Traffic Control Using Machine Learning

We are adding intelligence to the present traffic light system.

AdvancedShowcase (no instructions)6,391

Things used in this project

Hardware components

Arduino UNO

Raspberry Pi 3 Model B

SparkFun RFID Reader Breakout

Story

The Problem

The amount of time wasted in traffic signals is a huge problem for productivity. To avoid this problem we have designed an Intelligent Traffic System where the amount of time spent at the junction is reduced. For example:If there is a junction with four roads, the green signal remains for about 60 seconds in one particular lane, while the other three lanes remain in red. The cycle continues in every lane hence making a total time of 60 * 4 = 240 seconds. Now by ITS if there is more traffic on one particular road, there we will be an increase of the green signal time for an extra 9 seconds so it gets 69 seconds and other sides will get 57 seconds each (extra 9 seconds is divided to the other three lane) so that the total round off time is constant (I have taken one case so far, there might be cases where there will be traffic in two or more roads, as well).

Our Solution

We are placing a RFID tag in each and every vehicle; this sends a unique number and this unique number is linked to the vehicle details, insurance details and driving licence; and whensoever the vehicle approaches the traffic signal, the RFID tag sends the unique number to the receiver and finally the amount of unique numbers we get gives the number of vehicles on the road. With this number we are measuring the traffic density and determine the waiting time for the traffic signal. The illegal parking of the vehicles near the signal can be found if the same unique number from RFID tag is obtained for more than three rounds of cycles. If such a scenario occurs, then that vehicle must be stationary and it is informed to the nearest traffic police. If there is any emergency vehicle crossing the traffic signal it can be detected with its unique number and allowed then to pass through the signal without any delay. We can also reduce vehicle theft since the unique number is linked with the vehicle data.

We can't compare the number of vehicles in any two lanes because the time at which the lane is red is not unique for any two lane. So I used machine learning with linear recursion technique with my own algorithm.

MACHINE LEARNING CODE FOR ARDUINO UNO

int t=0,j,t1,t2,i,b[16],temp=0;
char a[16];
float s,sum=0,m=1,s1[15],d1;
void setup() 

{
 Serial.begin(9600); // put your setup code here, to run once:

}

void loop() 

{


i=0;
  
  // put your main code here, to run repeatedly:

t1= millis();
for(j=0;j<15;j++)
{
for(i=0;t2<=t1+2;i++)
{
if(Serial.available()>0)
{
  a[i]=Serial.read();
 // i++;

 t2=millis();
}
}

b[j]=i;
}


for(i=0;i<15;i++)
{
  s=s+b[i]-i*2-b[0];
  d1=d1+b[i]-(i-0.2)*2-b[0];
}

m=1;
if(s<d1)
{
  for(i=0;temp==1;i++)
  {
    t=0;
  for(j=0;j<15;j++)
  {
   
   s1[i]=s1[i]+b[j]-m*t-b[0];
  t=t+2;


    
  }

   m=m+.2;
   
  if((i>0)&&(s1[i-1]<s1[i]))
    {
      temp=1;
      m=m-.2;

      
    }
  }


}



else if(s>d1)
{
  //same as befor where m=m-0.2

  for(i=0;temp==1;i++)
  {
    t=0;
  for(j=0;j<15;j++)
  {
   
   s1[i]=s1[i]+b[j]-m*t-b[0];
  t=t+2;


    
  }

   m=m-.2;
   
  if((i>0)&&(s1[i-1]<s1[i]))
    {
      temp=1;
      m=m+.2;

      
    }
  }

}


Serial.println(m);
Serial.println(b[0]);
}