The all-electric ship (AES) with DC-grid configuration has demonstrated advantages compared to the traditional AC system and has become the state-of-the-art for ships with electric propulsion in the low to medium power range during the past decade. However, the integration with different power sources, such as fuel cells, batteries and diesel gen-sets, increases the system complexity and requires an advanced power management system (PMS) to handle vessel operation and to achieve optimal power control. This paper presents an optimized power management strategy to reduce the total cost of ownership of such vessels, considering not only the fuel cost and emission penalty, but also the power device degradation and equipment replacement cost. In this study, Model Predictive Control (MPC) and Reinforcement Learning (RL)-based PMS control methods are approached respectively. In order to demonstrate the performance of MPC and RL techniques, a typical tugboat load profile is simulated. The testing results are also compared with a traditional rule-based power management control.

The evolutionary approach of Genetic Programming (GP) has been applied extensively to model various non-linear systems. The distinct advantage of using GP is that prior assumptions for the selection of a model structure are not required. The GP automatically evolves the optimal model structure and its parameters that best describe the system characteristics. However, the evolution of an optimal model structure is highly dependent on the experimental designs used to sample the problem (system) domain and capture its characteristics. The literature reveals that very few researchers have studied the effect of various experimental designs on the performance of GP models and therefore the optimum choice of an experimental design is still unknown. This paper studies the effect of various experimental designs on the performance of GP models on two non-linear test functions. The objective of the paper is to identify the most robust (best) experimental design for effective modeling of non-linear test functions using GP. The analysis reveals that for the test function 1, the experimental design that gives best performance of GP models is response surface faced design and for test function 2, the best experimental design is 5-level full factorial design. Thus, the result concludes that the selection of the robust experimental design is a crucial preprocessing step for the effective modeling of non-linear systems using GP.

Genetic programming (GP) and its variants have been extensively applied for modeling of the stock markets. To improve the generalization ability of the model, GP have been hybridized with its own variants (gene expression programming (GEP), multi expression programming (MEP)) or with the other methods such as neural networks and boosting. The generalization ability of the GP model can also be improved by an appropriate choice of model selection criterion. In the past, several model selection criteria have been applied. In addition, data transformations have significant impact on the performance of the GP models. The literature reveals that few researchers have paid attention to model selection criterion and data transformation while modeling stock markets using GP. The objective of this paper is to identify the most appropriate model selection criterion and transformation that gives better generalized GP models. Therefore, the present work will conduct an empirical analysis to study the effect of three model selection criteria across two data transformations on the performance of GP while modeling the stock indexed in the New York Stock Exchange (NYSE). It was found that FPE criteria have shown a better fit for the GP model on both data transformations as compared to other model selection criteria.

Traditionally, complex systems were treated using centralized approaches; however, recent trends highlighted that the growing complexity can be best dealt by decomposing the entire system into subsystems and further treat them in a distributed way. The approach of Probability Collectives (PC) in Collective Intelligence (COIN) framework decomposes the entire system into a Multi-Agent System (MAS) or a collection of rational and self interested agents and further optimizes them in a distributed and decentralized way to reach the desired system objective. The complexity of the system increases when constraints are involved. The approach of PC is incorporated with a feasibility-based rule to handle the solution based on number of constraints violated, and further drives the convergence towards feasibility. Importantly for the first time, constrained PC approach has been tested solving a discrete problem such as 45-bar truss structure. The results are validated by comparing with the contemporary algorithms as well.

The complex systems can be best dealt by decomposing them into subsystems or Multi-Agent System (MAS) and further treat them in a distributed way. However, coordinating these agents to achieve the best possible global objective is one of the challenging issues. The problem becomes harder when the constraints are involved. This paper proposes the approach of Probability Collectives (PC) in the Collective Intelligence (COIN) framework for modeling and controlling the distributed MAS. At the core of the PC methodology are the Deterministic Annealing and Game Theory. In order to make it more generic and capable of handling constraints, feasibility-based rule is incorporated to handle solutions based on the number of constraints violated and drive the convergence towards feasibility. The approach is validated by successfully solving two test problems. The proposed algorithm is shown to be sufficiently robust and other strengths, weaknesses and future directions are discussed.

The best option to deal with a complex system that is too cumbersome to be treated in a centralized way is to decompose it into a number of sub-systems and optimize them in a distributed and decentralized way to reach the desired system objective. These sub-systems can be viewed as a multi-agent system (MAS) with self-learning agents. Furthermore, another challenge is to handle the constraints involved in real world optimization problems. This paper demonstrates the theory of probability collectives (PC) in the collective intelligence (COIN) framework, supplemented with a penalty function approach for constraint handling. The method of deterministic annealing in statistical physics, game theory and Nash equilibrium are at the core of the PC optimization methodology. Three benchmark problems have been solved with the optimum results obtained at reasonable computational cost. The evident strengths and weaknesses are also discussed to determine the future direction of research.

A complex system may be controlled and optimized in a more efficient and manageable fashion by treating it as a distributed Multi-Agent System (MAS). But the major challenge in such an approach is to make the agents work in a coordinated way to optimize the system objective via optimizing their individual local goals. This paper describes the modified Probability Collectives (PC) as an evolutionary and distributed approach to achieve the system objective. The approach is validated solving a combinatorial optimization problem such as the Single Depot Multiple Traveling Salesmen Problem (MTSP). Moreover, as constraint handling in evolutionary systems has remained a challenge for years, an effort towards developing a generalized technique incorporating constraints into the PC approach is also attempted. It is validated by solving the practical problem of a spring design. The optimum results are obtained at a reasonable computational cost.

Genetic algorithms (GA) do have some advantages over gradient-based methods for solving topology design optimization problems. However, their success depends largely on the geometric representation used. In this work, an enhanced morphological representation of geometry is applied and evaluated to be efficient and effective in producing good results via a target matching problem: a simulated topology and shape design optimization problem where a dasiatargetpsila geometry set is first predefined as the Pareto optimal solutions and a multiobjective optimization problem formulated such that the design solutions will evolve and converge towards the target geometry set. As the objectives (and constraints) are conflicting, the problem is challenging and an adaptive constraint strategy is also incorporated in the GA to improve convergence towards the true Pareto front.

This paper describes a technique for design under uncertainty based on hybrid genetic algorithm. In this work, the proposed hybrid algorithm integrates a simple local search strategy with a constrained multi-objective evolutionary algorithm. The local search is integrated as the worst-case-scenario technique of anti-optimization. When anti-optimization is integrated with structural optimization, a nested optimization problem is created, which can be very expensive to solve. The paper demonstrates the use of a technique alternating between optimization (general genetic algorithm) and anti-optimization (local search) which alleviates the computational burden. The method is applied to the optimization of a simply supported structure, to the optimization of a simple problem with conflicting objective functions. The results obtained indicate that the approach can produce good results at reasonable computational costs.

The growing number of components and communication in complex systems requires them to be treated as a collective of subsystems/agents having distributed and decentralized control. The major challenge in such approach is the coordination among agents optimizing their local goals and contributing towards optimization of the global objective. This paper implements the theory of Collective Intelligence (COIN) using Probability Collectives (PC) approach to achieve the global goal. This approach works on probability distribution, directly incorporating uncertainty and has deep connections to Game theory, Statistical Physics and Optimization. In this approach, the agents select actions over a particular range and receive some rewards on the basis of the overall system objective achieved because of those actions. The approach is illustrated using the problem of segmented beam minimizing total volume. Each segment is considered as an agent competing with one another to achieve the total minimum volume. The implementation produced encouraging results.

This paper demonstrates the automatic design of a compliant grip-and-move manipulator by topology and shape optimization using a hybrid genetic algorithm. It presents the novel idea of integrating both grip and move behaviors within one simple compliant actuator mechanism. The manipulator employs two identical path generating compliant mechanisms with two degrees-of-freedom each so that it can grip a workpiece and move it to anywhere within its working area. They are difficult to design because their motions have to be analyzed by finite element methods and the relationship between their geometry and their elastic behavior is highly complex and non-linear. The synthesis of such a mechanism in this work is achieved by a structural optimization approach based on a hybrid genetic algorithm and a morphological representation scheme for defining the structural geometry upon a finite element grid. The proposed hybrid algorithm integrates a simple local search strategy with a constrained multiobjective evolutionary algorithm. A novel constrained tournament selection is used as a single objective function in the local search strategy. The selection is utilized to determine whether a new solution generated in local search process will survive. Hooke and Jeeves method is applied to decide search path. Good initial solutions, the solutions to be mutated, are chosen for local search.

This paper describes a technique for design under uncertainty based on hybrid genetic algorithm. In this work, the proposed hybrid algorithm integrates a simple local search strategy with a constrained multi-objective evolutionary algorithm. The local search is integrated as the worst-case-scenario technique of anti-optimization. When anti-optimization is integrated with structural optimization, a nested optimization problem is created, which can be very expensive to solve. The paper demonstrates the use of a technique alternating between optimization (general genetic algorithm) and anti-optimization (local search) which alleviates the computational burden. The method is applied to the optimization of a simply supported structure under load uncertainties, to the optimization of a simple problem with conflicting objective functions. The results obtained indicate that the approach can produce good results at reasonable computational costs.

Decision support tools for assisting the human mission planner in the deployment of sensor networks is an important component of sensor management. The optimal selection of the number and types of sensors available from a suite of sensors, and their optimal placement in the terrain, is typically a multiobjective optimization problem with objectives defined based on the mission and scenario. One of the key advantages is applying multiobjective genetic algorithms for solving such problems is their ability to find multiple Pareto optimal solutions in a single run which then allows for the mission planner to select a final optimal solution based on higher-level considerations. The aim in this work is an effective genetic algorithm implementation of such a decision support tool for the deployment of sensor networks.

The structural topology optimization approach can be used to generate the structural design for some desired input-output (force-deflection) requirements. Optimization methods based on genetic algorithms (GA) have recently been demonstrated to have the potential for overcoming the problems associated with gradient-based methods. The success of the GA depends, to a large extent, on the structural geometry representation scheme used. In this work, some enhancements are incorporated into the recently developed morphological geometric representation scheme coupled with a GA. Based on the morphology of living creatures, a geometric representation scheme had earlier been developed that works by specifying a skeleton which defines the underlying topology/connectivity of a structural continuum together with segments of material surrounding the skeleton. In this work, the flexibility to turn on or off parts of the skeleton is integrated into the scheme. This improves the variability of topological and shape characteristics in the evolutionary process and enhances the representation's versatility. The methodology is tested by solving a multicriterion 'target matching' problem : a simulated topology optimization problem where a 'target' geometry is first created and predefined as the optimum solution, and design solutions are evolved to converge towards this target shape.

The genetic algorithm (GA) is a potent multiobjective optimization method, and the effectiveness of hybridizing it with local search (LS) has recently been reported in the literature. In this work, the proposed hybrid algorithm integrates a simple local search strategy with an effective constrained multi-objective evolutionary algorithm. A novel constrained tournament selection is used as a single objective function in the local search strategy. The selection is utilized to determine whether a new solution generated in local search process will survive. Hooke and Jeeves method is applied to decide search path. Good initial solutions, the solutions to be mutated, are chosen for local search. This paper also examines the following strategies in the implementation of local search: applying local search only to final solutions, applying local search to solutions only in the early generations, and initializing local search when mutation gives rise to improvement in the solution. Simulation results from a target matching test problem indicate that the hybrid algorithm outperforms the multi-objective method without genetic local search when the implementation of local search is appropriate. It is also shown that the hybridization can improve the convergence speed.

Very often real-world applications involve multiple objectives. Research on multiobjective evolutionary optimization algorithms have amply demonstrated that they are capable of finding multiple and diverse non-dominated solutions which allows the user to choose among many solutions. Although it is difficult to evaluate the importance of the various objectives quantitatively during the conceptual/preliminary stages of the design process, usually qualitative preference can be specified. This paper presents a novel, simple and intuitive way to integrate the user's preference into the evolutionary algorithm. This approach treats relatively more important objectives as adaptive constraints whose ideal values will be adaptively changed during the optimization procedure. Such changes will affect the region feasibility of the objective space which results in the variation of problem type (unconstrained problem, moderately constrained problem or highly constrained problem). As the selection criteria for mating partner depends on the type of problem in the algorithm used here, more selection pressure is put on adaptive constraints. The algorithm is validated using a target matching test problem. The results obtained indicate that the approach can produce good results at reasonable computational costs. The proposed algorithm efficiently guides the population towards the (preferred) region of interest, allowing a faster convergence and a better coverage of the preferred area of the Pareto optimal front based on the relative importance of the objectives.

This work describes a novel way of defining structural geometry for solving topology design optimization problems using a genetic algorithm (GA). It is a geometric representation scheme that works by specifying a skeleton which defines the underlying topology/connectivity of a structural continuum together with segments of material surrounding the skeleton. The required design variables are encoded in a chromosome which is in the form of a directed graph that embodies this underlying topology so that appropriate crossover and mutation operators can be devised to recombine and help preserve any desirable geometry characteristics of the design through succeeding generations in the evolutionary process. The overall methodology is first tested by solving a 'target geometry matching problem' - a simulated topology optimization problem in which a 'target' geometry is pre-defined as the optimum solution, and the objective of the optimization problem is to evolve design solutions to converge towards this 'target' shape. The second test problem is to design a complaint mechanism - a large-displacement flexural structure that undergoes a desired displacement path at some point when given a straight line input loading at some other point - by a process of topology/shape optimization.

A bit-array representation method for structural topology optimization using the GA is proposed. The importance of design connectivity is further emphasized and a hierarchical violation penalty method is proposed to penalize the violated constraint functions so that the problem of representation degeneracy can be overcome and the GA search can be driven towards the combination of better structural performance, less unusable material and fewer connected objects in the design domain. An identical initialization method is also proposed to test the performance of the GA operators. With the appropriately selected GA operators, the bit-array representation GA is applied to the structural topology optimization problems of minimum weight. Numerical results demonstrate that the present GA can achieve better accuracy with less computational cost and suggest that the GA performance can be significantly improved by handling the design connectivity properly.

Population-based, stochastic, zero-order optimization methods (e.g. genetic and evolutionary algorithms) are a popular choice in solving intractable, real-life optimization problems. These methods are particularly attractive as they are easy to use and do not require assumptions about functional and slope continuities unlike some of its gradient-based counterparts. Despite their advantages, these methods require the evaluation of numerous candidate solutions, which is often computationally expensive and practically prohibitive. We introduce a framework for optimization using approximate functions. The optimization algorithm is a population-based, stochastic, zero-order, elite-preserving algorithm that makes use of approximate function evaluations in lieu of actual function evaluations. The approximate function is constructed using a radial basis function (RBF) network and the network is periodically retrained after a few generations unlike other models which create and use the same approximate model repeatedly without retraining. A scheme for controlled elitism is incorporated within the optimization framework to ensure convergence in the actual function space. The computational accuracy and efficiency of the proposed optimization framework is assessed using a set of five mathematical test functions. The results clearly indicate that the optimization framework using approximations is able to arrive at reasonably accurate results using only a fraction of actual functions evaluations.