from
path.append(
import
Melper
x = Melper.returnFilenameTimestamp()
print x
Oh yes … !
Meetings for Feb-2009
Progress
Today got the Analysis/Mung code to work, when controlled by the Controller. Woot.
Worked out why my loop was failing — a pretty important array was not being cleared out during each loop. Problem resolved now though.
Ran out of time on the case of importing modules from other directories … but the answer can be found here!: http://stackoverflow.com/questions/456481/cant-get-python-to-import-from-a-different-folder
Changing the code to remove toroidality was remarkably straight forwards — just delete the offending section of code!! Boids now start within same starting area, but can move off to wherever they fancy … need to visualise this soon to actually see what they are doing.
Then moved onto the controller code, to start looping through the simulation to generate lots of data. However, for some reason code runs more times than anticipated .. for further investigation tomorrow methinks.
Back from hols today and trying to remember where everything got left, prior to a rather trying week at work …
The toriodality issue and mapping to a new plane is a non starter, as it will screw up the dynamics of a flock. Leaning to a huge world for the babies to play in, but starting them off in a smaller playpen, that does not have outlimits — only looking at the numbers, baby, not the true world.
How to get over the convex hull / toroidal issue:
Annoyingly this will have to be done in Model code, and not in Analysis code, unless a great thought strikes me from above.
To consider: Regarding toroidality of flocks: need to surely include a scale of the difference when moving boids over the boundary, not just a flat reset to w.g. widh parameter?
Nice quote on mathematical models btw from http://stats.lse.ac.uk/penzer/ST419materials/CSchpt3.pdf:
Mathematical models fall into two categories, deterministic and stochastic. For a deterministic
model, if we know the input, we can determine exactly what the output will be. This is not true
for a stochastic model; the output of a stochastic model is a random variable.
See above link and http://cran.r-project.org/doc/contrib/Lemon-kickstart/kr_rfunc.html for details on writing your own R functions.
Don’t you just hate it when your flock breaks through the toroid-ality of your environment, screwing up your convex hull calculation? If so, why not take your flock to the next dimension, chull their coords around a common reference point, and then return them to the dimension they came from … ?
NB Beginners Guide To R: http://cran.r-project.org/doc/contrib/Paradis-rdebuts_en.pdf
Stop the press!: R has a convex hull function in it, called chull() — could be useful. Will need to use matrix()
# read in area information and plot to screen
coords <- read.table(“c:/temp/file_coords.txt”, sep=”,”, fill=TRUE, row.names = 1)
coords = coords[,1]
time = 1:length(coords)
plot(coords ~ time)
# read in velocity information and plot to scren
vels <- read.table(“c:/temp/file_vels.txt”, sep=”,”, row.names = 1)
vels <- vels[,-length(vels)]
a <- list()
for (i in 1:length(x)) {a[[i]] <- as.vector(unlist(x[i,]))}
std_devs = sapply(a, sd)
time = (1:length(s))
plot (time ~ std_devs)
And saveplot() apparently saves your plot (maybe see http://www.nabble.com/Very-new—beginners-questions-to12125162.html#a12126867) — go figure …
New year, new problems … !
Playing around with R today, to try some initial analysis. Things learnt so far:
