MQTT for Wearable Robotics: How HiveMQ Cloud Enabled an Interactive Fursuit

TL;DR:

A passionate robotics engineer in Brazil built a full-body interactive fursuit with real-time digital eyes, voice/soundboard, LEDs, and a QR-code control app. To keep all components responsive, even with spotty conventional Wi-Fi, he used MQTT as the lightweight, bi-directional messaging layer and ran the broker on HiveMQ Cloud (managed, reliable, and quick to set up). The result: stable real-time control, easy onboarding for spectators, and an architecture that can scale to more features without reworking infrastructure.

In essence, MQTT + HiveMQ Cloud enabled a complex wearable cosplay system to become a reliable, real-time, multi-device experience.

Building a Real-Time Interactive Fursuit with MQTT and HiveMQ Cloud

Hi there! My name is Mekhy. I am a robotics engineer and owner of a fursuits/cosplay studio in Brazil. Through the years, I have been fascinated by full-body mascot costumes and wearable animatronics. These are used in controlled environments, such as movies/television shows and other forms of scripted performances, but not easily available for the general public to use. Here’s how Mekhy looks.

So, I set a goal to create an expressive cosplay costume (in this case a Fursuit of my original character) that I could use at events, using only consumer-grade electronics and a very easy-to-use control protocol. Cosplay is fundamentally about expression; turning imagination into something visible, emotional, and interactive. The goal of building more expressive costumes is simply the next step of that same idea: not just looking like a character, but letting the costume react in real time.

How I Built the Interactive Robot

The current version of the suit includes:

• Digital eyes rendered in real time

• Sound/voice capabilities using Spotify and Voicemod

• LEDs and additional wearable effects

• A personalized assistant

• A spectator-friendly control app people can open by scanning a QR code

Under the hood, this is a distributed system: multiple components need to communicate quickly and reliably, while the suit is moving through crowded convention spaces. Central to this architecture is a message bus and broker mechanism that ensures responsive and scalable communication across components, making MQTT an excellent fit for this and similar IoT-centric use cases.

Software Components

• Each part is open-sourced in its GitHub repository, allowing others to explore or adapt the system for their own creative projects.

• Mediapipe + OpenCV Python machine vision stack with custom scikit-learn model pipeline

• LiveKit agent using Silero for VAD, GPT 4o mini Transcribe for STT, GPT 4.1 Nano and Elevenlabs TTS

• Unity application using Live2D Cubism SDK for Unity (communication handled via websocket)

• Voicemod (communication handled via websocket using their control API)

• Spotify

• An MQTT broker running on HiveMQ Cloud

• Next.js control app, which runs on Vercel

• Self-hosted Telegram bot using the Telepota library

• C++ script running on AVR co-processor, which controls LEDs (communication handled via UART serial)

• AnyDesk

In summary, the current version of the suit utilizes a Lattepanda Delta 3 board for all processing and peripheral control. This board is installed inside a themed box located on the suit's back, and all peripheral components (HDMI displays, microphone, LEDs, etc.) are installed in a removable 3D-printed mask inside the head.

My own phone is positioned inside the box to act as a mobile network router. Software-wise, the control application is deployed on Vercel and is bound to a Telegram bot hosted by the suit. All actionable events, control intents, state changes, etc. are sent to an MQTT broker running on HiveMQ Cloud.

Image Source: https://www.lattepanda.com/blog-323319.html

You can check the build details in the following blog post by Lattepanda.

Why MQTT Was Ideal for My Wearable Robotics Project 

The Problem: Real-time Control with Unreliable Connectivity

Wearables used at conventions face a tough combination:

• Connectivity can be unpredictable (Wi-Fi congestion, signal drops, captive portals).

• The system is mobile and constantly changing network conditions.

• Control involves multiple clients (my phone, the suit, spectator devices).

• Interactions must feel instant because the lag ruins the magic.

I needed a protocol that was lightweight, bi-directional, supported multiple publishers, and designed for unreliable networks.

That’s where MQTT fit perfectly.

MQTT gave me exactly what I needed for a moving embedded system:

• Lightweight overhead (ideal for constrained or variable connections)

• Bi-directional communication (commands in, telemetry/state out)

• Multiple publishers/subscribers (spectators, control services, the suit itself)

• The ability to keep the experience stable even when connectivity is imperfect (a common reality at conventions)

In practical terms: MQTT lets the suit publish what’s available (controls, status, debug signals) while simultaneously receiving commands, without building a complicated custom backend.

Why I Used HiveMQ Cloud Instead of Self-Hosting

I could have self-hosted a broker, but I wanted to focus on the suit, not the infrastructure.

Unlike other MQTT Brokers like Mosquitto, HiveMQ offers a fully managed cloud broker in its free tier. It worked well for this kind of prototype-to-real-world pipeline. It offered:

• Fast setup with no complicated cloud setups. Just provide credentials and HiveMQ Cloud MQTT Broker handles it all.

• Complete with an easy-to-use web UI and excellent documentation

• A path to industry-grade reliability and scalability. It gives me the possibility to study and implement more complicated features in the future, without having to significantly change infrastructure.

Results

At the time of writing, Mekhy has been successfully utilized in 3 conventions with online connectivity available, and the feedback has been very positive. Community members who interacted with the control application were quite impressed by the ease of use and large amount of controllable parameters! This led to some crazy moments, especially when people 'discovered' the soundboard feature. This layer of interactivity feels functionally reliable yet full of surprises, perfect for the character and to begin creating content using this character. 

This project proves that such an art form and high-level computing can share a common link, and gave me the confidence to continue investing in future iterations and other similar endeavors.

Advice to the MQTT / IoT Community

As a piece of advice for the MQTT/IoT community: go crazy! IoT and M2M open doors for yet unexplored niches. If you are looking to add some magic to a project, use the tools at your disposal to turn your spectators into part of the show. It can take a long time to get it right, but when you do, all eyes will be on you.

Check out my project on Instagram: instagram.com/mekhy_w

Check out my Tiktok: tiktok.com/@mekhy_w