Temperature sensor with Raspberry Pi 3 and AWS

This tutorial describes how to set-up the communication between Raspberry Pi 3 and AWS cloud. As a concrete example, we will store temperature data tin the cloud and visualize it. This is accomplished in several steps:

  1. Connect Raspberry Pi 3 to AWS
  2. Read temperature from I2C sensor
  3. Store data time-series in DynamoDB
  4. Visualize the historical temperature data
  5. Receive alerts on low and high temperature

Connecting RPi 3 to AWS

As a starting point, we have the Raspberry Pi 3 board with Raspbian Linux  and fully configured WiFi connection. Let’s connect it to the Amazon IoT!

This process can be split up into two parts – set-up of security credentials and SDK installation. During the security credentials set-up, it does not matter which SDK you prefer. Amazon has pretty detailed description of how to connect RPi for Javascript SDK, but we can for the large part follow the same steps. Follow this AWS manual to

  • Create and Attach a Thing (Device)
  • Generate certificates and policy
  • Download certificates

Once all the certificates (i.e. device certificate, private key, and root CA certificate) are downloaded from AWS, let’s upload them into $ ~/certs folder on the device.

Now, we will install Python SDK rather than NodeJS. Simply following the instructions here and executing $pip3 install AWSIoTPythonSDK

Let’s also clone the Git repository so we can use its sample code to learn the SDK basics:

$ git clone https://github.com/aws/aws-iot-device-sdk-python.git
$ cd aws-iot-device-sdk-python/samples
$ ls 

We’ll take advantage of the basicPubSub.py script. It should be called together with the certificates (so that AWS IoT can identify the device):

# Certificate based mutual authentication
$ python3 basicPubSub.py -e <endpoint> -r <rootCAFilePath> -c <certFilePath> -k <privateKeyFilePath>

This sample code will subscibe to sdk/test/Python topic and publish to the same topic in the infinite loop. It is possible to use the external MQTT client (or the on built-into AWS IoT ) to see these messages and validate that the connection with AWS IoT is successful.

We are going to use the connection established in this way to publish ambient temperature to the appropriate topic. But first, we need to read the sensor data.

Reading Temperature Sensor…

I am using an affordable and easy to set-up TMP102 sensor from SparkFun, connected to the board using Pi Wedge. The connections are straightforward, as there are only four wires: SDA, SCL, +3.3V and GND.

Raspberry Pi with TMP 102
Raspberry Pi with TMP102 – I2C connections

Now let’s make sure that the I2C is enabled by typing  $ sudo raspi-config and going to the advanced options menu. If I2C is currently off,  enabling it requires reboot of RPi board. We also need to install i2c-tools by executing  $ sudo apt-get install -y i2c-tools  to be able to communicate through I2C. Let’s see what’s connected by executing i2cdetect command:

$ i2cdetect -y 1
  0  1  2  3  4  5  6  7  8  9  a  b  c  d  e  f
00:          -- -- -- -- -- -- -- -- -- -- -- -- --
10: -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- --
20: -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- --
30: -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- --
40: -- -- -- -- -- -- -- -- 48 -- -- -- -- -- -- --
50: -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- --
60: -- -- -- -- -- -- -- -- -- -- -- -- -- -- -- --
70: -- -- -- -- -- -- -- --

We have the TMP102 sensor at its standard address 0x48. The full temperature reading is stored in the first 12 bits with address 0. The first 8 bits provide  to get the rounded temperature reading:

$ i2cget -y 1 0x48 0 b

The returned result 0x19 = 25 C is the room temperature as measured by the sensor. The full 12 but reading will be used later to get more accurate data. For now,  we are assured that the sensor is connected and works. All we need is a python script that will read the temperature and send it to AWS.

…with Python

Let’s use pigpio library to communicate with I2C – seems like it comes pre-installed with Python3. Let’s also install the pyth0n-smbus library: $ sudo apt-get install python-smbus.

First, start a pigpio daemon by executing  $ sudo pigpiod.  After that, the script should acquire and print the temperature. We can start it by executing $ python tmp102.py The source code  of tmp102.py is below:

#reading TMP 102 sensor

import pigpio
import time

def tmp102_reading(byte0, word0):
 #calculation of the temperature based on 
 #the first word and the first byte
 # !!! not tested for negative temperatures !!!!
 #last 4 bits of the word0
 l4b = (word0 & 0b1111000000000000)>>12
 temperature = ((byte0<<4) | l4b) * 0.0625
 return temperature

#i2c bus of the Raspberry Pi 3
i2c_bus = 1
#TMP 102 address on the i2c bus
addr = 0x48

dev_pi = pigpio.pi()
dev_tmp = dev_pi.i2c_open(i2c_bus, addr, 0)
register_n = 0
 while True:
 t_byte = dev_pi.i2c_read_byte_data(dev_tmp, 0)
 t_word = dev_pi.i2c_read_word_data(dev_tmp, 0)
 t = tmp102_reading(t_byte, t_word)
 print(' Temperature: {} C'.format(t))
except KeyboardInterrupt:
print('Exiting the loop');
r = dev_pi.i2c_close(dev_tmp)

This script will output the sensor reading every second until Ctrl+C is pressed:

Temperature: 25.59375 C
Temperature: 25.59375 C
Temperature: 25.625 C
Temperature: 25.625 C
Temperature: 25.625 C
Temperature: 25.625 C
Temperature: 25.625 C

Sending data to AWS

Instead of printing the temperature in the console, we need to send it to AWS IoT and repeat the whole thing after a set interval – every minute, for example. The Raspberry Pi has already been connected to AWS during the first step. We will re-use the publish/subscribe code example (basucPubSub.py) from the AWS SDK example to publish to a new topic for our new sensor.

First, let’s create variables for the topic and delay between temperature updates. Plus, we can give our sensor a serial number – in case we build more than one and need to distinguish between them:

delay_s = 60
sensor_sn = '00000001'
topic = 'myrpi/'+sensor_sn

Let’s now read the temperature in the loop and send it to AWS (until Ctrl+C):

  while True:
    loopCount += 1
    t_byte = dev_pi.i2c_read_byte_data(dev_tmp, 0)
    t_word = dev_pi.i2c_read_word_data(dev_tmp, 0)
    t = tmp102_reading(t_byte, t_word)
    timestamp = datetime.datetime.now()
    print(' Temperature: {} C   Loop # {:d}'.format(t,loopCount))
    print(' Time: {} \n'.format(timestamp))
    msg = '"Device": "{:s}", "Temperature": "{}", "Loop": "{}"'.format(sensor_sn, t,loopCount)
    msg = '{'+msg+'}'
    myAWSIoTMQTTClient.publish(topic, msg, 1)
except KeyboardInterrupt:
print('Exiting the loop');
r = dev_pi.i2c_close(dev_tmp)
print('Disconnected from AWS')

Remember – this(get the entire file here)  is only a (slight) modification of the basicPubSub.py file provided in AWS Python SDK. Path to certificated is still necessary for it to run properly. The shell command to start publishing to AWS is therefore

$ python3 publis_temp.py -e <Your:AWS:endpoint> -r <rootCAFilePath> -c <certFilePath> -k <privateKeyFilePath>

The script publishes the following payload to the myrpi/00000001 topic:

 "Device": "00000001", 
  "Temperature": "27.0",
  "Loop": "2670"


The structure of this payload is very similar to the data sent by  the AWS  IoT button. Once the data is published we can time-stamp and log it the DynamoDB table. This was described in detail in the post about IoT button.

Getting started with Amazon IoT button (Part II)

Amazon IoT button
AWS IoT button

In the first part of this Getting started with IoT button post, we configured the WiFi connection and set up certificates for secure communication with AWS cloud. The data sent with every button press looks like that:

{"serialNumber": "G030JF05XXXXXXXX","batteryVoltage": "1592mV","clickType": "SINGLE"}

Besides sending, the device has to do operations with this data like publishing it on a MQTT topic. Since all actions in AWS are tightly controlled for security purposes, the device needs a permission to perform actions. Policy is a form of such permission.

Create a policy

Creating a policy starts with clicking Create a resource and choosing Create a policy tab.

Create AWS IoT policy
Create AWS IoT policy

In order to publish to a particular topic, select iot:Publish action (it is easy to find it in the drop-down list after typing the first few letters). The content of the resource field depend on the selected action.  Since we selected Publish, we need to provide topic identifier as a resource (read more about resources and topics here). Resources follow somewhat complex naming convention:


This is similar to the REST API endpoint, since it contains your AWS account number (endpoint subdomain) and region (for example, us-east-1). We already used the serial number as well when setting up WiFi access. With all previously collected info, it is easy to create the correct ARN:


Check the Allow box, click on Add statement  (single policy can contain multiple statements) and Create.  The policy is now visible in the list of resources.

Attach the policy and the thing to the device certificate

In order to work properly, the newly created policy has to be attached to the device certificate. Also, the certificate should be associated with the thing we created in the first part. To perform this association, select the certificate in the resource list and click on Actions menu. Attach both the policy and the thing to the certificate. The certificate is finally complete!

Attach policy & thing to the certificate
Attaching the policy & thing to the certificate
IoT policy properties
IoT policy properties after attachments








If you press a button on the device at this point, it will send the data and publish it to the iotbutton topic. LED indicator should turn solid green for several seconds, indication successful publish outcome. As a test, try to detach the policy – the LED will flash red. Also try detaching the thing. Surprisingly, the thing really does not make a difference – even without it the message is published.

You can subscribe to a topic and receive messages by using the MQTT client built into AWS IoT or with the standalone application like MQTT.fx (more MQTT clients here). Subscribe to the iotbutton/G030JF05XXXXXXXX topic to view messages for a specific device or iotbutton/+ if you have multiple buttons.

Create and configure a rule

AWS IoT can perform actions when a messages is published through the use of rules. Creating a rule is similar to creating any other resource. In the Create rule dialog, fill in the name and description fields first. We need to create an SQL query that will be used to monitor published messages. This is done by specifying the Attribute field and the Topic filter. Since we are interested in the IoT button topic, type in the familiar iotbutton/G030JF05XXXXXXXX. Attribute can be a specific field of the payload, such as clickType or batteryVoltage, but we can also use  for all fields. Condition is not required and can be left blank, unless we want to trigger the rule only when battery voltage falls below a certain threshold, for example.


Create a rule dialog
Create a rule dialog

Finally, we need to select an action from the expansive list of available actions. We are currently interested in the sending a text message or an email as the click notification, so select the SNS service.

Select action from AWS services
Select an action from the list of AWS services

We have not selected any targets for the notification. Let’s follow Create a new resource link to create a target for the AWS Push Notification Service (SNS). Targets for the rules are called topics in the SNS parlance.

Create SNS topic

Hit Create new topic button on the newly opened SNS dashboard and input name and display name of your topic.

Create SNS topic
Create SNS topic

Create the topic and click on its ARN in the list of topics. This should display topic details. As you see, the topic currently has no subscriptions – i.e. addresses or phone numbers to send notifications to. You can add a number or an email address by creating new subscription.

Creating new SNS subscription
Creating new SNS subscription

This process is straightforward, as Topic ARN field gets auto-populated, all you need to do is to choose the desired protocol (SMS, email or AWS Lambda function to name a few), enter the endpoint (cell number, email address, etc) and create the subscription. With the subscription and topic in place, lets go back to the IoT dashboard and continue with the rule creation.

Create a role
Create a role

SNS target is now the topic name. Message format field is not required and the only remaining thing to do is to specify the Role name. Click Create a new role  and specify role name – this is all. AWS will automatically add a new role into the Identity and Access Management (IAM) system and grant it a permission to push IoT notifications. With the role in place, Add action and Create  the rule. Notice that a rule can have multiple assigned actions.

Rules are independent of a particular device or certificate. They do not have to be attached to anything. AWS IoT rules engine will continuously monitor the messages and push notifications if the message matches SQL query specified in the rule.

Press the IoT button and receive text message

With all steps now complete, pressing a button sends a text message withing a few seconds. There are three distinct click types: single, double and long. In addition, the button sends voltage, which can be monitored over time. Counting clicks is another interesting project for learning AWS basics.

Final SMS notification
Final SMS notification from the IoT button

Continue reading  Amazon Web Services IoT and DynamoDB tutorials:

Storing and visualizing IoT button data

Connecting MQTT.fx to AWS IoT