Population-Based Variable Neighborhood Descent for Discrete Optimization

Many problems in smart solution development make use of discrete optimization techniques. It is expected that smart cities will have a constant need for parcel delivery and vehicle routing which is heavily reliant on discrete optimization. In this paper we present an improvement to the Variable neighborhood descent (VND) algorithm for discrete optimization. Our method makes the search procedure more exhaustive at the expense of time performance. Instead of keeping track of a single solution which is being improved, we allow branching of the solution into at most M promising solutions and keep track of them. Our experiments show that the proposed method produces results superior to VND. We analyze the impact on time complexity and give general usage guidelines for our method.