Inspiration
The signal lights can become a major opportunity to conserve fuel and climate mitigation.On an average every car has to pass through atleast 4 signal lights for 5 mile transit and the fuel cosumed in standby mode for waiting, the car transitional speeds impacting breaks life.Predicting the signal state prior helps in deccelarting or adjustingthe driving to reduce wait times.This micro driving techniquecan reduce transit times as well as emmissions providing both economical and environmental value.
What it does
Currently modeled with floating car data we used to estimate the staffic signal state ahead with in 2 Mile of transit route.This connected car data from real time estimates the congestion at traffic signal and the state of it as per the stop times of vehicle at the intersection.Correlating the data obtained by matching the distance matrix based on the routes wrt radius.Currently we are considering 2 mile radius of vehicles which cover upto 2-3 intersections from all directions.Based on the real time vehicle data, speed,possible congestion at intersection and wait times at particular signal we predict the traffic signal state ahead.Though there are several data models to optimize traffic signal times,there is none to predict the signal state ahead. A layover on the google maps as plugin which displays the signal state in map itself helpthe driver to micro manage their driving.We would like to optimize this model to predict the signal time left for at certain intersection.This model stands sound even for dynamic signals adopted in metro cities. Using the verizon 5G device to cast the real time transit data which is effecient enough to estimate for immediate intersection using the low latencyof AWS wavelength.We used AWS elastic to deploy the model to predict the statetimes based on the cars floating data.The prefed data of signals from Open Street Maps[OSM] helped to identify the point of interest which is intersection to estimate congestion and signal behaviour.The need for low latency is the traffic signal state varies for about average of 30sec to few minutes at a given intersection.To estimate the behaviour of signal state and predict for upcoming routes withing few mile radius is key for the effeciency and confidence scores of the model .
Challenges we ran into
Testing with NOVA without actual navigation, using navigation simulator.Understanding and choosing the training parameters from the car data obtained and designing the model.Working with wavelength zones for ecs, understanding the deployment and configurations in ecs.The model is as effecient as the speed of data obtained and computations in real time which is what the Verizon 5G as realtime fast data provider and AWS wavelenght low latency computations offered.
Log in or sign up for Devpost to join the conversation.