Original article can be found here (source): Artificial Intelligence on Medium
Together, We discovered the road bumps
I am going to tell you how!
it ‘s been a few since we added bumps on the map. Now, when you are navigating, you’ll be notified seconds before a road bump so you can balance your speed and drive with ease. This way, you and your passengers in the car won’t be hurt or troubled.
Btw, you can download the app here
This was the whole news! Now I am going to explain how it happened and how we discovered road bumps.
It was six months ago when we decided to find the road bumps with the help of users. We put an option for users to report the bumps. To accurately find bumps, we needed a lot of reports. We put all these reports together and started processing them into a series of locations.
There were problems though. Some of the reports ended as wrong outputs, especially on roads that there were no bumps at all. On the other hand, we missed bumps that were important due to the lack of reports.
The project was going to final(fail). We came up with something new. We realized that drivers usually slow down before a bump. Defining this in heavy traffic was a bit challenging. But during the calm hours like midnight, a sudden slowdown in the speed was a clear sign of a bump.
Using AI, we could identify these locations. All we had to do was to review midnight data and check slowdowns. There was a problem. That slow down happens before a bump, not on it. We could use this only to find out where we are close to a bump. Another obstacle was intersections where drivers usually slow down.
Again, we had to go back to user input. Using an algorithm, we defined the probable locations for bumps, followed by checking high levels of reports by users , we managed to do it.
We checked the first output in July and realized that there were still so many bumps that were not included. We checked 10–20 samples and found out that the algorithm did find the locations but reports were inadequate. We could wait for more reports but that would be late. That is why we asked users to report bumps more for a few days. The reports were more and now 16000 bumps are defined on the map in the country.
The joy in this project was to combine user intelligence and Artificial intelligence. We could finally lower the cost of front suspensions of cars in this economic circumstance.
Neshan can now display police, traffic, cameras and road bumps. AI and the users’ help shine the way for us. Please be with us more and let’s show all the police, traffic cameras and road bumps.