Jonathan Mendenhall
Published © GPL3+

Automatic Road Quality Detector

This device automatically detects and uploads the location of road hazards, creating a safer driving environment!

IntermediateFull instructions provided3 hours1,250
Automatic Road Quality Detector

Things used in this project

Hardware components

Particle Argon
Interfaces with the MPU6050, GPS and EEPROM chip to automatically detect significant bumps in the road as the user is driving.
Inertial Measurement Unit (IMU) (6 deg of freedom)
Inertial Measurement Unit (IMU) (6 deg of freedom)
Spikes in the Z axis acceleration signify a possibly hazardous bump in the road.
GPS Module
Any NMEA GPS module is suitable. This is the one I used.
AT24C256 EEPROM Module
Stores the GPS position of each bump detected in the road as the user is driving. After the next startup, data from the EEPROM module is read and any remaining entries are uploaded to the central database.
Breadboard (generic)
Breadboard (generic)
All components will be plugged into the breadboard.
Jumper wires (generic)
Jumper wires (generic)
All jumper wires used are Male-Male.
3.7v LiPo Battery
Any 3.7v LiPo battery is suitable, but a smaller capacity will limit the lifespan of the device.
JST PH 2.0 cables
I happened to have my batteries already but the connector was not the same as what the Particle Argon uses, so I needed these to make an adapter cable. You may not need these if you have a LiPo battery with a JST PH 2.0 connector already.
Double Sided Adhesive
This will be used to firmly mount the bread board and GPS module in the enclosure, and to mount the device firmly to the car.

Software apps and online services

Google Firebase
Particle Build Web IDE
Particle Build Web IDE

Hand tools and fabrication machines

Soldering iron (generic)
Soldering iron (generic)
Might be necessary if making an adapter cable for the LiPo battery to the Particle Argon LiPo connector


Read more


IoT Device

New Fritzing parts for the Particle Argon are not available yet, so I was unable to make a schematic in Fritzing. This should be sufficient though.


Particle Argon Code

This can be used in the Particle Build IDE to program the Particle Argon
#include <MPU6050.h>
#include "NMEAGPS.h"

#define EEPROM_ADDR 0x50

// fill these values in with the network address of the computer running the Python script

byte[] PYTHON_SERVER = {192, 168, 1, 18};

// Causes the Particle device to start the code immediately rather than waiting for a WiFi connection


// holds the data for each sensor reading that will be stored in EEPROM

typedef struct {
    int32_t t;
    int32_t lat;
    int32_t lng;
    float accz;
} sensorReading_t;

// local variables used for reading data from the MPU6050 and GPS

MPU6050 mpu;
gps_fix fix;
sensorReading_t reading;

// variables used to maintain a rolling mean of the sensor data used as a baseline for detecting spikes in the acceleration data

#define ACCZ_THRESHOLD 0.17 // 0.17G's

int32_t avgAZ;
uint32_t lastAvgAZ;
uint32_t avgAZSamples;
uint16_t numStoredRecords;
int32_t circleBuf[ROLLING_WINDOW];
uint32_t circleBufIndex;
int32_t circleBufSum;

bool detectedBump;
float detectedAccZ;

// these LED status modes can be activated to display when the device is running or when it detected a bump


// I2C functions for writing data buffers to the EEPROM module

void writeEEPROM(uint16_t addr, uint8_t data ) {
    Wire.write(addr >> 8);   // MSB
    Wire.write(addr & 0xff); // LSB

void writeEEPROM(uint16_t addr, uint8_t* buffer, uint8_t length ) {
    Wire.write(addr >> 8);   // MSB
    Wire.write(addr & 0xff); // LSB
    Wire.write(buffer, length);
uint8_t readEEPROM(uint16_t addr) {
    Wire.write(addr >> 8);   // MSB
    Wire.write(addr & 0xff); // LSB
    Wire.requestFrom(EEPROM_ADDR, 1);
    if (Wire.available()) 
    return 0xff;

uint8_t readEEPROM(uint16_t addr, uint8_t* buffer, uint8_t length) {
    Wire.write(addr >> 8);   // MSB
    Wire.write(addr & 0xff); // LSB
    uint8_t read = 0;
    Wire.requestFrom(EEPROM_ADDR, length);
    while(Wire.available()) {
        *buffer++ =;
    return read;

// the first 2 bytes in the EEPROM represent how many records there are in the remainder of the EEPROM

uint16_t getNumStoredRecords() {
    uint16_t num;
    readEEPROM(0, (uint8_t*)&num, 2);
    return num;

void setNumStoredRecords(uint16_t num) {
    writeEEPROM(0, (uint8_t*)&num, 2);

// this function will attempt a connection to the device's known WiFi network
// if it connects successfully, it will attempt to communicate with the Python script to upload its records to the Firebase Database

void tryConnectAndUpload() {
    // try to connect to the WiFi network
    long t0 = millis();
    while(!WiFi.ready() && millis() - t0 < 30000); // wait for a connection or timeout after 30s
    if(!WiFi.ready()) {
    } else {
        TCPClient client;
        Serial.println("Connecting to batcher");
        // try to connect the the Python server based on the address at the top of this code
        if(client.connect(PYTHON_SERVER, 5000)) {
            client.print(System.deviceID());                                        // send the device ID to Python
            client.write((uint8_t*)&numStoredRecords, 2);                           // send the number of records to Python
            // send the binary data for the records to Python
            for(uint16_t i = 0; i < numStoredRecords; i++) {
                readEEPROM((1 + i) * 16, (uint8_t*)&reading, 16); // read each record from the EEPROM module into the 'reading' variable which is a sensorReading_t struct
                NeoGPS::time_t t = (NeoGPS::clock_t)reading.t;
                String data = String::format("{\"t\": \"%d-%02d-%02dT%02d:%02d:%02dZ\", \"lat\": %f, \"lng\": %f, \"accz\": %f}", t.year+2000, t.month,, t.hours, t.minutes, t.seconds, / 1.0e7, reading.lng / 1.0e7, reading.accz);
                // send the raw byte data
                client.write((uint8_t*)&reading, 16);
        } else {
            Serial.println("ERROR"); // the device was unable to connect to the Python server
        numStoredRecords = 0;
        Serial.println("Cleared all records");

    // turn WiFi off to conserve battery power;
    Serial.println("WiFi off");


void setup() {
    // begin communication with Serial and I2C
    Serial1.begin(9600);            // start Serial connection to NMEA GPS module

    // initialize the MPU6050 module
    Serial.println("Initializing MPU6050");
    Serial.println(mpu.testConnection() ? "MPU6050 connection successful" : "MPU6050 connection failed");
    // check how many records are stored in EEPROM
    numStoredRecords = getNumStoredRecords();
    Serial.print("Stored records: ");
    if(numStoredRecords > 0) {
        tryConnectAndUpload();                                  // try to upload the records if there are any
    statusNormal.setActive(true);              // turn off the status led by setting statusNormal to active
    lastAvgAZ = millis();

void loop() {
    // try to read data from the GPS module
    while(gps.available(Serial1)) {
        fix =;
        // if a bump was detected since the last reading, save the time and location to the EEPROM module
        if(detectedBump) {
            reading.t = (time_t)fix.dateTime;
   = fix.latitudeL();
            reading.lng = fix.longitudeL();
            reading.accz = detectedAccZ;
            // reset detection variables for the rolling mean comparison
            detectedBump = false;
            detectedAccZ = 0;
            Serial.print(reading.t); Serial.print(' ');
            Serial.print( / 1e7); Serial.print(' ');
            Serial.print(reading.lng / 1e7); Serial.print(' ');
            Serial.print(reading.accz); Serial.print('\n');
            // write the buffer to EEPROM and increase the counter for the number of records
            writeEEPROM((1 + numStoredRecords) * 16, (uint8_t*)&reading, 16);
            // flash the LED to signal a bump was stored
    // add the current acceleration to the total to get an average acceleration every 20ms
    avgAZ += mpu.getAccelerationZ();
    // use the average acceleration every 20ms
    if(millis() - lastAvgAZ >= 20) {
        lastAvgAZ = millis();
        avgAZ /= avgAZSamples;
        // if the circle buffer is full, compare the current average acceleration with the current rolling mean of the data
        if(circleBufIndex >= ROLLING_WINDOW) {
            float relAmpAccZ = abs(avgAZ * ROLLING_WINDOW - circleBufSum) / 16384.0 / ROLLING_WINDOW;         // computationally cheap way to compute the rolling mean of the acceleration data
            // if the amplitude of the current acceleration is beyond a threshold away from the rolling mean, a bump was detected 
            if(relAmpAccZ > ACCZ_THRESHOLD && relAmpAccZ > detectedAccZ) {
                detectedAccZ = relAmpAccZ;
                detectedBump = true;
            // subtracts the last element of the circular buffer from the current sum of the data to keep computation speed fast
            circleBufSum -= circleBuf[circleBufIndex % ROLLING_WINDOW];
        // append the average Z axis acceleration to the circle buffer for maintaining a rolling window of the baseline acceleration data
        circleBuf[circleBufIndex++ % ROLLING_WINDOW] = avgAZ;
        circleBufSum += avgAZ;

        // reset the average acceleration data for every 20ms
        avgAZ = 0;
        avgAZSamples = 0;

This code runs on a host computer (Laptop, Desktop, or even Raspberry Pi) that is connected to the same WiFi network as the Particle Argon. It will receive sensor readings from the Argon and create new records in a Firebase Firestore Database so the Particle Argon does not have to deal with complicated HTTPS Rest API's
import time
import socket
import struct
import datetime
from import firestore
import firebase_admin
from firebase_admin import credentials
from firebase_admin import firestore

# load the firebase credentials from a local file

cred = credentials.Certificate('firebaseAdminCredentials.json')

# get a connection to the firebase firestore database

db = firestore.client()

# create a TCP socket that will listen for a connection from the Particle Argon

sock = socket.socket(socket.AF_INET, socket.SOCK_STREAM)
sock.setsockopt(socket.SOL_SOCKET, socket.SO_REUSEADDR, 1)
sock.bind(('', 5000))

# begin listening and waiting for a connection

while True:
		while True:
			print('waiting for connection')
			conn, addr = sock.accept()

			# read the deviceID and number of records from the Argon

			deviceID = conn.recv(24).decode('utf-8')
			numReadings = struct.unpack('<H', conn.recv(2))[0]
			print(f'deviceID: {deviceID}, readings: {numReadings}')
			# read all of the binary data for the sensorReading packets

			readings = []
			for i in range(numReadings):

				# each packet is 16 bytes long
				buf = conn.recv(16)

				# unpack the values into time, latitude, longitude, and accelerationZ
				t, lat, lng, accz = struct.unpack('<iiif', buf)
				t = datetime.datetime.fromtimestamp(t + 946684800)	# the GPS library on the Particle Argon gives time in seconds since 2000 rather than typical unix time since 1970
				lat /= 1e7
				lng /= 1e7

				# add all of the received readings to a list to be uploaded later
				print(t.strftime('%Y-%m-%dT%H:%M:%SZ'), lat, lng, accz)
				readings.append((t, lat, lng, accz))
			# close the connection so the Argon is not waiting for Python to upload all of the readings

			# add all of the records in a batch to make upload speed faster
			batch = db.batch()
			for reading in readings:
				t, lat, lng, accz = reading
				datapoint_ref = db.collection('datapoints').document()		# get the reference to a new document in the 'datapoints' collection
				# set the contents of the document to the data from the reading
				batch.set(datapoint_ref, {
					'device': deviceID,
					'time': t,
					'location': firestore.GeoPoint(lat, lng),
					'acceleration': accz

			# commit the batch to start the upload to the Firestore Database

	except KeyboardInterrupt: # pressing Ctrl+C (Linux/MacOS) or Ctrl+Break (Windows) will close the program if necessary


Jonathan Mendenhall

Jonathan Mendenhall

2 projects • 17 followers
I've been working with hardware and software for 5 years, and I have 3 years of professional software development.


Add projectSign up / Login