XIAO-S3 NASA Shuttle Orientation Dashboard

A high-performance IIoT visualization project utilizing the Seeed Studio XIAO ESP32-S3 and the MPU-9250 (9-DOF) sensor. This system provides real-time 3D orientation tracking (NASA Shuttle) in a web browser via WebSockets and Three.js.

Overview

• Real-time Sensor Fusion: Implements a custom Madgwick filter to convert raw IMU data into stable quaternions.
• Low-Latency Communication: Uses asynchronous WebSockets to stream orientation data at 60 Hz.
• Dual-Core Architecture: Core 1 runs sensor reading + fusion at 200 Hz; Core 0 handles WiFi AP + WebSocket broadcast.
• Decoupled Frontend: A Three.js-based dashboard hosted on LittleFS and served directly from the ESP32.
• Autonomous Power: Integrated LiPo battery management for a completely wireless 3D-printed model.

Technical Stack

• Hardware: Seeed Studio XIAO ESP32-S3 (Dual-core, Wi-Fi/BT, LiPo charging).
• Sensor: InvenSense MPU-9250 (I2C @ 400 kHz).
• Environment: PlatformIO / VS Code.
• Firmware: C++ with ESPAsyncWebServer and Bolder Flight Systems MPU9250.
• Frontend: HTML5, CSS3, JavaScript (Three.js r128), WebSockets.

Dependencies

Library	Version	Source	Notes
Bolder Flight Systems MPU9250	1.0.2	PlatformIO Registry	See Known Incompatibilities
ESPAsyncWebServer (lacamera fork)	3.1.0	GitHub	ESPHome-maintained fork, includes AsyncTCP-esphome 2.1.4
ArduinoJson	^7.0.0 (resolves 7.4.x)	PlatformIO Registry

Known Incompatibilities

Bolder Flight Systems MPU9250 v5.x (v5.3.0 - v5.6.0) is NOT compatible with Arduino/PlatformIO on ESP32. The v5.x line was rewritten for Bolder Flight's own build system and introduces the following breaking issues:

1. Eigen dependency conflict: The v5.x library depends on Bolder Flight Systems Eigen which defines a variable B1 in BDCSVD.h. This collides with Arduino's binary.h macro #define B1 1, causing compilation errors.
2. Undeclared transitive dependencies: The v5.x mpu9250.cpp requires units.h (another Bolder Flight library) which is not declared in library.properties, so PlatformIO cannot resolve it automatically.
3. Namespace change: v5.x uses bfs::Mpu9250 with a completely different API (Begin(), Read(), ConfigAccelRange()) compared to v1.x (begin(), readSensor(), setAccelRange()).

The v1.0.2 API is stable, lightweight (no Eigen), and fully compatible with ESP32 Arduino framework.

ESPAsyncWebServer (me-no-dev original) is abandoned. The original me-no-dev/ESPAsyncWebServer@1.2.4 and me-no-dev/AsyncTCP@1.1.1 are no longer maintained. We use the lacamera/ESPAsyncWebServer fork (maintained by the ESPHome team), which bundles AsyncTCP-esphome as a dependency automatically.

Hardware Wiring (I2C)

XIAO ESP32-S3	MPU-9250 Pin	Description
3.3V	VCC	Power Supply
GND	GND	Common Ground
D4 (GPIO 4)	SDA	I2C Serial Data
D5 (GPIO 5)	SCL	I2C Serial Clock
B+ (Pad)	LiPo Red (+)	Battery Positive
B- (Pad)	LiPo Black (-)	Battery Negative

Project Constants

Constant	Value
Board	seeed_xiao_esp32s3
SDA / SCL	GPIO 4 / GPIO 5
I2C Frequency	400 kHz
MPU-9250 Address	0x68
WiFi Mode	AP — SSID NASA-Shuttle-IMU, pass 12345678
AP IP	192.168.4.1
WebSocket Endpoint	ws://192.168.4.1/ws
Sensor Loop	200 Hz (Core 1)
WS Push Rate	60 Hz (Core 0)
Madgwick Beta	0.1
Filesystem	LittleFS

Project Structure

platformio.ini          # Board config, LittleFS, dependencies
src/main.cpp            # WiFi AP, dual-core tasks, WebSocket server
src/sensor_fusion.h     # Custom Madgwick filter (header-only, 6-DOF)
data/index.html         # Dashboard HTML shell
data/app.js             # Three.js scene, WS client, rolling charts
data/style.css          # Dark theme styling

System Architecture

graph TD
    subgraph Physical_Layer ["Layer 1: Physical Hardware & Power"]
        direction TB
        BATT["3.7V LiPo Battery"]
        XIAO["XIAO ESP32-S3"]
        MPU["MPU-9250 IMU Sensor"]

        BATT -- "Positive (+) Red" --> BPAD["B+ Pad (Bottom of XIAO)"]
        BATT -- "Negative (-) Black" --> BNAD["B- Pad (Bottom of XIAO)"]

        XIAO -- "3.3V (VCC)" --> MPU
        XIAO -- "GND" --> MPU
        XIAO -- "GPIO 4 (D4 / SDA)" --> MPU
        XIAO -- "GPIO 5 (D5 / SCL)" --> MPU
    end

    subgraph Processing_Layer ["Layer 2: Firmware & Data Processing"]
        direction LR
        LFS[(LittleFS: HTML/JS/CSS)]
        Fusion[Madgwick Filter: 6-DOF]
        WSS[Async WebSocket Server]

        XIAO --> LFS
        MPU -- "Raw Accel/Gyro Data" --> Fusion
        Fusion -- "Quaternions (w,x,y,z)" --> WSS
    end

    subgraph Visualization_Layer ["Layer 3: Web Frontend"]
        Browser["Web Browser"]
        ThreeJS["Three.js Engine (CDN r128)"]
        Model3D["3D NASA Shuttle Model"]
        Charts["Rolling Charts (Canvas 2D)"]

        WSS -- "JSON @ 60 Hz" --> Browser
        Browser --> ThreeJS
        ThreeJS --> Model3D
        Browser --> Charts
    end

Loading

Build & Upload

# Add PlatformIO to PATH (add to ~/.zshrc for permanence)
export PATH="$HOME/.platformio/penv/bin:$PATH"

# Build firmware
pio run

# Upload firmware + filesystem
pio run -t upload && pio run -t uploadfs

# Monitor serial output
pio device monitor

Usage

1. Power up the ESP32-S3 with the MPU-9250 connected
2. Connect to WiFi network NASA-Shuttle-IMU (password: 12345678)
3. Open http://192.168.4.1 in a browser
4. The 3D shuttle model rotates in real-time matching the physical sensor orientation

Performance Budget

Resource	Usage
RAM	13.5% (44 KB / 320 KB)
Flash	24.5% (818 KB / 3.3 MB)
LittleFS	< 20 KB (index.html + app.js + style.css)
WebSocket bandwidth	~7.2 KB/s (120 bytes x 60 Hz)
Core 1 CPU	~5% (sensor + Madgwick @ 200 Hz)
Core 0 CPU	~15-20% (WiFi + JSON + WS)