Keith Briggs

This page was last modified 2024-01-21  

.
.


home 
·publications 
·thesis 
·talks 
·meetings 
·records 
·maths notes « 
·software 
·languages 
·music 
·travel 
·cv 
·memberships 
·students 
·maps 
·place-names 
·people 
·photos 
·links 
·ex libris 
·site map 


.

The travelling salesman problem for randomly distributed nodes

Suppose n nodes are uniformly distributed (iid) on a unit 2-torus. I use a torus to avoid edge effects. What is the expected path length L(n) of the optimum travelling salesman problem (TSP) tour in Euclidean norm?

It is known that limn→∞ L(n)/sqrt(n) exists and is about 0.7120 (see [1]). Here is my data, which is an attempt to get a better value for this constant. The error bars are standard errors of the means. For each value of n, I solved 105 TSP instances exactly. All TSP data was computed using Concorde.

references

[1] Allon Percus: The stochastic traveling salesman problem This website uses no cookies. This page was last modified 2024-01-21 10:57 by Keith Briggs private email address.