Metaheuristics Based on a Set of Solutions
Abstract
This chapter describes metaheuristics evolving a set of solutions and generating new solutions by either combining existing ones or by making them cooperate through a learning process for vehicle routing problems. It presents distinguished population-based approaches, which combine solutions selected from a population stored in memory from swarm methods such as particle swarm optimization (PSO) or ant colony optimization (ACO) based on a cooperation of homogenous agents in their environment. Genetic algorithms (GAs) are subsumed on population-based approaches. Three variants are identified, namely the basic version (GA), its advanced variant using local search procedures, called memetic algorithm (MA), and a further enhanced method which includes a population management mechanism (MA|PM). Scatter search (SS) also follows a reproduction scheme that combines solutions in a systematic way. Another metaheuristic approach, namely path relinking (PR), uses two solutions from a population to generate new ones, but without crossover operators.