Week 5 - Oct 02 - 08
This week I have planned to design and build module #1 which would the biggest and the main module in my trafficpredictor project. In this blog, I have covered up the machine learning concept/ algorithm to be used. Along with this, I am going to unveil the case to be used for mounting the Nucleo - L476RG kit in the vehicle.
I am still exploring ways to utilize the kit's storage capacity, once done will share the same in my future blogs.
Module 1 - Traffic prediction with machine learning on an added advantage of mass storage capability of the STM32 Nucleo-64 development board
So far everything to kickstart my trafficpredictor project has been completed.
Code Editor & Software IDE required
Yet to come
App (Android / Web)
Code Snippets
MATLAB GUI
Machine Learning New
Machine learning (ML) has been the talk of the town along with Artificial intelligence (AI). ML is an interesting area because we can teach the computer how to identify objects/ face through computer vision (CV) / predict data. The reason I have opted for machine learning to predict traffic is its ability to process a huge amount of data and convert them into a single output.
There are two main types of machine learning - Classification and Clustering. I have chosen to classify data into either best route with less traffic or route with traffic, this would fall under type classification. This is one way because of my prior experience with classification type machine learning in MATLAB and also it better suits the project purpose.
Data refers to GPS data - latitude & longitude from API plus time at which data was recorded.
Output refers to resultant of machine learning process - probably a route which is predicted to have less traffic.
GUI for machine learning
I have a GUI developed in MATLAB already for machine learning smells which can be seen below. The graph plotted to the left (Classification) separates the fragrance of one perfume from other (separated by a line). The graph to the right (Clustering) clusters the smell values around the 'X' marked and thus separates them.
Link to share your thoughts in Discussions related to the above project "Odor Detector", if interested - Is it possible to transmit and receive smell?
| {tabbedtable} Tab Label | Tab Content | ||||
|---|---|---|---|---|---|
| Building a case for the kit Updated | Check out the next tab for"How to connect Nucleo-L476RG board to USB Charger Power Bank
The case has been built to hold/mount the Nucleo-L476RG kit onto the vehicle therefore while on travel we can record the GPS location data in future modules to identify real-time traffic The case includes two provisions one to hold the MCU board with expansion boards and the other to hold the battery pack power bank Though the case is not an enclosure type it is hard enough to protect the case hold the kit and battery in position even in tough terrains and allows easier access I have added below photos video of the kit inside the case and mounted onto my bike
For those who are not able to afford 3D - Printers like me, I would suggest Metal Mechanix - 5. It has been much help in bringing my designs into life at a much faster rate.
Travel with the newly built case - Video
Please do not attempt to record audio / video while driving the vehicle alone or without any safety harness and auto recording equipments. | ||||
| How to connect Nucleo-L476RG board to USB Charger Power Bank | Mobility would be the key criteria when it comes to moving the prototype to real-traffic to gather data for machine learning. So I have decided to use my 20,000mAh power bank with 5.1V 2A output as the external power source.
Steps to connect Nucleo-L476RG board to USB Charger Power Bank
Nucleo-L476RG board connected to power bank
Reference for this article: Nucleo-L476RG Datasheet |
Please find below links to my previous blogs on my traffic predictor project for IoT on Wheels design Challenge,
Blog 1 - The Official Announcement
Blog 2 - Quest for the code editor
Blog 4 - Finally, the kit arrives
Top Comments