LED-bot overview

A small device based on an microcontroller with a square LED array. Attached with a microphone. This project was my first attempt to use AI supported coding. I used claude api. The results were quite sobering:

• claude is very forgetful. I had to repeat my goals very often.
• claude is very expensiv. For this little project I had to pay about $30.
• The code is well documented but very poor structured. It's hard to find the desired function.

YouTube Video: LED-bots listening to music

Rear view of WS2812 8x8 panel with ESP32-C3 and Mems module mounted

bot capatibilities

• Each node is part of a swarm but also can live alone.
• each node has 3 inputs:
  • message from neighbour
  • recorded noise from microphone
  • randomly generate time slot
• each node has 2 output.
  • message to neighbours
  • animation on led array
• each node has the following states:
  • booting
  • dreaming
  • following (music, talks)
  • panic level 1
  • panic level 2
  • panic level 3
• each bot has a set preset animations displayed on the led array:
  • breathing(frequency, colorset)
  • letters(string, color, bgcollor)
  • blink(frequency, colorset)
  • sparcling(intensity, colorset)
  • rotation(frequency, colorset, direction)
  • spectrumLevel(bass, midrange, treble)

Hardware

• Esp32c3 super mini
• ws2812 based 8x8 rgb led array
• INMP441 mems module on i2s
• LiIon 18650 battery / usb-charger

Software

• Framework: ESP-IDF
• Build-System: PlatformIO with framework = espidf

State machine

• when a node is started it will

  • annouce its id via a letters-animation(mac-adress)
  • initialize the interrupt driven message recieving routines
  • wait 5 seconds
  • broadcast a request for rssi via wifi
  • all hearing nodes will answer their mac and rssi
  • the ids (mac) of the 2 nodes with the highest rssi will become its neighbours (- after a second all other nodes will also perform this procedure to refind their 2 neighbours)
  • the microphone will be initialized
  • using the mic meassurment the random generator will be seeded
  • a random time periode between 2 and 30 minutes will be generated and a timer started
  • the main loop is started. Within this the following controll steps are done:
    • analyse the sound buffer

• when a node recieves a request for rssi

  • it will send the rssi to the requester
  • it will initiate a rssi request to update its own neighbourhood.a request storm / loop must be prevented.

• when it recieves a panic level message from a neighbour

  • it will set the panic level flag
  • starts a delay panic timer. the panic animation should start with a delay of 200ms to achieve the effect of a color change wave in the swarm.
  • it will forward the panic level message to other neighbour (not the sender).

• when a random timer message is recieved:

  • the ramdom animation should start at once.

• when the sound analyser routine detects:

  • "silense" for at least 1 minute the breathing animation is started
  • "talking" the rotation anmation is started
  • "music" the spectrumLevel animation ist started
  • "hand claps":
    • the blink animation is started
    • a panic level 1 message is send to the 2 neighbours
    • the level animation should be held for 10 seconds
  • "screaming":
    • the blink animation is started with other values
    • a panic level 2 message is send to the 2 neighbours
  • "thunder":
    • the blink animation is started with other values
    • a panic level 3 message is send to the 2 neighbours
    • the level animation should be held for 2 seconds

• when the random timer is fired: - the blink animation is started with other values - a new random time is generated and the timer set - the random message is broadcasted to the swarm