# NavList:

## A Community Devoted to the Preservation and Practice of Celestial Navigation and Other Methods of Traditional Wayfinding

**Cocked Hat Monte Carlo Program**

**From:**Dave Walden

**Date:**2007 Mar 17, 08:33 -0700

Repeating, so it's all together. I've successfully compiled with g77 under linux and f77 (an old DEC/HP/MS compiler) under NT.

Questions?

Below is a short FORTRAN program to do Monte Carlo simulations of a three point LOP. It is rough and uncommented. I provide it as is since there seems to be quite some interest at the moment. I have found it, and other versions and modifications, very helpful as others have said recently such a program might be. It's late, and I apolgize in advance if I mis-speak
below.

It assumes, without loss of generality, the true location is at x=0, y=0. sd is the standard deviation of altitude uncertainty. For each Monte Carlo run, for each LOP it finds the altitude error based on a random selection from a distribution of errors. h45 for example. (NORRAN is the subroutine. The choice of random number generators can require some attention, as many may realize.) It checks if the true location is inside the cocked hat, itest=3. It finds the center of the cocked hat, the a,b,c,d,e,g calculation from the Nautical Almanac.
It calculates the distance from the true location to the center of the cocked hat, dist. It accumulates some variables to provide statistics on this distance as well as x and y distance from the true location. It calculates the length of the side of this cocked hat, len, via CALL CHECK. It calculates the number of times the point is with a cirle of radius of .5887*len of the true location. It writes out dist, len, and itest. By analyzing this file, one can calculate probabilites of such things as small cocked hats that do not contain the true location, how often large cocked hats do contain cocked hats, how often cocked hats of various sizes occur, etc.

These quantities can be calculated analytically and I have found the Monte Carlo results to tend towards these results as the number of runs increases. (A million only takes a minute on an old, slow PC.)

In the sample output, next, one sees that of the 100,000 runs for 25,230 cases, the true location was inside the cocked hat. Approaching 25%. The average x and y errors are -.0012 and .0008 respectively. Approaching zero as expected.

ii 100000 number less than 0.37134999 *len away ratio= 26119 in, inside cocked hat= 25230 ii,sumsq,sum 100000
0.331236064 0.510192156 sd 0.26634568 sum, sumsq,sqrt( sumsq- sum**2) 0.510192156 0.331236064 0.26634568 xsum,xsumsq,sqrt(xsumsq-xsum**2) -0.00127442868 0.16588138 0.407283396 ysum,ysumsq,sqrt(ysumsq-ysum**2) 0.0008370672 0.165351331 0.406633288 sh-3.00#

The first few lines from the output file,

0.057159 1.233867 3

0.724052 0.217705 1

0.387911 0.322774 1

0.608533 0.352897 1

0.635789 0.097189 1

1.136409 1.546760 1

0.175892 0.142276 1

show for example, a case where the distance from the center of the cocked hat to the true location is .057 min (or naut miles) and length of the sides of the cocked hat is 1.233 min
and the true location is inside the cocked hat, itest=3.

Don't be flakey. Get Yahoo! Mail for Mobile and

always stay connected to friends

--~--~---------~--~----~------------~-------~--~----~

To post to this group, send email to NavList@fer3.com

To unsubscribe, send email to NavList-unsubscribe@fer3.com

-~----------~----~----~----~------~----~------~--~---