Overview / Usage

**What if we could control priority of traffic lights by using sensor data from cars and adapting traffic light signals to demand in real-time?  **This can achieve in near real-time but will require predictive analytics of when a certain volume(number) of cars will arrive at a particular spot and giving priority based on the demand. The solution proves to be more proactive than by using traffic light cameras to control traffic only when signs of congestion start to arise. The challenge posed by such a solution is that it will require fast processing of vehicle speed and coordinate data in order to determine the time it will take for certain number of cars to arrive at a particular destination and therefore give priority based on demand in real-time. Such high volume of data can be processed on a heterogeneous system with multiple GPUs or FPGAs for near real-time decision making.

Methodology / Approach

The project uses DPC++ to calculate the approximate distance that a motor vehicle will arrive at a particular spot (where a traffic light post is placed). Data about the current position of the vehicle is gotten from the users mobile device GPS sensor. Because of the current nature of numerous cars on the road, data ingested into the system will be numerous and requires processing in real-time to estimate the time the car will arrive at a given spot. Based on the current demand(load) on a particular road, traffic light systems can be adjusted to meet demand in real-time.

To calculate the distance between the car and a traffic light spot, we use the haversine function to estimate the distance between the two spots. This can be offloaded to an accelerator to speed up the computation and make real-time decisions.

Technologies Used

Intel Devcloud

DPC++

Documents and Presentations

Repository

https://github.com/JohhnyKinuthia/oneapi-challenge