## 24 September update — V0.4 control components
+
+The [ordered V0.4 control set](hardware_v04.md#control-electronics-ordered) is **$39.60 per rover**, including its optional snap-on enclosure, or **$34.65** without that enclosure. Two sets cost $79.20 plus $21.83 shipping; import charges are not included. The Cytron MDD3A is selected to replace the two MD10Cs; its purchase price is not supplied in this update.
+
+The three totals above remain the **dated 21 September comparison**. Do not add the whole new order on top of a one-robot BOM: it contains two sets and replaces existing controller/INA lines. Recalculate a separate V0.4 total once all replacement lines, including the driver, are priced and reconciled.
diff --git a/docs/bom.md b/docs/bom.md
index 21bfb3c..4cbac72 100644
--- a/docs/bom.md
+++ b/docs/bom.md
@@ -44,6 +44,12 @@ Not printed: metal HEX12 hubs, bearings, shaft collars, motor–shaft couplings,
Cabling: same wire for the main runs and the driver branches, runs of 25–30 cm at most. Conductor bundle measured 1.2 mm in diameter; section estimated at 0.75–1 mm², not manufacturer-verified.
+## V0.4 selection (next revision, not built)
+
+The next revision selects a QT Py ESP32-S3 (5426), screw-terminal BFF (6495), optional enclosure (6505), INA219 STEMMA QT (904) and 300 mm cable (5384). One set is **$39.60** including the enclosure, before shipping and import charges. A Cytron MDD3A is selected to replace the two MD10Cs. The Waveshare Bus Servo Adapter remains.
+
+See [the V0.4 update](hardware_v04.md) for quantities, unit prices, CAD status and remaining fit checks. This does not overwrite the as-built V0.3 inventory above. The new Adafruit INA footprint still needs reconciliation with the current hull model.
+
## Mechanical hardware
- Steel shafts 5 / 6 / 8 mm (cut to length)
- Bearings 608 and 626 (625 also in stock)
diff --git a/docs/build_log.md b/docs/build_log.md
index 4631a93..8d96ec8 100644
--- a/docs/build_log.md
+++ b/docs/build_log.md
@@ -109,3 +109,13 @@ What the first metres taught: the single-sided track adjustment leaves the tensi
The CAD already has the second tensioner on each track; one of them still lacks its adjustment screw, to be added next. No autonomous behaviour yet; this is a drive test.
+
+## 2026-09-24 — V0.4 hull and simpler wiring
+
+The physical prototype remains V0.3 and continues assembly and testing after its first drive. In parallel, V0.4 CAD now includes the revised hull and component arrangement: Jetson seat and USB-side cable space, compact motor driver and servo adapter, QT Py / terminal-block provision, a larger battery envelope and motor fasteners accessible from outside.
+
+Two Adafruit control sets have been ordered: QT Py ESP32-S3, Terminal Block BFF, snap-on enclosure, INA219 STEMMA QT and a 300 mm cable. They are not installed yet. The existing INA model in CAD is approximate; the newly ordered Adafruit board still needs a footprint and connector-fit check. Details and costs are in [the V0.4 update](hardware_v04.md).
+
+A [40-second hardware-tour animation](../media/video/MOSS_V04_Inside_and_Action_20260924.mp4) shows the exterior leaving frame, an orbit of the hull and labelled components, reassembly and a scripted pickup. This is CAD animation; it does not report physical autonomous pickup or a new Jev experiment.
+
+Remaining work includes the additional track tensioners, real connector fit, bin magnet mounts, wheel/roller screw recesses, new controller firmware mapping and physical V0.4 validation. The integrated lower platform beneath the arm is still fixed; making it detachable is deferred.
diff --git a/docs/hardware_v04.md b/docs/hardware_v04.md
new file mode 100644
index 0000000..154a793
--- /dev/null
+++ b/docs/hardware_v04.md
@@ -0,0 +1,63 @@
+# Hardware V0.4 — development update
+
+24 September 2026. The physical rover remains **V0.3**. **V0.4** is the next hardware revision, being developed from assembly feedback while the existing prototype is tested. It is not yet a printed, assembled or qualified replacement.
+
+## Control electronics ordered
+
+Two of each item below were ordered, providing two sets. Quantities in the first numeric column are for one rover. Prices are the builder's order prices, not a complete robot cost.
+
+| Item | Per rover | Ordered | Unit USD |
+|---|---:|---:|---:|
+| [QT Py ESP32-S3, 8 MB Flash / no PSRAM, PID 5426](https://www.adafruit.com/product/5426) | 1 | 2 | 12.50 |
+| [Terminal Block BFF, PID 6495](https://www.adafruit.com/product/6495) | 1 | 2 | 10.95 |
+| [Snap-on enclosure, PID 6505](https://www.adafruit.com/product/6505) | 1 optional | 2 | 4.95 |
+| [INA219 STEMMA QT, PID 904](https://www.adafruit.com/product/904) | 1 | 2 | 9.95 |
+| [STEMMA QT / Qwiic JST-SH cable, 300 mm, PID 5384](https://www.adafruit.com/product/5384) | 1 | 2 | 1.25 |
+| **One set, including enclosure** | | | **39.60** |
+
+Two sets: **$79.20** in parts, **$21.83** shipping, **$101.03** order total before any separately collected import charges. Without the optional enclosure the parts used in one integrated hull cost **$34.65**. This is a control-electronics subtotal, not a revised total BOM.
+
+The BFF provides screw terminals for the harness. Headers still need soldering to the QT Py; the wires themselves terminate in connectors or screw terminals. One STEMMA QT cable connects each INA219 to its controller. The QT Py and BFF have different I2C pin routes on this ESP32-S3 variant, so the firmware must match the connector actually used. The old DevKitC pin numbers are not a drop-in QT Py wiring map.
+
+The INA219 remains a **battery-voltage monitor** in the present design. Its stock shunt measurement range is ±3.2 A; this update does not put the complete rover current through it or validate current-based battery monitoring.
+
+Ordered is not received or tested. The existing V0.3 wiring and firmware remain the reference for the physical prototype until the new electronics are fitted and tested.
+
+## Other selected changes
+
+- One **Cytron MDD3A dual-channel driver**, ordered earlier, is selected to replace the two MD10C drivers in V0.4. The first-drive prototype still uses the MD10Cs. The MDD3A has not been validated on MOSS in this update; its purchase price is not supplied here.
+- The **Waveshare Bus Servo Adapter (A) V1.1** remains the separate SO-101 servo interface. The General Driver for Robots is already removed from the selected architecture.
+- A **30 cm USB-A to right-angle USB-C cable** is ordered for the RealSense D455f. USB3 operation and the actual plug envelope remain to be tested at delivery.
+- Panel-mount **5.5 × 2.1 mm DC jack** selected for the charging connection. The supplied DC-022B drawing specifies an M8 × 1 mounting barrel. This is a connector provision, not a charging circuit.
+- Four **40 × 10 × 5 mm magnets** are ordered to retain the bin. Their mounting pockets and retention arrangement are still to be designed.
+- The fuse remains in the power path, in an **inline cable holder**. Removing the separate printed fuse-holder mount does not remove electrical protection or establish a final fuse rating.
+
+## Current hull CAD
+
+The latest local assembly is named `MOSS_Hull_V04_Cartes_INA90.step`. It includes the hull, electronics, battery envelope, motor supports and motor reference models. Its companion hull-only STEP/STL is `MOSS_hull_V04_cartes_INA90`.
+
+- Jetson retains its original plastic base. The removable caddy is replaced by an integrated seat and retaining pieces, with a cable route on the USB side.
+- Integrated mounting provision for the QT Py / BFF. The purchased snap-on enclosure is optional and is not the mounting interface used in this hull CAD.
+- Cytron and Waveshare positions have been exchanged. The Waveshare is rotated 90 degrees for USB access; the INA219 and its mounting features are also rotated 90 degrees for wiring access.
+- The existing battery measures **67 × 74 × 27 mm**. The blue **75 × 82 × 32 mm** volume is a candidate larger-pack envelope only, with wire and XT connector space. It establishes neither a purchased battery nor capacity or runtime.
+- The **lower platform integrated into the hull beneath the arm** is still fixed. Making it detachable has been deferred.
+- Motor-support fastening is designed from the outside through recessed M4 heads in the hull into captive M4 nuts in the supports. The local study uses standard M4 × 20 screws; final assembly access and printed fit still need validation.
+
+The latest hull export passed local solid/mesh and clearance checks. Those checks are not physical assembly validation. The INA reference is an approximate model of the **existing module**, not proof of fit for the newly ordered Adafruit board. Its footprint, mount positions and STEMMA connectors must be reconciled before printing that revision. The motor reference model also does not certify the actual motor screw pattern or encoder/cable clearance.
+
+Only the hull and explicitly requested component placements were advanced in this work. Unrelated printed parts were not silently replaced. The manufacturing files are still awaiting release review.
+
+## Assembly feedback still to close
+
+- Fit the second tensioner on each track and validate even tension. A track shortened by one tooth is a candidate for testing, not a validated new belt length.
+- Add screw-head recesses in wheels and rollers.
+- Use commonly available screw lengths and head types; do not require custom 26 mm screws or routine screw shortening in the released design.
+- Keep the corrected flank geometry. The suspected tensioner conflict was partly due to trying the wrong rail; do not relocate the fourth fixing solely on that initial diagnosis.
+- Validate the more continuous front cover / hull interface separately. No replacement cover is being released by this hull update.
+- Finish the bin magnet retention, new INA mounting, real connector fit and the electronics/firmware migration before calling the V0.4 BOM reproducible.
+
+## Animation
+
+[V0.4 hardware tour and pickup animation](../media/video/MOSS_V04_Inside_and_Action_20260924.mp4) — 40 seconds, 1920 × 1080, 30 fps, silent.
+
+The exterior moves out of frame, the camera orbits the current hull and components with readable labels, then the rover reassembles, rolls and picks up a can. This is a **scripted CAD presentation**, not footage of autonomous pickup, a Jev run or a new physical test. Exterior geometry is reused from the previous exploded view; the interior reflects the saved V0.4 assembly. The separate Jev / MuJoCo demo is unchanged.
diff --git a/docs/printed_parts.md b/docs/printed_parts.md
index 471387c..f7563f2 100644
--- a/docs/printed_parts.md
+++ b/docs/printed_parts.md
@@ -242,7 +242,7 @@ G/D and −1/+1 retain the CAD naming so parts can be matched to their source fi
## Naming and release status
-The current physical prototype is **V0.3**. R8, R8H, R10, R11 and V2-01 inside filenames are historical component identifiers; retain them for traceability and do not treat them as the current robot version. The next hardware revision is not assigned a new number here. A successful print and an assembled, validated part are separate milestones.
+The current physical prototype is **V0.3**. R8, R8H, R10, R11 and V2-01 inside filenames are historical component identifiers; retain them for traceability and do not treat them as the current robot version. The next hardware revision is V0.4; its in-progress changes are tracked separately in `hardware_v04.md`. This inventory remains the V0.3 reference. A successful print and an assembled, validated part are separate milestones.
## Source records
diff --git a/docs/roadmap.md b/docs/roadmap.md
index e312c2a..d41407d 100644
--- a/docs/roadmap.md
+++ b/docs/roadmap.md
@@ -1,9 +1,11 @@
# Roadmap
-- **Now (V0.3 prototype)**: the print set is complete and the rover rolled for the first time on 22 September. Fix the track tension (second tensioner per track), measure ticks per revolution, write the Jetson–ESP32 interface then speed control; fit the arm. What turns out wrong is logged in [assembly_feedback.md](assembly_feedback.md); some fixes go on the prototype, the others into the next hardware revision (not numbered yet), which also gets mounts for the ESP32-S3, the servo adapter and the real power distribution. Manufacturing files are published part by part once their validation is decided.
+- **Now (V0.3 prototype)**: the print set is complete and the rover rolled for the first time on 22 September. Fix the track tension (second tensioner per track), measure ticks per revolution, write the Jetson–ESP32 interface then speed control; fit the arm. What turns out wrong is logged in [assembly_feedback.md](assembly_feedback.md); some fixes go on the prototype, the others into V0.4, which also gets mounts for the ESP32-S3, the servo adapter and the real power distribution. Manufacturing files are published part by part once their validation is decided.
- **Next**: first roadside run, teleoperated with a gamepad through dimOS; first recordings.
- **Then**: mount the arm and gripper; collect pick-up demonstrations; publish the first pooled dataset.
- **Later**: first policy; first autonomous pick on video; kits so the first hundred builders do not have to source parts themselves.
No dates. Each step is posted when it is real.
- **Lower-cost Pi build (experimental)**: Raspberry Pi 5 8 GB with either Mighty Camera or STEVAL-VL53L9 3D ToF, plus a colour camera. Start with teleop and data collection, then evaluate navigation and detection with dimOS on the Pi. Next: mount and test both sensors with dimOS, teleop and data collection first, then navigation and detection. Nothing is validated until it runs on the rover.
+
+- **V0.4 hull and wiring**: current CAD includes the revised electronics arrangement, Jetson cable route, battery envelope and exterior-access motor fasteners. QT Py, screw-terminal BFF and INA219 STEMMA QT replace the DevKitC and the bare INA module. Reconcile the new INA footprint and real USB connectors, port the controller pin map and validate the printed assembly before release. [Detailed status](hardware_v04.md).
diff --git a/hardware/README.md b/hardware/README.md
index 03a8f7f..40d52f9 100644
--- a/hardware/README.md
+++ b/hardware/README.md
@@ -1,6 +1,6 @@
# Hardware
-Design in Fusion 360. The current prototype is **V0.3**, being assembled from its first print set (113 pieces, see `docs/printed_parts.md`). Fit problems are logged in `docs/assembly_feedback.md`: some are fixed on the prototype, the others wait for the next hardware revision, which has no number yet.
+Design in Fusion 360. The current prototype is **V0.3**, being assembled from its first print set (113 pieces, see `docs/printed_parts.md`). Fit problems are logged in `docs/assembly_feedback.md`: some are fixed on the prototype, the others wait for V0.4. The current hull CAD and ordered electronics are described in [the V0.4 update](../docs/hardware_v04.md); this is not yet a tested replacement for the prototype.
Manufacturing files (STL/STEP) are published here part by part once their validation is decided, with the fastener list (type, thread, length, head, quantity, what goes into PETG with inserts) and the sources they derive from (SO-101 arm, NormaCore gripper) with exact revisions and local changes. Until then, the photos in `media/`, the inventory and the BOM are the reference.
diff --git a/media/images/MOSS_V04_Inside_20260924.jpg b/media/images/MOSS_V04_Inside_20260924.jpg
new file mode 100644
index 0000000..4d437c8
Binary files /dev/null and b/media/images/MOSS_V04_Inside_20260924.jpg differ
diff --git a/media/images/moss_v04_hardware_tour.gif b/media/images/moss_v04_hardware_tour.gif
new file mode 100644
index 0000000..2ab49aa
Binary files /dev/null and b/media/images/moss_v04_hardware_tour.gif differ
diff --git a/media/video/MOSS_V04_Inside_and_Action_20260924.mp4 b/media/video/MOSS_V04_Inside_and_Action_20260924.mp4
new file mode 100644
index 0000000..8748f9c
Binary files /dev/null and b/media/video/MOSS_V04_Inside_and_Action_20260924.mp4 differ