In a rapidly urbanizing world, the effects of air pollution, noise pollution, and traffic congestion on our populations will drastically increase. The first step to solving these problems is to understand them through data collection. Implementing a dynamic, extensible sensor network to cover a city is not an easy task. Designing a system applicable to almost any urban environment with varying levels of infrastructure is even more challenging.

One of the major limitations of IoT projects of this scale is in the name, the requirement of internet access. To overcome this problem, our design focuses on peer-to-peer communication. With each unit not depending on an available network, this system can be deployed more quickly, with less overhead, and for a lower cost, meaning more of the worlds cities can be supported.

The network will have a single master device that requires internet connectivity through an available access point, LTE connectivity, or direct connection to a local server to store and distribute the data produced. This device will handle storing the location and index of each new device on the network so that the data it receives can be easily analyzed and passed along through the proper channels.

To implement a robust, large IoT network, the complexity of deployment and maintenance needs to be minimized. Our goal is to have homogenous device software that requires no specific order of installation, fully automated network setup for each device, and no manual location logging. Deploying the network will be as easy as strapping the unit to a light pole, switching it on, and moving on to the next intersection.

When a device is initialized, it will first get its location from GPS, then search for neighboring devices. When it finds another unit, it will pass along its MAC address and location data to the network, which will relay the information to the master device on the network. The master device will then return a device ID, which will prepend all future data transmissions to increase legibility and simplicity of the network allowing for devices can easily be identified, located, and tracked.

For maintenance, each unit will have protocols in place to report errors with its sensors, network, or battery faults, which can then be easily passed on to allow for quick and painless replacement or repair. In the event a devices loses all functionality, the lack of data from it will trigger an error report. Replacing a unit will be as easy as installing a new one and turning it on, its neighbors will be smart enough to welcome the new device and include it in the current network.

The general plan for design is to have a single PCB with all sensors integrated for easy and cost-effective mass production. The sensors will include an electret microphone for audio sampling, a DHT22 or equivalent for temperature and humidity sensing, an SDS011 air quality sensor, and possibly other sensors for chemical pollution measurement, cost prohibiting. For wireless communication and sensor processing, we plan to use the popular ESP32 NodeMCU SoC. The reason for these selections was primarily the popularity of the devices. They all have well documented supporting libraries for easy integration into our project, which means the cost and time of development will be minimized, and they are all available in large supplies to easily scale the production.

The ESP32 is our current choice due to our experience with it. In our own tests, we were able to achieve peer-to-peer wifi ranges of around 150 meters using the ESP-Now protocol. One of the best features of the 32-bit dual-core MCU is the devices low power consumption of 2.5 µA in hibernation. All of this hardware will be powered with 18650 cells, maintained by a small solar panel topping each unit.

The network will have multiple layers of data, and each later can be independently visualized, managed, and distributed. Visualization of data will involve overlaying it on a map of the network, with how depending on what type of data it is. For example, temperature, humidity, and noise pollution could all be viewed as heat maps so anyone could easily identify what areas have cleaner air, more car horns, or other information. Microphones could be used to detect and pinpoint possible gunfire and report it to dispatch. Other layers will display network diagnostics as vectors that show the direction each unit sends data and the volume of data it sends.

These maps could be distributed through many different means, local news stations could share live data about the air quality and noise pollution. Government could do traffic studies using car horn counts from different intersections and use air quality data to ensure local industry isn’t polluting. The goal is to make the data legible, accessible, and valuable to all members of society so that it can then be leveraged to make future cities better for all.


As a fan of Keysight equipment, I'm signed up for notifications through multiple channels, one of which gave me the information of this challenge.


Voting is closed!