GPS data visualization

Selecting GPS data from raw telematics data

kable(head(mytelematics, 100), format = "html") %>% 
  kable_styling() %>% 
  scroll_box(width = "100%", height = "300px")

Even though there are many variables related to other sensors and accuracy of GPS, let us choose only four variables as follows;

gpsdata <- mytelematics %>% select(Timestamp, Long, Lat, Speed.m.s.)
colnames(gpsdata) <- c("Timestamp", "x", "y", "speed")

# convert speed unit (from m/s to mph)
gpsdata$speed <- gpsdata$speed * 2.23694 

Thus, the GPS data we are working on in this article consists of time, longitude, latitude, and speed information. Note that I changed the unit of the speed from m/s to mph.

kable(head(gpsdata, 100), format = "html") %>% 
  kable_styling() %>% 
  scroll_box(width = "100%", height = "300px")

Converting Time variable

As we can see in the above table, the Timestamp variable contains the information of the time, but it is string data. Since we want to use the data for the animated plot, we need to convert them into the relative time scale, whose starting value should be zero seconds.

My solution is a little bit inefficient and possibly has an error if the recording time is around midnight (if this is the case then we need to think about the date too), but since my driving data mostly during the afternoon, I converted the time string into the data with unit second as follows;

# seperate date and time
temp_time <- as.character(gpsdata$Timestamp) %>% strsplit(" ") %>% unlist

# select the time information and split into three part
temp_time <- temp_time[seq(2, length(temp_time), by = 2)] %>% strsplit(":")

# conversion formula for hh:mm:ss to seconds. 
convert_time <- function(timelist){
    timelist <- unlist(timelist)
    hr <- as.numeric(timelist[1]) * 3600
    min <- as.numeric(timelist[2]) * 60
    sec <- as.numeric(timelist[3])
    result <- hr + min + sec
    result
}

# apply convert_time function to temp_time
gpsdata$Timestamp <- lapply(temp_time, convert_time) %>% unlist()
gpsdata$Timestamp <- gpsdata$Timestamp - gpsdata$Timestamp[1]

Now, the timestamp of the GPS data starts with zero as follows;

kable(head(gpsdata, 50), format = "html") %>% 
  kable_styling() %>% 
  scroll_box(width = "100%", height = "300px")

Frequency of GPS data

One thing you may notice is that the frequency of our timestamp is higher than 1 Hz, which means the gap between each time point is less than 1 second. Even though GPS data has been recorded as 1 Hz, it is recorded with higher frequency, like 25 Hz in this case, since other recorded sensors use a higher frequency recording rate, so GPS data is also recorded at the same rate. However, the actual information about the position of a vehicle is changed every 1 second, which can be confirmed by drawing the time vs. speed graph as follows;

p <- ggplot(data = gpsdata, aes(x = Timestamp, y = speed)) +
        geom_line(alpha = 0.3) +
        labs(x = "time (sec)", y = "speed (mph)",
             title = "Trip speed") +
        theme_bw()
p

Thus, since we do not need redundant information, let us select the data for every 25 data points.

gpsdata <- gpsdata[seq(1, length(gpsdata$Timestamp), by = 25),]

Thus, the final data set for the GPS visualization is as follows;