Memetic cooperative models for the tool switching problem

Abstract

This work deals with memetic-computing agentmodels based on the cooperative integration of search agents endowed with (possibly different) optimization strategies, in particular memetic algorithms. As a proof-of-concept of the model, we deploy it on the tool switching problem (ToSP), a hard combinatorial optimization problem that arises in the area of flexible manufacturing. The ToSP has been tackled by different algorithmic methods ranging from exact to heuristic methods (including local search meta-heuristics, population-based techniques and hybrids thereof, i.e., memetic algorithms). Here we consider an ample number of instances of this cooperative memetic model, whose agents are adapted to cope with this problem. A detailed experimental analysis shows that the meta-models promoting the cooperation among memetic algorithms provide the best overall results compared with their constituent parts (i.e., memetic algorithms and local search approaches). In addition, a parameter sensitivity analysis of the meta-models is developed in order to understand the interplay among the elements of the proposed topologies.

The job sequencing and tool switching problem: state-of-the-art literature review, classification, and trends

“…Only the results from the cooperative SDI technique seemed promising, but they were still inferior to the MA plus HC proposed by Amaya, Cotta, and Fernández (2008). At about the same time, Amaya, Cotta, and Leiva (2010a) and Amaya, Cotta, and Fernández-Leiva (2011) extend the above mentioned cooperative scheme and combine the features of memetic agent models supported by different LS mechanisms, namely HC and TS. The results show that especially heterogeneous memetic agents that exchange current best solutions outperform individual agents.…”

“…The computational results show that the MA significantly outperforms GA and HC when the number of jobs increases. Leiva (2010a, 2010b) and Amaya, Cotta, and Fernández-Leiva (2011) address the area of parallel meta-heuristics, particularly cooperative search algorithms, which means that several running algorithms or so-called parallel cooperating agents search for a solution in the whole solution space. Amaya, Cotta, and Leiva (2010b) propose four cooperative methods, three with a specific interaction structure endowed with LS, and one model with a specific search, diversification, and intensification (SDI) technique.…”

A Simple and Effective Hybrid Genetic Search for the Job Sequencing and Tool Switching Problem

“…Population-based methods have also been successful. Hybrid methods combining genetic algorithms (GA) with other search procedures can be found in (Ahmadi et al, 2018;Amaya et al, 2008Amaya et al, , 2011Amaya et al, , 1305Amaya et al, , 2012Amaya et al, , 2010Chaves et al, 2016). Amaya et al (1305Amaya et al ( , 2008Amaya et al ( , 2012 combined GA with local search-based procedures such as hill climbing, simulated annealing and tabu search, leading to hybrid methods which are also known under the name of memetic algorithms.…”

“…Amaya et al (1305Amaya et al ( , 2008Amaya et al ( , 2012 combined GA with local search-based procedures such as hill climbing, simulated annealing and tabu search, leading to hybrid methods which are also known under the name of memetic algorithms. Amaya et al (2011Amaya et al ( , 1305Amaya et al ( , 2010 combined GA with a multi-agent approach or cross-entropy methods. Finally, Chaves et al (2016) combined clustering search (CS) with a biased random-key genetic algorithm (BRKGA), while Ahmadi et al (2018) solved the problem as a TSP of second order (Jäger and Molitor, 2008) with a learning-based GA. Table 2.1 summarizes the SSP studies in chronological order, lists their main contributions as well as the origin of the benchmark instances considered.…”

“…We also refer to Calmels (2019) for a comprehensive literature review and classification of SSP variants. Tang and Denardo (1988a) 1988 Exact methods + heuristics Tang and Denardo (1988a) Bard (1988) 1988 Exact methods + heuristics Bard (1988) Crama et al (1994 1994 Heuristics Most of the aforementioned GA-based implementations use traditional parent-selection strategies (Whitley, 2019) such as roulette wheel (Ahmadi et al, 2018) and binary tournament (Amaya et al, 2008(Amaya et al, , 2011(Amaya et al, , 2012. Survivor selections are solely based on individual objective values, typically obtained with KTNS (Ahmadi et al, 2018;Amaya et al, 2008Amaya et al, , 2011Amaya et al, , 1305Amaya et al, , 2012Chaves et al, 2016).…”