This work investigates some uses of self-monitoring in classifier systems (CS) using Wilson's recent XCS system as a framework. XCS is a significant advance in classifier systems technology which shifts the basis of fitness evaluation for the Genetic Algorithm (GA) from the strength of payoff prediction to the accuracy of payoff prediction. Initial work consisted of implementing an XCS system in Pop11 and replicating published XCS multiplexer experiments from (Wilson 1995, 1996a). In subsequent original work, the XCS Optimality Hypothesis, which suggests that under certain conditions XCS systems can reliably evolve optimal populations (solutions), is proposed. An optimal population is one which accurately maps inputs to actions to reward predictions using the smallest possible set of classifiers. An optimal XCS population forms a complete mapping of the payoff environment in the reinforcement learning tradition, in contrast to traditional classifier systems which only seek to maximise ...

This paper develops a reusable computing paradigm based on genetic algorithms to transform the "unsolvable problem" of optimal designs to a practically solvable "nondeterministic polynomial problem", which results in computer automated designs directly from nonlinear plants. The design methodology takes into account practical system constraints and extends the solution space, allowing new control terms to be included in the controller structure. In addition, the practical implementations using laboratory-scale systems demonstrate that such "off-the-computer" designs offer a superior performance to manual designs in terms of transient and steady-state responses and of robustness. Various contributions to the genetic algorithm technique involving the construction of fitness functions, coding, initial population formation and reproduction are also presented.

The Guided Local Search method has been successfully applied to a number of hard combinatorial optimisation problems from the well-known TSP and QAP to real world problems such as Frequency Assignment and Workforce Scheduling. In this paper, we are demonstrating that the potential applications of GLS are not limited to optimisation problems of discrete nature but also to difficult continuous optimisation problems. Continuous optimisation problems arise in many engineering disciplines (such as electrical and mechanical engineering) in the context of analysis, design or simulation tasks. The problem examined gives an illustrative example of the behaviour of GLS, providing insights on the mechanisms of the algorithm. 1.

Personnel rostering problems are highly constrained resource allocation problems.

There has been significant progress in the area of short-term scheduling of batch processes, including the solution of industrial-sized problems, in the last 20 years. The main goal of this paper is to provide an up-to-date review of the state-of-the-art in this challenging area. Main features, strengths and limitations of existing modeling and optimization techniques as well as other available major solution methods are examined through this paper. We first present a general classification for scheduling problems of batch processes as well as for the corresponding optimization models. Subsequently, the modeling of representative optimization approaches for the different problem types are introduced in detail, focusing on both discrete and continuous time models. A comparison of effectiveness and efficiency of these models is given for two benchmarking examples from the literature. We also discuss two real-world applications of scheduling problems that cannot be readily accommodated using existing methods. For the sake of completeness, other alternative solution methods applied in the field of scheduling are also reviewed, followed by a discussion related to solving large-scale problems through rigorous optimization approaches. Finally, we list available academic and commercial software and briefly address the issue of rescheduling capabilities of the various optimization approaches.

This paper develops high performance system identification and linearisation techniques, using a genetic algorithm. The algorithm is fine tuned by simulated annealing, which yields a faster convergence and a more accurate search. This global search technique is used to identify the parameters of a system described by an ARMAX model in the presence of white noise and to approximate a nonlinear multivariable system by a linear time-invariant state space model. Results obtained show that simple step input can be used for effective system identification and linearisation with much higher performance than conventional means. I. INTRODUCTION System identification techniques have been used in many fields for building accurate mathematical models of dynamic systems, based on observed input-output data. The area is already fairly mature, with various conventional techniques being developed and implemented in practical applications. However, most of the identification methods, such as those bas...

This paper develops a genetic algorithm based design automation method for linear control systems. It unifies the design and avoids the need for pre-selection of control schemes. Using this method, best performance is obtained for controllers described by a transfer function. The genetic algorithm encoded in decimal numerals is fine tuned by incorporating a simulated annealing technique for a more accurate search. It is shown that the design can be applied to both linear and nonlinear plants without manual calculations and can include practical constraints imposed upon the performance requirement. This method also allows the step of linearising nonlinear plants to be bypassed. 1. Introduction With rapid and exciting development in modern control theory and algorithms, control engineers face an increasing challenge of complexity in selecting an appropriate controller and then in optimising its coefficients to meet given performance requirements in specific practical applications. Thes...

The mathematical programming based approaches to driver scheduling have had most success, including the IMPACS and TRACS II systems developed by the authors and their associates. However there is scope for improvement in terms of computational efficiency, robustness, and capability to tackle large data sets. This paper describes on-going research into using the Genetic Algorithm (GA) approach to replace the mathematical component of TRACS II, which solves a set covering model. The question "What makes a good fit amongst potential shifts in forming a schedule?" is pursued to identify combinatorial traits associated with the data set. Such combinatorial traits are embedded into the genetic structure, so that they would play some roles in the evolutionary process. They could be effective in narrowing down the solution space and they could assist better to evaluate the fitness of individuals in the population. The mathematical component of TRACS II first solves a relaxed LP. The first stag...

This paper develops novel neural networks suitable for direct embedment within a feedback loop. The structure of the embedded system is inspired by the proportional plus derivative control with moving average excitations. A parameter vector based uniform description of the problem of neural network design is presented. Difficulties associated with traditional mathematicallyguided design methods are discussed, which leads to the development of a genetic algorithm based evolutionary design method that overcomes these difficulties and makes direct neurocontrollers possible. Techniques are also developed to optimise the architecture in the same process of parameter training, leading to Darwin neural machines. The proposed methods are verified by examples of direct neurocontroller design for a delayed linear plant and a nonlinear plant.

Abstract not found