0

a guy asked me to solve the MCPP using heuristics. He pointed out simulated annealing. I'm doing my research and as far as I'm concerned this algorithm can not be applied for this problem. Why? Cause to find a solution some roads (edges, arcs) have to be doubled in most of the scenarios for MCPP. That gives me an infinite set of possible choices for the next edge, cause I can always re-use one of the edges that have been used before (e.g. the one that got me to the current node just a moment ago). Am I wrong? If so, what I am missing? Or does pure simulated annealing simply not fit this problem?

user6321065
  • 11
  • 1
  • Can you explain a bit more? – Dieter Meemken May 11 '16 at 15:18
  • Simulated Annealing is a Meta-Heuristic. That means, that you still define some inner local-search stuff. SA has the sole purpose of controlling the acceptance of your local-modifications (some improve the solution, some do not; so SA is always a valid approach). Look up SA and in general Meta-Heuristics in an generic form. There should be a lot of examples for SA + local-operators regarding TSP from which you could learn. – sascha May 11 '16 at 16:29
  • Ok, I agree,it is a meta heuristic. My point is, in problems like TSP I I can go from each node to any other nodes I like till I visit all of them. And in MCPP, lets say that the only edge remaining to visit is on the other side of the graph. Then I have to help myself with another algorithm to find a way to get there first, using the already used edges/arcs. Would that still be SA then? – user6321065 May 11 '16 at 17:06
  • Sure, because the other inner algorithm is the local-search part of SA. In SA-terminology: the inner-algorithm is proposing the neighbor-state. neighbor(solution_t)) -> solution_t+1 (solution implicitly says these are feasible / correct / valid solutions) – sascha May 11 '16 at 17:14

0 Answers0