CODEWITHSUNDEEP
r
umair durrani
umair durrani

Reputation: 6155

How to compare two consecutive rows with a reference value in R?

I have a data frame of vehicle trajectories. Here's a snapshot:

> head(df)
  vehicle frame globalx class velocity lane
1       2    43 6451214     2    37.76    2
2       2    44 6451217     2    37.90    2
3       2    45 6451220     2    38.05    2
4       2    46 6451223     2    38.18    2
5       2    47 6451225     2    38.32    2
6       2    48 6451228     2    38.44    2

where, vehicle= vehicle id (repeats because same vehicle is observed in several time frames), frame= frame id of time frames in which it was observed, globalx = x coordinate of the front center of the vehicle, class=type of vehicle (1=motorcycle, 2=car, 3=truck), velocity=speed of vehicles in feet per second, lane= lane number (there are 6 lanes). I think following illustration will better explain the problem: enter image description here The 'frame' represents one tenth of a second i.e. one frame is 0.1 seconds long. At frame 't' the vehicle has globalx coordinate x(t) and at frame 't-1' (0.1 seconds before) it was x(t-1). The reference location is 'U'(globalx=6451179.1116) and I simply want a new column in df called 'u' which has 'yes' in the row where globalx of the vehicle was greater than reference coordinate at 'U' AND the previous consecutive globalx coordinate of this vehicle was less than reference coordinate at 'U'. This means that if df has 100 vehicles then there will be 100 'yes' in 'u' column because every vehicle will meet the above criteria once. I have tried to do this by running the function with ifelse and also tried to do the same using a for loop but it doesn't work for me. The output should have one new column:

vehicle frame globalx class velocity lane u

I tried using ifelse inside for loop and a function but it doesn't work for me.

Upvotes: 2

Views: 1613

Answers (2)

Georg Schnabel
Georg Schnabel

Reputation: 631

I assume the data frame is sorted primarily for vehicle and secondarily for globalx. If it's not you can do it by:

idx <- with(df,order(vehicle,globalx))
df <- df[idx,]

Now, you can perform it with the following vectorized operations:

# example reference line
U <- 6451220
# adding the extra column
samecar <- duplicated(df[,"vehicle"])
passU <- c(FALSE,diff(sign(df[,"globalx"]-U+1e-10))>0)
df[,"u"] <- ifelse(samecar & passU,"yes","no")

Upvotes: 1

Nicholas Hamilton
Nicholas Hamilton

Reputation: 10506

Here is my solution:

First create dummy data, based on your provided data (I have saved it to data.txt on my desktop), duplicate the data so that there are two cars with the same identical data, but different vehicle id's:

library(plyr)
df <- read.table("~/Desktop/data.txt",header=T)
df.B <- df; df.B$vehicle = 3 #For demonstration
df <- rbind(df,df.B); rm(df.B)

Then we can build a function to process:

mvt <- function(xref=NULL,...,data=df){
  if(!is.numeric(xref)) #Input must be numeric
    stop("xref must be numeric",call.=F)
  xref = xref[1]

  ##Split on vehicle and process.
  ddply(data,"vehicle",function(d){
    L = nrow(d) #Number of Rows
    d$u = FALSE #Default to Not crossing

    #One or more rows can be checked.
    if(L == 1)
      d$u = (d$globalx > xref)
    else if(L > 1){
      ix <- which(d$globalx[2:L] > xref & d$globalx[1:(L-1)] <= xref)
      if(length(ix) > 0)
        d$u[ix + 1] = TRUE
    }

    #done
    return(d)
  })
}

Which can be used in the following manner:

mvt(6451216)
mvt(6451217)

Upvotes: 0

Related Questions