Unity 6000.6.3 removed Assets/Resources/GDKEditionAutoGen when it opened the
project (Microsoft GDK package settings; the project doesn't use the GDK).
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
- TOP_GROW: top shell 10 mm taller (room behind the display); the bottom
shell is unchanged, dependent heights follow Z_SPLIT / Z_BOT.
- Nextion dimensions as measured by the owner: board 100.6 x 55, LCD
85.7 x 55 x 3.86 plus a 1.3 mm flex cable, 11.21 mm glass to back of J1.
LCD pocket allows the cable on either short side; bezel 1.0 mm (the
display sat on the Teensy).
- Display posts tapped M3 for M3 x 5 machine screws, 1.2 mm skin under the
top face (was M2.5 self-tapping). README screw list and PLANNING.md
updated; STLs regenerated.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Dongle (app):
- bridge mode: stands in for the robot's radio (ch 96 / 250k / HXPD1,
auto-ack, dynamic payloads, ack payloads). Received packets go to the PC
as A5-framed 'R' frames; the PC's 'A' frames become the ack payload of
the next packet, like robot/Comms.cpp; dropped after ~0.5 s without a
refresh. 'S' status once a second.
- USART RX through circular DMA into a ring buffer, overrun stall
disabled, error flags cleared: a burst (e.g. SignalRGB probing the port)
used to deafen the app until a replug.
- RX test mode ('x') accepts protocol v4's dynamic-length packets.
- safeboot_flash.py --knock: keeps knocking across a replug to catch
safeboot's cold-boot window when the app doesn't answer.
Unity:
- SimRemote sources: Keyboard, or the real remote via the dongle
(DongleLink.cs: kernel32 serial, auto-finds the CH340, all port I/O on
one background thread). Packets go into the firmware byte for byte, the
robot's telemetry goes back on the ack, so the remote's display shows
the sim. Tested end to end with the real remote.
- Dropped the USB joystick and robot-serial-harness inputs and the old
Serial/Serial2/Serial3/HandleHoystick scripts (and their components in
Main.unity): no USB links to remote or robot any more.
- Test panel: dongle status; no sideways scroll bar.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
make_remote_hmi.py now builds the project with Editor display direction 270
(landscape, flipped) before adding pictures and fonts, so they're stored for
that direction. Checked in Nextion Editor and on the real display.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
- Nextion is on Serial2 (pins 7/8) - already in the last commit; docs updated.
- MPU6050 on Wire (18/19) so everything is on the edge headers (owner's
choice); the two stick wires from 18/19 move to A7/A8 (21/22).
- Stick mapping checked on the real remote: left stick X/Y/twist on
A3/A7/A8, right on A0/A1/A2, all inverted but the right twist.
- RemoteIMU: full device reset before waking. The remote's MPU6050 (likely
a clone) answers and wakes without it but every data register reads 0.
Axis defaults confirmed (+pitch nose down, +roll right side down).
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Protocol v4: CmdPacket grows to 26 bytes (cal_cmd, cal_servo, cal_value);
new 30-byte Telemetry rides back on the nRF24 ack payload (robot state,
IMU, battery, 18 joint angles, packets/s, servo-cal echo). Each end drops
wrong-length packets, so version mismatches still fail closed.
Robot (owner-approved): SERVO_CAL state while disarmed, per-servo centre
trims (+-200 us) in EEPROM, telemetry, battery reading; the IMU estimate
now always runs (leveling still gated by ROBOT GYRO).
Remote: generated HMI (scripts/make_remote_hmi.py) with pages splash,
drive, status, servos, cal, setup, upload; page-aware NextionUI that only
writes the showing page and only what changed; servo calibration and
UPDATE DISPLAY on the touchscreen and serial console; battery module.
The display is on Serial2 (pins 7/8), found on the real remote; the old
sketch's Serial1 was never connected to it.
Sim: RF24 shim supports ack payloads, API version 8
(hxsim_get_ack_payload); sim_smoke checks telemetry and every CAL command.
Unity CmdPacket mirror updated.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
The nxhmi MCP server (.mcp.json) edits and renders the remote's Nextion HMI;
scripts/make_remote_hmi.py uses its Python package from the submodule.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
The old HMI's pages are 0 splashsceen, 1 main, 2 servocal, 3 page0,
4 uploadscreen, 5 joysticks; the link-status texts are main.t1 and
page0.t1 (not page1.t1/page3.t1, which the display silently ignored).
remote.HMI is the 2019 project re-saved by Nextion Editor 1.68, to work
on here instead of in 2019-teensy-nrf24/.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
- Confirmed the status LED is on PA0 (active-high). README pin table updated:
amplifier (AT2401C) control is one of PA1/PB1/PF0/PF1, still to distinguish.
- app.c gains bring-up modes for mapping the amplifier pins:
- 'x' RX on the hexapod protocol (ch96/250k/HXPD1, 22-byte) + packet counter,
to find the RX-enable pin by which pin lets the remote's packets arrive;
- 'f'/'n' carrier-detect (RPD) finder (too noisy alone, kept for reference);
- 'a' sweep every usable pin HIGH ~3s for LED/scope checks.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
- dongle_flash.py aborts before erasing unless the ROM-bootloader link tests
clean (8/8 GETs), so a flaky software-jump link can't wipe a working stick.
- open_racing() retries opening the port to win the race against SignalRGB,
which grabs CH340 ports on enumeration.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Reverse-engineered the AliExpress "USB_24RF Ver3.1" stick (STM32F030F4P6 +
Si24R1 + AT2401C PA + CH330N) and replaced its closed AT-command firmware so it
can be the PC-side radio for the hexapod.
- Pin map found over SWD and confirmed at runtime: USART1 PA9/PA10 (CH330N),
SPI1 PA5/6/7, nRF CSN PA3 / CE PA4 / IRQ PA2. LED + AT2401C control are among
the 5 free pins, still to map.
- Two-stage firmware so wire-free flashing can't be broken by an app update:
- safeboot (4 KB, pages 0-3): permanent recovery bootloader, fixed-baud flash
protocol over USB, only it survives every update;
- app (12 KB, pages 4+): the radio firmware, the only part updates rewrite.
The app updater never erases safeboot; recovery works even with a dead app.
- tools/dongle_flash.py: one-time install via the STM32 ROM bootloader (the only
time the BOOT0 wire is used). tools/safeboot_flash.py: wire-free app updates.
- dongle/README.md documents the hardware, pin map, RDP1 (factory fw can't be
dumped), the memory map and the flashing workflow.
- README: add dongle/ to the layout and the PC-radio plan.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
- remote_v2.scad: main board is the measured 56 x 80 mm perfboard, held by
4 side clamps (M3 button head + washer into inserts), no holes in the board
- export_3mf.py writes remote_v2_top_inlays.3mf (top shell + honeycomb +
ON/OFF on filaments 1/2/3); remote_v2_top_multicolor.3mf stays the
owner's OrcaSlicer project
- STLs re-exported, README printing table and PLANNING notes updated
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
- Bottom shell: the brace pivot bosses stand on pillars and the lip's tie
band steps out at 45 deg (both were floating overhangs); the kickstand
barrel sits on the bed. Checked by slicing the STLs layer by layer: only
bridges are left. STLs re-exported.
- remote_v2_top_multicolor.3mf: now the owner's OrcaSlicer project (all
parts on two plates); export_3mf.py regenerates the plain multi-part file.
- README "Planned": a USB nRF24 stick on the PC so Unity can talk to both
the robot and the remote (what to work out), plus pointers to the other
notes for later.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
remote_v2.scad: a new remote for the same electronics, 216 x 116 x 32 mm,
ASA, two shells printed without supports and joined with M3 heat-set
inserts. Hangs level from a neck strap on a balance-point eyelet (top face
or front edge) with a fold-out body brace; kickstand for the table.
JH-D300X-R1 sticks clamp the plate; main perfboard (Teensy 4.0, nRF24,
IMU) clamped at its sides under the display; UPS board with its LEDs
through the floor. Red honeycomb and white ON/OFF inlays in the first
3 layers. Every part checked for collisions.
- stl/remote_v2/: print-ready STLs, plus a multi-part 3MF (export_3mf.py)
with the inlays on filaments 2 and 3
- ARCHIVE.md: the original remote with the owner's photos; README shows v2
and how to print it
- Notes for later: battery-voltage reading, IMU on Wire (18/19) with two
stick axes moved to A7/A8, a stable 6 V supply for the 18 servos
- remote.scad: its sticks are JH-D300X-R1 (the model uses the D400X drawing)
- *.3mf through Git LFS; repo README says Teensy 4.0 remote
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Board teensy41 -> teensy40. All pins exist on the 4.0; the MPU6050's
Wire2 (SDA2 25 / SCL2 24) is on the underside solder pads.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
hardware/openscad/ rebuilds the broken Fusion models as OpenSCAD:
body (top/bottom plate, standoffs, battery bay, switch), legs (coxa
bracket, femur link, tibia, horn adapters, TPU socks, servo cables),
the whole robot, and the remote (shell, lid, Nextion 3.5", JH-D400X-R4
sticks, 18650 UPS board, IMU, Teensy 4.0, USB-C breakout). Every part
was checked by slicing against the original STEP solids; the original's
bugs are fixed and listed in PLANNING.md.
The robot now uses a hexagon leg layout (corner legs 60 deg round from
the mid legs; the corner coxa brackets need turning on their splines).
Physical limits from a collision sweep of the model are in Config.h:
per-leg coxa limits and LEG_CLEARANCE_MM 56 (a leg is 55.2 mm wide at
the joints); the sim API exports them (version 7). sim_smoke passes.
Unity loads the model (export_unity.py -> Resources/OpenScadModel,
OpenScadModel.cs builds a new rig in place of the old scene's robot).
The PCB model's mounting holes now use the exact EasyEDA grid. README
shows renders of both.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
The Teensy 3.6 carrier board mounted inside the robot body (Teensy
peeks through the top-body hole). Contents:
- PCB_test_2026-11-13.json : EasyEDA PCB source (editor v6.5.61).
- OBJ_test.obj + .mtl : EasyEDA 3D OBJ export (slab + holes only).
- top.png, bottom.png : EasyEDA renders of each side; bottom.png
is the authoritative source for the
JP1-JP7 silkscreen numbering.
- 20190524_234512.jpg : photo of the populated bottom side.
- 20190524_234518.jpg : photo of the top side, powered on.
- main-pcb.scad : parametric OpenSCAD reconstruction. Real
board outline, 4 M3 holes, and every major
component (Teensy 3.6, nRF24L01+, GY-521
IMU, 7805, ZENER, JP7 battery JST, six
JST-PH 2mm servo headers JP1-JP6, POWER
LED + 510R resistor, "Hexapod v1" silk) on
the correct side of the board. Placements
are the pad-group centres pulled from the
EasyEDA JSON, not eyeballed.
- README.md : explains the files, the EasyEDA format and
unit conversion, which side each part is
on, and an authoritative placement table.
Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
Fusion 360 STL exports of the body, legs, hinges, foot, servo horn and
test legs into robot-parts/, and an early remote part into remote-case/.
Found in an old Hexapod folder; the remote files that came with them were
already in the repo.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
PDF version of the IK derivation. It was in the Hexapod-Archief repo from its first commit (2014) until the 2016 cleanup removed it, so it was only left in that repo's history.
- original-frame/ moves into 2009-fpga/original-frame/ (the FPGA version introduced the frame).
- Media regrouped by the owner: the two 2011-07-01 PNGs and MOV02256-02259 to FireFly, SNC00198.jpg to the FPGA version.
- Removed from the current version (still in the history): docs/math wisFORlive.docx, wisFORlive3.docx, wisFORlive4.docx and 'z axis rotation.url'.
- READMEs, 2026-current/CLAUDE.md and the root timeline updated to match.
Comments in Config.h and the README referred to LEGACY/, which is folded into the generation folders. Also 'from the repo root' -> 'from the 2026-current folder'. Comment and documentation changes only; approved by the owner.
The root README has the timeline of all generations since 2009. Each generation folder explains its hardware and what is in it. .gitattributes turns off line-ending conversion so all files stay as they were.
The Arduino IDE needs the folder name to match the .ino: HexapodRobot/, Hexapod_NEW/ and Hexapod_Teensy/. References in 2026-current/CLAUDE.md updated.
Visual Studio build output (.obj, .pdb, .tlog, .sdf, .suo, ipch) of the 2012 IK visualiser and Unity's .vs, Logs and UserSettings of the 2018 simulation. They stay in the history. The built 'Hexapod app.exe' (Debug) is kept. Also drops the per-folder .gitattributes of the Processing sketch (the repo root now defines how files are stored) and the 2016 README of the old Hexapod repo (its only content was a GitLab link).
The LEGACY copies of HexapodRobot, Hexapod_NEW, Hexapod_Teensy and the old remote sketch were identical to the imported repos and are removed. The old Nextion design moves to 2019-teensy-nrf24/remote-nextion/, the 2019 remote case files to 2019-teensy-nrf24/remote-case/. Paths in CLAUDE.md, READMEs, remote/ comments, sim/build.ps1, the Unity DLL message and .vscode/settings.json now point to the new places. robot/ is not touched (its CLAUDE.md asks to check with the owner first).
Body bottom, bottom and mid leg, servo horn (v5 and v6), Hitec HS-485HB servo model and the remote case (v4 and v40), with their preview images and exports. From D:\Fusion 360 export.
Files from Electronics\Hexapod whose content differs from the archive copies. They keep their original file name; a date is added where the name would clash.
Everything created after 2 Nov 2011 20:44:07, incl. the 'Hexapod 29-11-11' film and media.jpg. From Electronics\Hexapod\media, Verwerkte media and media.jpg.
Math (IK, rotation), planning, research papers, concept images, the Arduino Mega + SSC-32 notes and the two 2011 Windows programs. From Hexapod-Archief documentation/ (files only; its history stays in the archived repo).
From https://mangavalk.com/Hexapod/Hexapod @ 7945005 (its Hexapod_Teensy
is identical to the copy already in LEGACY). Earlier electronics
generation: on-board IK driving an external SSC-32-style servo
controller. Notes on what still applies are in CLAUDE.md.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
- LEGACY/Hexapod_Teensy/: old robot sketch (reference only).
- platformio.ini: robot env targets teensy36 (owner-confirmed board).
- Config.h: per-servo SERVOS table (pin, pulse at 0 deg, signed
us/deg, pulse limits) built from the old sketch's pins, ranges, trims
and the old Unity servo directions. Rear-left/front-left coxa offsets
set to 0: the owner re-horns those two servos (the old +-90 left them
at the end of travel; the sim showed them clamping).
- ServoDriver: 18 servos on hardware PWM (50 Hz, 16-bit); the Teensy 3.6
Servo library stops at 12.
- Radio on CE 39 / CSN 15 with SPI SCK on pin 14.
- sim: analogWrite/PWM shims, pulses decoded through SERVOS (API v6).
- Docs: CLAUDE.md, READMEs.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
- LEGACY/remote/teensy/: old remote sketch from
https://mangavalk.com/Hexapod/remote.git @ 492dece (reference only).
- remote/Config.h: sticks on A0-A5, Nextion on Serial1 (0/1), as in the
old sketch; MPU6050 moved to Wire2 (25/24) since A4/A5 are Wire's pins.
- NextionUI: controls on a vscope-global "control" page added to the
old HMI; keeps the old t1 status texts; the old page-1 b2 button (and
serial 'U') enter Nextion upload passthrough, following the PC's baud.
- Docs: NEXTION.md, remote README wiring, CLAUDE.md.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Simulation
- sim/: the unchanged robot firmware compiled into hexapod_sim.dll
against PC stand-ins (Arduino, Servo, Wire + register-level MPU6050,
RF24 with matching-config radio), a kinematic body model (support
plane + planted-foot odometry) that also drives the simulated IMU,
and a headless smoke test (sim_smoke). build.ps1 builds and tests.
- Hexapod_Simulation_Unity/ folded in (history kept in a bundle
outside the repo). Assets/HexapodSim loads the DLL, drives the
existing leg rig, and offers joystick / keyboard / robot-serial
input plus a test panel.
Robot (all found or confirmed in the simulation)
- Config.h geometry taken from the Unity model.
- IK femur elevation sign, gait stance direction, body pose
rotation direction, yaw lever arm.
- Stance feet integrated from the current command; gait restarts at
mid-stance; heading integrates the applied turn rate, only while
walking.
- Leg clearance and reach checks (Limits::feetFeasible); walk command
and body pose scaled as a whole, pose scale rate-limited, planted
feet guarded so the body never snaps or legs collide.
- Robot-gyro self-leveling as an integrator.
- Heading mode: steering relative to the body; S walks backward.
- Standing height from the remote (STANCE_HEIGHT_RANGE_MM).
Protocol v3 (reflash both boards)
- CmdPacket gains body_height (22 bytes).
Remote
- Height slider (Nextion hHeight/nHeight, serial [ ]).
- Fixed unspecified I2C read order in RemoteIMU.
LEGACY/: owner's old Nextion HMI and remote hardware assets.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Not yet verified on hardware.
- shared/HexapodProtocol: CmdPacket (now 20 bytes, adds remote_roll /
remote_pitch), FLAG_* bits and radio settings as the single protocol
definition; static payload size = sizeof(CmdPacket) on both radios.
- robot: use the shared protocol; vx = forward (harness w/x <-> q/e);
InputMapper adds remote tilt on top of the right stick when
FLAG_REMOTE_GYRO is set, and robot.ino no longer zeroes stick
roll/pitch when it's off.
- remote (Teensy 4.1): Sticks (calibration, EEPROM), RemoteIMU,
RemoteState (interlocks, link-loss disarm, smoothing), PacketBuilder,
RadioLink, NextionUI, SerialConsole.
- CLAUDE.md, README.md, remote/README.md, remote/NEXTION.md.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Robot-side firmware (Teensy 4.0) imported unchanged as extracted.
platformio.ini defines [env:robot] (teensy40) and [env:remote]
(teensy41); scripts/select_src_dir.py points each env at its own
sketch folder so robot/ and remote/ stay Arduino-style sketches.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>