Project Caretaker. Part 1. Beginning

Project Caretaker - a remotely controlled robot on a tracked platform with camera streaming.

Why do I need it? For fun - this project is practically the quintessence of my hobbies - electronics, 3D printing, microcontroller programming, and generally everything that can be included in the DIY sphere. But if I were to come up with a practical application - it would be convenient when a leak sensor is triggered to be able to remotely check what's happening before rushing from work/travel (there have been several false-positive cases).

Then

I conceived this project, dare I say, 10 years ago. I already had a repository on GitHub with a description of what I wanted and even some implementation. I believe this was my first GitHub repository, which I maintained through the website without any git knowledge. Back then it was an Arduino Uno connected to a motor driver with 2 ultrasonic rangefinders, without a camera. I didn't have the energy or resources to fully implement what I had in mind, so the project was put on hold indefinitely.

This implementation could move forward and, based on the rangefinder readings, try to avoid obstacles. If both sensors showed the same readings, the robot would back up and turn at a random angle.

Now

But 10 years have passed, and a bunch of new possibilities and devices have emerged. I got a 3D printer (even 3), mastered printing with various materials, gained programming experience, smart home technologies, and all that.

Therefore, the new plan for the project looks as follows:

• esp32 cam as a controller and video streaming device;
• custom firmware with WiFi control and video transmission, flashlight, 2 control options, adaptive layout and video quality switching, possibly even connecting an app (Radiomaster TX12) via elrs;
• custom device design with AI;
• self-developed chassis - 2WD, possibly 4WD (technically even 6WD could be connected, but that would be pointless);
• fully 3D-printed body and TPU tracks;
• charging station that can be remotely approached and docked;
• plugin for Home Assistant integration.

Everything is quite realistic and even half-implemented already. At least the custom chassis, design, and basic firmware version are ready.

I'll reveal some details, but without too many spoilers.

Firmware

Written in C with the help of a large number of AI models (I tried almost all the top models because implementing video streaming on esp32 cam turned out to be an almost impossible task).

A detailed description of the firmware is available on GitHub, but in brief:

• control via virtual joystick or two sliders (since we have a tracked platform);
• SD/HD video transmission modes, SD for maximum performance, HD for image quality (FPS may drop);
• LED control - 3 positions 0%, 50%, 100% brightness - the built-in LED in esp32cam is quite bright;
• display of controller core load statistics (yes, esp32 has 2 cores) by various tasks; I wrote a separate article about the epic journey of enabling these functions in esp32 and platformio.

This isn't everything I plan to do in the project, but it's enough for an MVP.

Hardware

For now, I'll show a simplified diagram based on the same Rover 5 platform:

What we have here:

• DFRobot Rover 5 - a tracked platform purchased about 10 years ago from @amperka;
• 3x 18650 batteries with holders;
• L298N 2-channel motor driver;
• ESP32-CAM (you can also find this board under the name AI Thinker);
• regular switch for breaking the power circuit;
• DC-DC step-down converter, you can use a universal one with trimmer resistors - 2 pieces (12.6v > 9v, 12.6v > 5v).

Demo

Flash the firmware, open the device's IP address in a browser, and control it:

Video demo from a phone: