====== TSP with multiple agents ======
Data: {{ :courses:a0m33eoa:en:semestral_tasks:mtsp:mtsp_data.zip |data_mtsp}} for 3, 4 and 5 agents

===== Problem description =====

This is a modification of the classical [[https://en.wikipedia.org/wiki/Travelling_salesman_problem|Traveling Salesman Problem]].
 
Given is an undirected graph with $N$ nodes (cities), one of them serving as a **depot**. The depot is a node where all routes start and end. 
Further, there are $M$ agents (traveling salespersons). The goal is to find a set of $M$ routes such that the length of the longest one is minimal subject to
  * all routes start and end at the pivot node,
  * all nodes, except the depot, are visited exactly ones by some of the agents,
  * no route can have zero length.

{{:courses:a0m33eoa:semestralni_ulohy:mtsp:mtsp.png|}}

===== Possible representations =====
  * [permutation of nodes][break-points]
    * example: depot=1, [2-4-3|5-8-6-7|10-9][3,7] implies routes r1=[1,2,4,3,1], r2=[1,5,8,6,7,1], r3=[1,10,9,1]
  * ...

===== Uniform evaluation function =====
Solution quality is determined by the length of the longest tour, that is to be minimized.

<WRAP center round info 100%>
If solved as a pure black-box optimization problem, one could use information about cities locations or the distance matrix only in the evaluation function. Here, it is allowed to use or analyze such a kind of information elsewhere as well, e.g., in recombination operators.  
</WRAP>