Fine for a side-project web app. Does not work for an inverted pendulum spinning on your desk.

The Pit and the Pendulum

MathWorks just published a Tech Talk by Brian Douglas. It is worth watching if you build hardware.

The setup is a Quanser cube servo with an inverted pendulum, a Raspberry Pi running the controller, and a coding agent wired into Simulink through the Simulink Agentic Toolkit. The agent’s job is to swing the pendulum up and balance it.

Balancing an inverted pendulum is a classic control problem because it’s unstable by nature. Get the math wrong and the pendulum falls. Get the code wrong and a 12V motor jerks the cube into something it shouldn’t.

What Happens If You Just Prompt It

The tempting move is the one most of us have done with web apps. Open a chat, describe the system, ask for the code. One prompt, big blob, ship it.

The MathWorks engineers walk through why that falls apart for hardware:

“If you rely on AI to generate code directly, that process is much less structured… it’s often unclear exactly how or why a particular implementation was created, making it harder to trust and verify.”

Brian Douglas

Hardware is unforgiving. You can’t roll back a destroyed motor with a git revert. When a physical system fails, you need to know which assumption broke, which equation got the wrong constant, which sample rate the controller is actually running at. A black box of generated C code doesn’t tell you any of that.

“You don’t want to let your AI agent skip parts of your engineering workflow. That’s how you can get yourself into trouble with a final product that doesn’t meet your objective or worse, is just purely unsafe.”

Brian Douglas

The Workflow Is the Script

The answer is to keep the AI inside the workflow engineers already trust: Model-Based Design. The agent doesn’t replace the workflow. It runs each step inside it, faster.

Define quantifiable requirements. Before any code, the human and the AI agree on what success looks like. Spin-up speed. Wobble limits. The 12V motor constraint. These targets become the benchmarks for verification later.

Use a trusted model. Instead of letting the AI invent a custom physics engine, the workflow points it at an existing Simulink model. The agent runs linearization through the Simulink Agentic Toolkit and reports the system’s poles back to the human. The math comes from validated tools, not from a language model freestyling differential equations.

Use a trusted controller. Same logic. The agent is told to use the MPC Toolbox. Model Predictive Control has decades of theory behind it and a deterministic implementation. If something looks wrong, you can tune it manually. You’re not debugging an optimizer the AI invented on the fly.

Verify in layers. Three stages before the code touches hardware:

• MIL (Model-in-the-Loop): the controller runs against the theoretical model
• SIL (Software-in-the-Loop): the generated embedded C code runs on a computer, behaving exactly like the simulation
• PIL (Processor-in-the-Loop): the C code runs on the actual target Raspberry Pi

Each layer catches different bugs. Skip one and you’ll catch that bug on hardware instead.

Deploy and iterate. Only after layered verification does the code go on the real pendulum.

“Instead of asking an agent to solve your problem however it wants, you tell it to follow an engineering workflow that you understand.”

Brian Douglas

The AI executes. The human directs.

The Hardware Truth

Here’s the part of the video I liked best. They ran the verified controller on the real pendulum and it didn’t quite work. The physical hardware behaved differently from the model.

This is what the vibe-coding fantasy ignores. Reality disagrees with simulation, especially the first time. You will need to iterate. The thing that matters is whether you can.

Because the team followed the workflow, they could. The model was traceable. The requirements were quantifiable. The controller was inspectable. They used the AI to update the physics model, ran it back through MIL, SIL, and PIL, and tried again. No starting over.

The workflow doesn’t promise perfect code on the first pass. It promises code you can fix.

The Director Pattern

A few months ago, I wrote about Seth DeLand’s framing of agentic AI for engineers. The headline idea: you’re not being replaced, you’re being promoted. The engineer becomes the one who defines the problem, sets the constraints, and shapes the space where the AI works.

The MathWorks Tech Talk is what that looks like in hardware. The AI isn’t the engineer. It’s the agent. You’re the director.

“Agentic AI is powerful, but for engineering, the value comes from how you use it. You get the speed and productivity benefits of AI without sacrificing trust, safety, or understanding.”

Brian Douglas

You don’t get speed and safety together by handing the AI a prompt and crossing your fingers. You get there by keeping the workflow you trust and letting the agent move faster inside it.

Speed without losing safety. Productivity without losing traceability.

For software, vibe coding is mostly fine. The blast radius is a wasted afternoon.

For hardware, the blast radius is a sparking motor and a pendulum on the floor.

Borrowing ideas from my Lofi Girl CheerLights Shadowbox project, George and I made a CheerLights Gift for under our Christmas tree. This gift changes color based on the latest CheerLights color and anyone on Twitter or Discord can change its color and the color of all of the other CheerLights projects.

The CheerLights Gift project is based on the Pimoroni Plasma Stick 2040 W and NeoPixel string lights for illumination. The Plasma Stick ships with CheerLights code main MicroPython file, so all you have to do is change the Wi-Fi network name and password. The microcontroller makes a request to the ThingSpeak API to get the latest CheerLights color, sets the color of the NeoPixel LED lights, waits 10 seconds, and repeats. This is about the fastest way to build your own CheerLights display.

I used a metal gift box that I found in a craft store to house the lights and controller. All we needed to do was stuff the LEDs inside and glue them in place. I only burned my fingers twice.

The project turned out well and kept my five-year-old’s interest for the duration of the build. I was happy that the Plasma Stick 2040 W made it really easy to get the project going.

Join the CheerLights IoT community at CheerLights.com.

The Lofi Girl Shadowbox is backlit with the latest color of CheerLights. It’s always changing based on what people set the color to on Twitter and Discord.

Parts

• Lofi Girl poster – I am a 1 A.M person
• Shadowbox
• Black foam board
• Foam block
• Pimoroni Plasma Stick 2040 W (based on the Raspberry Pi Pico W)
• LED string lights
• Micro USB cord and power supply
• Other things: utility knife, straightedge, hot glue gun, and repositionable glue squares

Assembly

The poster that I got from the Lofi Girl shop was 30cm x 45cm and on really thin paper. I used the repositionable glue squares to stick the poster on the black foam board. This gives the poster some structure. I cut a hole in the back of the shadowbox and feed the USB cable through. I cut out a rough shape of the Pimoroni Plasma Stick 2040 W and wrapped the string lights around the foam block.

Programming

The best part of the whole project is that CheerLights is baked into the Pimoroni Plasma Stick 2040 W. All you need to do is program the device with your Wi-Fi settings. Open Thonny and open the WIFI_CONFIG.py file. Change SSID to your wireless network name, PSK to your password, and COUNTRY to your country code like US or GB. Click Save.

Pimoroni has an excellent tutorial on getting Thonny installed and an explanation of the Pimoroni Plasma Stick 2040 W kit. I have also written a tutorial on the CheerLights blog.

Chill

Now that you’ve done all of the hard work, it’s time to chill and enjoy the Lo-Fi vibes tuned to CheerLights. While I was making this project today, I had Lofi Girl on YouTube, and the lights were just bright enough to see what I was doing without cutting myself. I am looking for a wall to permanently install my Lofi Girl Shadowbox. I am seriously thinking of drilling a hole in the wall to hide the USB power adapter.

I have been exploring MicroPython ever since I got a Raspberry Pi Pico W. The community is moving fast and making more and more examples and libraries for MicroPython. Having the Arduino team lean into making the Arduino Lab IDE for MicroPython is a good indicator where things are headed. I was able to build an awesome CheerLights display using MicroPython running on a Raspberry Pi Pico W connected to a Pimoroni Galactic Unicorn PED panel.

Consider Arduino Labs for MicroPython as pre-release, experimental software. Things are moving fast and the project will stabilize. Follow the project on GitHub and file any problems that you come across on GitHub Issues.

Recently, my college at MathWorks, Christopher Stapels, introduced me to the Raspberry Shake. The”shake” is a personal seismograph that laterally allows you to watch the Earth move and the things that cause the earth to move.

Christopher went on to collaborate with Alan Kafka at Weston Observatory, Boston College, and Jay Pulli at Raytheon to explore traffic data analysis using seismic data collected by the Raspberry Shake and sent to ThingSpeak.

The team explored the data using MATLAB and tried to establish some correlation between vibration and traffic density. Check out the project on the MathWorks IoT Blog. Whole Lotta Shakin’ Goin’ On.

For this season, I wanted to try out the Blinkt! kit from Pimoroni. I really what I am seeing from Pimoroni. Their starter kits and accessories for the Raspberry Pi are really well designed, colorful, and useful. If you are in the US, some of the Pimoroni kits are on Amazon.

I do recommend the Pimoroni Raspberry Pi Zero W Starter Kit if you are using the Raspberry Pi Zero W for the first time. The main reason for using a starter kit is that the ports on the Raspberry Pi Zero W need some adapters. You need to get a mini to full-size HDMI adaptor and a Micro-USB adapter to be able to connect a monitor, keyboard, and mouse to the Raspberry Pi Zero W. The kit also includes a 16GB micro-SD card with an operating system pre-loaded.

Connect a monitor, keyboard, and mouse to the Raspberry Pi Zero W and connect the USB cable to a reliable 5V USB power supply and you are ready to start programming.

I followed Pimoroni’s getting started tutorial to install the Blinkt! python library. The Blinkt! board is a strip of LEDs that sit on top of the Raspberry Pi’s GPIO pins. Blinkt! is a cool way to get started with controlling things with the Raspberry Pi. Consider turning on the LEDs for the first time using code as your “Hello World” program for physical computing.

After connecting your Raspberry Pi Zero W to your Wi-Fi network, open a terminal window and enter the following comment to install the Blinkt!~ libraries (type ‘y’ or ‘n’ when prompted).

curl https://get.pimoroni.com/blinkt | bash

Create a new file in your home directory called “cheerlights.py”. This file will contain the code that gets the latest CheerLights color on ThingSpeak and sets the color of the Blinkt! LEDs. Copy this code into the “cheerlights.py” file and save it.

python cheerlights.py

After a few seconds, you should see the latest CheerLights color on the Blinkt!

To keep the “cheerlights.py” running all of the time, you need to add the script to crontab. This will make the script stays running in case of an error or reboot of the device.

sudo crontab -e

Then, add this line to the end of the crontab file and save the file.

@reboot sleep 60 && sudo python /home/pi/cheerlights.py

A couple of years ago, I noticed that a maker space along with Qwirkshop created a desktop light kit for the Raspberry Pi Zero W and Blinkt! board. The kit was sold for a brief time on Etsy and I picked up a few kits. Well, I just found them in my storage closet! The kit is a bunch of laser cut pieces that you assemble together. It is an ingenious kit in that it was designed in layers and holds together with no glue. The results were beautiful.

Let me know if you build something to display the latest CheerLights color. I am always looking for new ideas for the CheerLights blog.

Let’s stay connected!

I recently was looking around my game room for some inspiration for a quick project. I have a custom arcade cabinet based on RetroPie running on a Raspberry Pi attached to a smart TV. It has been a lot of fun playing my video games from 30+ years ago. My project this time was to add a custom acrylic sign with an RGB backlight. I wanted the cabinet to look more like an actual arcade cabinet.

My go-to site for graphic printing is Signs.com. I have used it many times for upgrading Arkham Horror: The Card Game and for other projects around the house. I designed a sign using Adobe Illustrator and got it made out of acrylic. I cobbled together vector logos of many of my favorite systems plus the RetroPie logo and named the arcade cabinet, “The Scharcade”. See what I did there? Do you see it?

I used an RGB Led Strip with IR Remote Control as a backlight for the sign and to edge light the TV screen. The IR remote is handy to change the color of the lights. Just make sure that you expose the IR receiver on the power supply so you can change the color from the front of the cabinet.

My upgrade to the upgrade was to frost the arcade cabinet sign to make it look better. The original sign was completely transparent and you could see the lights shining through from each one of the LEDs. LEDs love diffusers. To fix this issue, I used Rust-Oleum Frosted Glass Spray Paint on the backside of the sign. I tested out the spray paint on a scrap piece of acrylic. I was really worried about ruining the custom sign. Once I was confident with the results, I went ahead and sprayed one even coat of the paint on the backside of the sign. It takes about 20-30 minutes in order for the paint to activate. I wanted in anticipation for 20 minutes and I gradually saw the clear acrylic turn frosted. I was really happy with the results. The 20 minutes felt light an eternity.

MQTT operates under a publish/subscribe model in which clients connect to the MQTT broker and either publish data or subscribe to topics to receive data from other clients.  MQTT messages can be sent using unsecured TCP, which is the easiest and has the lower RAM, CPU, and bandwidth requirements.  MQTT messages can also be sent using WebSockets.  This may be needed for situations when the standard port for MQTT, 1883, is blocked on the network.  When using MQTT over WebSockets, it is possible to encrypt the communication with SSL.  If your hardware is capable of using SSL over a WebSocket, that is the recommended method.

On the Raspberry Pi, an MQTT client library is required to send data to ThingSpeak.  Paho is an open-source MQTT client library that will be used for this example.  It has been adapted to many languages, including Python.

This example will demonstrate how to publish data to a ThingSpeak channel using the Paho library.  The CPU and RAM usage of the Raspberry Pi will be measured and published to ThingSpeak every 20 seconds.

Setup

1) Create a new channel as shown in Collect Data in a New Channel with two fields named “CPU (%)” and “RAM (%)”.

2) Create an MQTT device by clicking Devices > MQTT at the top of the ThingSpeak page, then Add Device. When setting up the device and adding the new channel to its authorized list, click Download Credentials > Plain Text. For details, see Create a ThingSpeak MQTT Device. Use the saved credentials in the Code section below.

3) Download the Paho MQTT client library for Python. You can use the command line to install the libraries. If you are using Python 2, use this code:

sudo pip install paho-mqtt
sudo pip install psutil

If you use Python 3, use this code:

sudo pip3 install paho-mqtt
sudo pip3 install psutil

Publishing to ThingSpeak Example

This example shows how simple it can be to send data to ThingSpeak using MQTT.

Three MQTT connection methods are demonstrated in this example: TCP, WebSockets, and SSL. To use a conventional TCP socket on port 1883, set useUnsecuredTCP=True.  This connection method is the simplest and requires the least system resources.  To use WebSockets on port 80, set useUnsecuredWebsockets=True.  Websockets are helpful when the default MQTT port is blocked on the local network.  Finally, if encryption is required, set useSSLWebsockets=True.   Be sure only to set one of the connection method variables to True and leave the other two False.

Also, be sure to change the channel ID and API key to your channel’s settings.

# Publich to a ThingSpeak Channel Using MQTT
# 
# This is an example of publishing to multiple fields simultaneously.
# Connections over standard TCP, websocket or SSL are possible by setting
# the parameters below.
#
# CPU and RAM usage is collected every 20 seconds and published to a
# ThingSpeak channel using an MQTT Publish
#
# This example requires the Paho MQTT client package which
# is available at: http://eclipse.org/paho/clients/python
import paho.mqtt.publish as publish
import psutil
import string

# The ThingSpeak Channel ID.
# Replace <YOUR-CHANNEL-ID> with your channel ID.
channel_ID = "<YOUR-CHANNEL-ID>"

# The hostname of the ThingSpeak MQTT broker.
mqtt_host = "mqtt3.thingspeak.com"

# Your MQTT credentials for the device
mqtt_client_ID = "<YOUR-CLIENT-ID>"
mqtt_username  = "<YOUR-USERNAME>"
mqtt_password  = "<YOUR-MQTT-PASSWORD>"

t_transport = "websockets"
t_port = 80

# Create the topic string.
topic = "channels/" + channel_ID + "/publish"

while (True):

    # get the system performance data over 20 seconds.
    cpu_percent = psutil.cpu_percent(interval=20)
    ram_percent = psutil.virtual_memory().percent

    # build the payload string.
    payload = "field1=" + str(cpu_percent) + "&field2=" + str(ram_percent)

    # attempt to publish this data to the topic.
    try:
        print ("Writing Payload = ", payload," to host: ", mqtt_host, " clientID= ", mqtt_client_ID, " User ", mqtt_username, " PWD ", mqtt_password)
        publish.single(topic, payload, hostname=mqtt_host, transport=t_transport, port=t_port, client_id=mqtt_client_ID, auth={'username':mqtt_username,'password':mqtt_password})
    except (keyboardInterrupt)
        break
    except Exception as e:
        print (e) 

This site uses Akismet to reduce spam. Learn how your comment data is processed.