1. Required hardware

(click on each section for more information and alternatives):

1.1 Components:

• 1x ESP-32 or similar
• 1x L298N motor driver
• 1x 7.4v 2s battery pack
• 1x MPU-6050 – throughout the guide we will call this the IMU (inertial measurement unit)
• 2x 5v DC gear motors (with wheels)
• Jumper wires (lots)
• 2x 0.96″ OLED displays
• 1x Perfboard (or any other connection medium)

1.2 Fasteners:

• 4x ~30mm M3 bolts + nuts (for motors)
• 8-10x ~10-15mm M3 bolts + nuts (for attaching components)
• 12x ~15mm M3 self-tapping wood screws (for attaching 3D printed parts)

1.3 Equipment:

• 4x ~30mm M3 bolts + nuts (for motors)
• 8-10x ~10-15mm M3 bolts + nuts (for attaching components)
• 12x ~15mm M3 self-tapping wood screws (for attaching 3D printed parts)

2. Download the project files

We have provided all of the project files in one organised .ZIP file.
Download the correct file corresponding to your use of OLED’s

Download the correct file set variant:

Extract this file somewhere sensible on your computer. We will reference it throughout the guide.

3. Testing the circuit

Before we begin to build anything, It is always a good idea to test the electronics beforehand. This is where a breadboard can be useful. You should build a circuit according to one of the schematics and instruction sets bellow, on a breadboard or any other temporary connection medium. Insure that you select the correct wiring diagram for whether you want to use OLED’s or not.

Wiring schematic and instructions – WITH OLED’s

This wiring schematic outlines the connections that must be made between components. This schematic is for use WITH OLED’s. Please ensure you upload the correct code.

Important information:

Both the OLED’s and MPU-6050 will communicate via an I²C connection protocol, this helps reduce wiring complexity but creates a few issues when processing data. Our main issue is that the ESP-32 needs to be able to communicate over 30 frames per second to multiple OLED displays while simultaneously reading the tilt angle from the IMU over 1000 times a second.

On a standard, single-bus I²C setup all of this traffic would share the same physical wire connections. In practice, we found that combining these components together introduced huge amounts of latency causing the OLED’s to become de-synchronised and uneven; the IMU readings inconsistent; and jittery robot balancing.

To solve these issues, we use a multi-bus I²C setup operating simultaneously (something many Arduinos do not support). We also run both OLED displays on the same address so the ESP-32 treats them as one component.

One drawback of this system is that we need to manually solder pullup 4.7kΩ resistors to the MPU-6050 bridging the SCL and SDA lines with VCC.

Wiring schematic and instructions – WITHOUT OLED’s

This wiring schematic outlines the connections that must be made between components. This schematic is for use WITHOUT OLED’s. Please ensure you upload the correct code.

---

3.1 Testing the circuit

Inside the project file you downloaded earlier you should see a “Without web server” folder. Inside that folder you will find the folder containing the correct sketch for your OLED variant.

It should be labelled either:

Code_NOwebserver_NOoleds or Code_NOwebserver_WITHoleds

you should then upload that code to your ESP-32.

You will also need to download the following library’s using the Arduino IDE built-in library manager:

Adafruit MPU6050

Adafruit Sensor

If your using OLED’s you will also need:

Adafruit SSD1306

Adafruit GFX

These can all be downloaded from the library manager menu in Arduino IDE, but I have provided links to the corresponding GitHub pages.
Also note, if you ARE using Oleds – download the RoboEyes library. It’s a free, open‑source library for OLED‑based robot eye animations and is essential for this project – you can’t get this one from the library manager.

if your not using OLEDs then you don’t need to worry about this

---

You should expect the motors to spin into action and change speed/direction as you rotate the IMU. If you are using OLED’s you should also expect both OLED’s to display one eye.

If you run into any issues or something doesn’t work, please leave a comment on the respective YouTube video and we will be happy to help you out in a timely manner.

4. Printing the chassis

Next, we need a chassis (frame) for our robot. You will need to 3D print the required quantity of each part

The project file you download earlier should contain a “3D Print files” folder. Within this folder, you should see the following:

You will need 1x:

• – Layer_1.stl
• – Layer_2.stl
• – Layer_3.stl
• – Layer_4.stl

You will need 2x:

• – Motor_Mount.stl

The print settings don’t particularly matter for this build but this is what should be used:

• Support: NO supports
• Infill: 20% gyroid
• Walls: 3 walls
• Layer height <0.2mm

5. Assembly

This is the fun part, we will now construct our robot, it’s pretty simple and self-explanatory but this guide will take you through the process

You should now gather all of the 3D printed parts, M3 nuts and bolts, M3 self-tapping screws.

---

5.1 Mounting the motors

1. 1. Add superglue into the rectangular divots on the bottom plate (1)
2. 2. Press the motor mounts (2) into place (I needed a hammer).

3. 3. Use ~30mm M3 bolts (3) to secure the motors
4. 4. Mount your battery in the centre – either with screws or glue.

---

5.2 Logic board

Next, make the main logic board for the robot. According to the correct schematic for your version (refer to section 2). You must leave connections for the L298N and OLEDs as empty pin headers so I could attach jumper wires later. When laying our your logic board, ensure the ESP-32’s USB port is accessible – we will need it later.

Here’s how mine looked. I used a 70mmx90mm perfboard.

(ignore the extra cables and USB‑C port — I broke the one on the ESP32)

I also used a TTGO T7 esp32 which is why it looks different, this was purely because I had it laying around. But the connections remain the same.

---

5.3 Power supply

Next, make the main logic board for the robot. According to the correct schematic for your version (refer to section 2). You must leave connections for the L298N and OLEDs as empty pin headers so I could attach jumper wires later

I also used a separate breadboard power supply

5.4 The Motor Driver

Attach the L298N to the correct plate (6) in the orientation shown using the shorter M3 screws. Add jumper wires and route them through the hole. If using OLEDs, glue them to the front using hot glue (7) and route their wires through the cable hole (8).

5.5 Final assembly

Now you’ve build your robot!

Don’t worry if it doesn’t balance at the moment, we’re going to get to that

6. PID Tuning

Now, we need to tune the PID values for the robot – this part is slightly less fun.

6.1 Upload the new code

Now you must upload the web-server enabled version of your code. Inside of the file you downloaded earlier there should be a folder labelled “With web server”.

Inside that folder, there should be a folder containing a sketch of your OLED variant. It should be called either of the following:

Code_WITHwebserver_WITHoleds or Code_WITHwebserver_NOoleds

Upload the correct variant to your ESP32 – ensure you have the correct library’s installed (note “3.1 – uploading the code“).

6.2 Check motor direction

Now you can power on your robot. We need to identify whether the motors are spinning in the correct direction or not. To do this you should just try and keep the robot balanced upright; if it is just slamming straight into the ground or doing other strange stuff then your motors are going the wrong direction – it will be fairly obvious.

To reverse the motor direction, you have two options, either:

Reverses the direction in which the respective motor wires are connected to the L298N (note “5.5 – final assembly):

Or, if both of your motors are spinning the same direction; just incorrect. Then you can reverse it in code:

---

6.3 Tune the robot

In your browser, open:

http://192.168.4.1

this will load the tuning interface. If this link doesn’t connect anywhere make sure your device is connected to the Balancing_Robot WIFI network

This interface allows for fast, real-time tuning and means we don’t have to upload the code every time we want to tweak a value. IMPORTANT: when the robot looses power, it will forget all of your PID tuning hard work so it is imperative that you either write down or screenshot important checkpoints/values that are working for you to return too and if you ever wish to return to defaults; just restart the robot.

Our robot uses a PID control loop. This is a simple yet affective method for achieving a desired equilibrium point. The downside is each of the three sections requires individual tuning.

For more information on which each section does, how it works and how we implement it into the robot, check the YouTube guide. (top of the page)

In short:

• Too much P: jittery and twitchy balancing
• Too little P: weak, slow to react and cannot catch itself
• Too much I: big delayed swings, oscillates (vibrates) when falling away
• Too little I: persistent lean, never fully upright
• Too much D: almost nervous-looking, reacts sharply to noise and disturbance
• Too little D: slow, wobble and most importantly – overshoots

---

Where to begin?

Find equilibrium:

Set the power multiplier to 0, then try and balance the robot at a perfect point without any motor aid. Read this angle from the web page and note it down. Write this value into the “setpoint” section.

Start with only P:

Set:

P = 20, I = 0, D = 0

Increase P until the robot starts oscillating. Then back off slightly –  this gives the backbone of the balancing.

Introduce a little D:

Increase D in small increments until the oscillations calm down. Stop as soon as it becomes smooth

Finally, add I:

Increase I in tiny increments just enough so it corrects slow drift. If it starts doing big, delayed swings then reduce I. I is the easiest way to ruin an otherwise good tune.

Other sections:

The landing page will host a variety of other options for tuning and tweaking, if you wish to experiment then feel free but you can safely leave them as-is.

What works for you works:

If your robot doesn’t need any of a certain term (usually I). Then don’t have it!

If your robot doesn’t have PID values at all similar to others. That’s fine!

Your robot will have specific values and tuning that work for you.

SAVE!!:

Once you have found the values. TAKE A PHOTO. Or note them down somewhere. We will need them in the next step.

7. Finishing touches

Now that we have found the optimal PID values it is good practice to hard-code them into the robot so we don’t have to start the webserver every time we turn on the robot.

---

Go back to the project files you downloaded earlier and open the Inside the “Without web server” folder. Inside that folder you will find the folder containing the correct sketch for your OLED variant.

It should be labelled either:

Code_NOwebserver_NOoleds or Code_NOwebserver_WITHoleds

you should then upload that code to your ESP-32. (exactly like in step 3.1).

At the top of the code you should see this section:

Hit upload and your done!