This paper proposes an extension method for Ant Colony Optimization (ACO) algorithm called Dynamic Impact. Dynamic Impact is designed to solve challenging optimization problems that has nonlinear relationship between resource consumption and fitness in relation to other part of the optimized solution. This proposed method is tested against complex real-world Microchip Manufacturing Plant Production Floor Optimization (MMPPFO) problem, as well as theoretical benchmark Multi-Dimensional Knapsack problem (MKP). MMPPFO is a non-trivial optimization problem, due the nature of solution fitness value dependence on collection of wafer-lots without prioritization of any individual wafer-lot. Using Dynamic Impact on single objective optimization fitness value is improved by 33.2%. Furthermore, MKP benchmark instances of small complexity have been solved to 100% success rate where high degree of solution sparseness is observed, and large instances have showed average gap improved by 4.26 times. Algorithm implementation demonstrated superior performance across small and large datasets and sparse optimization problems.
Optimization algorithm has been designed for traveling salesman problem (TSP) described in Dorigo [1] doctoral theses in 1992. ACO algorithm has been successfully implemented to solve number of different problems. Routing problems [2], scheduling and sequencing problems [3] [4] [5] [6] [7], subset problems [8]. Furthermore, ACO has been used to solve large scale optimization problems as demonstrated in these research papers [9] [10] [11]. ACO also has been successfully used for multiple types of scheduling problems. For resource constrained project scheduling problem [3], tardiness problem [4], job shop scheduling problem [5] [6], and other types. Dorigo and Stützle in this research [12] explain implementation methods in detail to solve various scheduling problems using Ant Colony Optimization algorithm.
Microchip manufacturing is a complex process that utilizes expensive machinery. Tight manufacturing schedules are used in order to run processes at maximum efficiency, minimize machinery down time and always have enough stock of product in demand. Often predicted microchip demand does not meet observed real demand, and microchip production schedule must be altered accordingly to meet newly specified demand. Microchip manufacturing scheduling problems have been researched from various points of view. Scheduling robotic arms of two-cluster tools in microchip manufacturing facilities [13], transport scheduling in automated material handling systems for wafer manufacturing plants [14], wafer production scheduling as a job shop scheduling problem [15]. In this paper the research focus is on the resource constrained production scheduling. Microchip manufacturing plant production floor scheduling is a difficult task, as the nature of the problem does not allow to have a set heuristic information on each edge that enables ants to efficiently navigate the search space. MMPPFO is a scheduling problem. Scheduling problems are proven to be strongly NP-Hard combinatorial problem [16].
To solve scheduling problems variety of different algorithms have been explored. The comparison of Ant Colony Optimization (ACO) versus Genetic Algorithm (GA) and Simulated Annealing (SA) has been conducted buy Huang et. al. [17]. Researchers have found ACO algorithm to be the most effective obtaining feasible solutions for NP-complete scheduling problems. Moreover, semiconductor wafer fabrication scheduling using Ant Colony Optimization was explored in [15] and showed ACO algorithm to be highly effective on a large optimization problem.
In addition to solving real world optimization problem this research further proved the validity of proposed methods by solving a theoretical Multi-Dimensional Knapsack Problem (MKP) as well as compare the results with previously published research papers. The goal of MKP is to maximize the total profit of the items taken into knapsacks, where all items have multi-dimensional weights for each knapsack and each knapsack have a capacity that must not be exceeded [18]. The nature of packing different size items in all knapsacks simultaneously makes the feasible region of the search very sparse [19]. Such sparsity is a great challenge for optimization algorithms where good solutions are obtained by iterative convergence.
Combinatorial search algorithms are designed to explore large search spaces efficiently and converge to a good solution quickly. The efficiency is achieved using metaheuristic methods that allows the search space to be explored more in areas of greater reward. These combinatorial search algorithms usually have multiple hyper parameters, that are often tricky to find such that the best convergence speed is achieved. Hyper-heuristics are methods introduced by Burke et. Al. [20] to generate or choose heuristic that enables combinatorial metaheuristic algorithms converge faster. Hyper-heuristics has been adopted to use multiple low-level heuristic algorithm search results as a search space [21]. For Ant Colony Optimization algorithm such hyper-heuristic usually tunes 𝛼 , 𝛽 , and 𝜌 search hyper parameters [22]. However, using similar approaches, it is possible to have more sophisticated search with introduced lower level heuristics within metaheuristic algorithm. The Stochastic Gradient Ascent introduced by Dorigo et. Al. [23] introduced manipulation of the Ant Colony pheromone matrix. Such pheromone correction allows Tuba and Jovanovic [24] avoid algorithm stagnation. Furthermore, sub-heuristics are the heuristic methods applied within a core of search algorithm that acts upon the state of incomplete partial solution. authors at [25] has utilized such heuristic for Ant Colony Optimization algorithm for probability calculation where a branching can occur while building the solution. This sub-heuristic method allowed them to have transition operation that otherwise could not be accounted from the solutions previously explored.
Our research on MMPPFO has real optimization data for initial testi
This content is AI-processed based on open access ArXiv data.
