Skip to content

Make something from Roborock S8 Robot Vacuum Cleaner — Smart Home Patrol Bot

TL;DR (one-liner)

  • Attach a Raspberry Pi camera + Pi controller, integrate with Roborock local control (python-miio or Valetudo), set patrol routines, stream video, publish alerts via MQTT/Telegram.

Safety & legal note

  • Respect privacy — inform housemates/guests.
  • Don’t use for spying on others or unlawful surveillance.
  • Warranty risk: flashing firmware (Valetudo) may void warranty. Proceed only if comfortable.

What you’ll need (parts)

  • Roborock S8 (base).
  • Raspberry Pi 4 (2GB+) or Pi Zero 2 W (for light use).
  • Pi Camera Module (or USB webcam).
  • USB power bank (5V 2–3A) or 5V supply near dock.
  • MicroSD card (16GB+) for Pi OS.
  • Optional: USB microphone / speaker for voice alerts.
  • Optional: PIR sensor or ultrasonic sensor (for local motion triggers).
  • Ethernet cable or reliable Wi-Fi.
  • Small velcro/3M mounting strips and a lightweight bracket to attach Pi to robot or base.
  • Laptop for setup.
  • (Advanced) TTL serial adapter, screwdriver kit — only if flashing Valetudo.

Two approaches — pick one

  • Non-invasive (recommended): Keep Roborock firmware intact. Use local LAN control (python-miio) or cloud control to send start/stop/go_to commands; Pi handles camera, streaming, motion detect, alerts. Lower risk.
  • Full local control (advanced): Install Valetudo or other local firmware for direct map+goto control and MQTT integration. More powerful but riskier (voids warranty; requires serial access).

Prep — common setup steps

  1. Flash Raspberry Pi OS to SD: Raspberry Pi Imager → Raspberry Pi OS Lite.
  2. Boot Pi, enable SSH, connect Wi-Fi.
  3. Update:
    • sudo apt update && sudo apt upgrade -y
  4. Install basics: sudo apt install python3-pip git curl -y

Step A — Non-invasive setup (full guide)

A1 — Get Roborock token (LAN control)

  • Install Mi Home (Xiaomi) / Roborock app, add device.
  • Export device token/credentials (tools: python-miio token extraction via backup or use miio instructions).
  • NOTE: token extraction methods vary; follow community guide specific to your app/OS.

A2 — Install python-miio on Pi

  • pip3 install python-miio
  • Quick test (replace TOKEN & IP):
pip3 install miio
mirobo --ip 192.168.1.55 --token YOUR_TOKEN raw_command get_status
  • If success → Pi can start/stop/vacuum/go_to.

A3 — Pi camera streaming

  • Enable camera: sudo raspi-config → Interfaces → Camera → Enable.
  • Install motion or ffmpeg streamer. Example using ffmpeg + simple HTTP stream:
# install
sudo apt install ffmpeg -y

# start simple stream (adjust device)
ffmpeg -f v4l2 -framerate 15 -video_size 640x480 -i /dev/video0 \
 -f mpeg1video -b 800k http://0.0.0.0:8081/feed1.ffm
  • Or use mjpg-streamer for MJPEG: sudo apt install mjpg-streamer (or build).

A4 — Motion detection + alerts

  • Install motion (motion detection daemon) or write Python using opencv:
    • pip3 install opencv-python numpy paho-mqtt python-telegram-bot
  • Minimal motion script (pseudo):
# detect_motion.py (sketch)
import cv2, time, paho.mqtt.client as mqtt
cam = cv2.VideoCapture(0)
mqttc = mqtt.Client(); mqttc.connect("mqtt-broker.ip",1883,60)
ret, frame1 = cam.read(); ret, frame2 = cam.read()
while True:
    diff = cv2.absdiff(frame1, frame2)
    gray = cv2.cvtColor(diff, cv2.COLOR_BGR2GRAY)
    blur = cv2.GaussianBlur(gray,(5,5),0)
    _, thresh = cv2.threshold(blur,20,255,cv2.THRESH_BINARY)
    if cv2.countNonZero(thresh) > 3000:
        mqttc.publish("home/robot/patrol/motion","1")
    frame1 = frame2
    ret, frame2 = cam.read()
    time.sleep(0.2)
  • Hook MQTT → Telegram or push notifications.

A5 — Patrol logic (Python controller)

  • Logic: scheduled runs + on-demand + motion-triggered investigate.
  • Example flow:
    • Schedule at night: cron → run patrol.py.
    • patrol.py sends miio commands: start_clean, go_to known coordinates, or spot_clean.
    • While moving, Pi streams camera; motion triggers higher-res capture and notification.
  • Example patrol.py skeleton:
from miio import Vacuum
v = Vacuum("192.168.1.55", token="YOURTOKEN")
# start
v.start()
# wait, then spot or go_to if supported
# v.goto(x,y)  # requires advanced firmwares / Valetudo
# stop & return
v.home()

A6 — Map + checkpoints

  • Without Valetudo, exact go_to may be limited. Use timed segments:
    • Start vacuum → let run X seconds → pause → rotate camera capture → resume.
  • With Valetudo (advanced) you can go_to coordinates precisely.

A7 — Notifications & remote view

  • MQTT topics:
    • home/robot/patrol/status
    • home/robot/patrol/motion
  • Hook to Telegram bot:
    • Create bot, get token, send images on motion event (via python-telegram-bot).

Step B — Advanced: full local control with Valetudo (power user)

Only if comfortable opening vacuum and using TTL serial. Warranty risk.

  1. Open robot — find serial header on mainboard.
  2. Connect TTL adapter (3.3V) to board.
  3. Back up original firmware.
  4. Flash custom bootloader + install Valetudo (or other local firmware).
  5. Configure Valetudo → exposes REST API + MQTT + map/goto commands.
  6. Use Pi to send REST/MQTT commands: precise goto, resume, pause.
  7. Use mapping to create checkpoints and run patrol route reliably.

Example: simple end-to-end patrol script (non-invasive)

  • Precondition: python-miio control works; Pi camera streaming online; MQTT broker reachable.
  1. cron nightly: */30 22-06 * * * /usr/bin/python3 /home/pi/patrol.py
  2. patrol.py does:
    • v.start()
    • wait 90s
    • v.pause()
    • capture 10s of camera frames → analyze for motion
    • If motion → publish MQTT alert + send Telegram image + v.home() and stop
    • else resume and continue next segment
  3. Results: periodic checks, live stream, alerts on movement.

Example code snippets (copy-paste ready)

  • Install libs: pip3 install python-miio paho-mqtt python-telegram-bot opencv-python
  • Minimal motion + notify (combine from earlier snippets). (User can request full scripts and I’ll paste entire files.)

Mounting & power tips

  • Mount Pi near docking station rather than on robot to avoid weight/obstruction.
  • If mounting on robot: use lightweight Pi Zero + small camera; secure with velcro.
  • Power: use small powerbank with pass-through so robot can still charge; or run Pi off wall supply at docking station.

Practical patrol ideas / behaviors

  • Nightly perimeter sweep (quiet) → send snapshots on motion.
  • Doorway-focused checks: aim camera near main door while robot is near dock.
  • Passive monitor: robot stays docked; Pi camera + PIR does the detecting. Robot dispatched only on alert.
  • Map-based patrol (Valetudo): create loop of waypoints → goto each → capture → move on.

Troubleshooting

  • python-miio token errors → re-extract token; ensure LAN mode enabled.
  • Camera lag → reduce resolution/framerate.
  • Robot stuck during manual move → monitor and abort via v.stop().

Extensions (nice extras)

  • Integrate with Home Assistant: expose entities, automation, UI.
  • Add two-way audio: microphone + speaker for warnings.
  • Add thermal sensor camera (FLIR) for night detection (advanced, expensive).
  • Create web dashboard with map, live feed, last alerts.

Quick maintenance checklist

  • Test dry runs in daytime.
  • Lower camera resolution for reliability.
  • Log events with timestamps.
  • Keep emergency stop accessible.

Tags: