A Continuous-GRASP Random-Key Optimizer

This paper introduces a problem-independent GRASP metaheuristic for combinatorial optimization implemented as a random-key optimizer (RKO). CGRASP, or continuous GRASP, is an extension of the GRASP metaheuristic for optimization of a general objective function in the continuous unit hypercube. The novel approach extends CGRASP using random keys for encoding solutions of the optimization problem in the unit hypercube and a decoder for evaluating encoded solutions. This random-key GRASP combines a universal optimizer component with a specific decoder for each problem. As a demonstration, it was tested on five NP-hard problems: Traveling salesman problem (TSP); Tree hub location problem in graphs (THLP); Steiner triple set covering problem (STCP); Node capacitated graph partitioning problem (NCGPP); and Job sequencing and tool switching problem (SSP).