Das and Whitley have shown that functions which are easy in theory for geneticalgorithms (GA) can be easily optimized with a global search method (Das and Whitley 1991). In this paper, we show that this global search...
详细信息
Das and Whitley have shown that functions which are easy in theory for geneticalgorithms (GA) can be easily optimized with a global search method (Das and Whitley 1991). In this paper, we show that this global search method can be simply adapted in order to solve also GA-hard functions and more generally consistently deceptive functions. The resulting algorithm, called GSC1, generates a set of binary strings and outputs the string that wins the first order schemas competitions as well as its binary complement. According to the theory of deceptiveness in GAs, this method solves GA-easy and GA-hard problems efficiently, as shown effectively in the reported experiments. This method is however only well suited for these functions, and does not deal with partially deceptive functions. It is then shown how it could be combined with a GA.
This paper presents a genetic algorithm for the design of an optimal mesh network. The problem is of relevance in the design of communication networks where the backbone switching network takes the form of a highly co...
详细信息
This paper presents a genetic algorithm for the design of an optimal mesh network. The problem is of relevance in the design of communication networks where the backbone switching network takes the form of a highly connected mesh in order to provide reliability in the event of switch/link failure. The proposed algorithm addresses two important aspects of the problem - topology design and capacity allocation. The optimization is done with respect to connection costs subject to performance (delay), connectivity and capacity constraints. Connection costs are assumed to depend on distance and link capacity. Though the algorithm was designed with mesh networks in mind, it can be applied to the simpler problem of the constrained minimum spanning tree. The algorithm has been tested on a tree network and two mesh networks. The results compare very favourably with those obtained from existing design techniques.
This paper introduces a niching technique called GAS which dynamically creates a subpopulation structure (taxonomic chart) using a radius function instead of a single radius, and a `cooling' method similar to simu...
详细信息
This paper introduces a niching technique called GAS which dynamically creates a subpopulation structure (taxonomic chart) using a radius function instead of a single radius, and a `cooling' method similar to simulated annealing. GAS offers a solution to the niche radius problem with the help of these techniques. A method based on the speed of species is presented for determining the radius function. Speed functions are given for both real and binary domains. Finally we discuss the sphere packing problem on binary domains using some tools of coding theory to make it possible to evaluate the output of the system.
The paper is devoted to the problem of learning decision policies in multi-agent games. This problem is a simple but appealing computational model of several important real-world problems in such domains as parallel c...
详细信息
The paper is devoted to the problem of learning decision policies in multi-agent games. This problem is a simple but appealing computational model of several important real-world problems in such domains as parallel computing, optimization, and control on one hand, and economy, social, and political sciences on the other hand. We describe a general framework for studying games of intelligent agents, extending the basic model of games with limited interactions, and its specific realization based on learning classifier systems. Simulation results are presented that illustrate the convergence properties of the resulting system. Avenues for future work in this area are outlined.
Effective optimal power system planning and operation is limited by: (i) the high dimensionality of power systems; and (ii) the incomplete domain dependent knowledge of power system engineers. The first limitation is ...
详细信息
Effective optimal power system planning and operation is limited by: (i) the high dimensionality of power systems; and (ii) the incomplete domain dependent knowledge of power system engineers. The first limitation is addressed by numerical optimisation procedures using gradient approximations to calculate the search directions in various nonlinear programming formulations or by linear programming solutions to imprecise models. The advantages of such methods are in their mathematical underpinnings, but disadvantages exist also in the sensitivity to problem formulation, algorithm selection and undue focus on local minima. Here, the author examines the use of geneticalgorithms in addressing the above problems.
The paper proposes a novel constraint handling technique to improve the efficiency with geneticalgorithms, which could effectively attack the problem of economic-environmental power dispatch (EED) under the constrain...
详细信息
The paper proposes a novel constraint handling technique to improve the efficiency with geneticalgorithms, which could effectively attack the problem of economic-environmental power dispatch (EED) under the constraint of generation rate. EED is a sophisticated and difficult task because of the conflicting requirements of minimizing generation cost and reducing environmental pollution. The problem is made even more formidable with the additional constraint of generation rate. geneticalgorithms (GAs) are a promising approach to solving the highly constrained multi-objective problem involved in the solution. However, the performance is affected crucially by the way a constraint is handled in the GA implementation. The conventional method, which passes only a constraint equation to the fitness function, exhibits great difficulties for a GA to reach the global optimum. In this paper, the proposed constraint handling method incorporates the concept of how far a solution is away from the feasible region so as to improve the genetic search ability. To enhance the GA performance further, the paper also employs a unique fitness function formulation alongside the proposed constraint handling technique. The modified GA is applied to the highly constrained EED problem on a four generator system. The study shows that the proposed distance based constraint handling is performed better than the conventional equation based method, and could provide easy and efficient means to attack the difficult problem presented by economic - environmental dispatch.
The authors review the developments in artificial intelligence techniques for power system voltage control. They discuss: expert systems; artificial neural nets; hybrid systems of expert systems and neural nets; fuzzy...
详细信息
The authors review the developments in artificial intelligence techniques for power system voltage control. They discuss: expert systems; artificial neural nets; hybrid systems of expert systems and neural nets; fuzzy control; and geneticalgorithms. The advantages of these technologies as well some of the drawbacks are discussed.
A genetics Based Machine Learning (GBML) method is proposed and analyzed for learning and enhancing the control of a microrobot with stepping motor drives. This approach tries to combine several advantages of fuzzy lo...
详细信息
A genetics Based Machine Learning (GBML) method is proposed and analyzed for learning and enhancing the control of a microrobot with stepping motor drives. This approach tries to combine several advantages of fuzzy logic and genetics-based machine learning using slightly modified classifier systems, and then a discussion is presented about the learning capabilities of the proposed control system. The PID gains of a conventional controller were tuned run-time in order to minimize the effect of the nonlinear disturbances (nonlinear variable load torque applied to the controlled plant). The tuning is based on a predictive estimation method of the controller's gains, performed by a GA driven fuzzy classifier system, which has to evolve an adequate rule set to tune properly the controller's gains.
This paper describes a genetic Algorithm (GA) for the design of optimal finite word length (FWL) infinite impulse response filters with arbitrary response functions using a cascade of second order sections. Such struc...
详细信息
This paper describes a genetic Algorithm (GA) for the design of optimal finite word length (FWL) infinite impulse response filters with arbitrary response functions using a cascade of second order sections. Such structures could be used for the implementation of filters for an unrestricted range of responses and are usually characterized by having low sensitivity to variations in word length and simple stability tests. A distinct feature of the GA is that it allows the single-step-design from a specified filter template hence avoiding additional errors being introduced as is the case with the two-stage-design approach. Depending on the specified template, the GA could be used to design filters with arbitrary response functions including those of the main classical filters with considerably reduced coefficient wordlengths. The paper commences by highlighting FWL effects in digital filters and reviewing existing attempts in using GAs in digital filter design. Next, the paper describes the implementation technique adopted here and reports on the results obtained using examples of 8th and 10th order filters.
Electrical distribution networks are built as interconnected, meshed networks, while in the operation they are arranged into radial, tree structures. The network configuration problem is to find a radial operating str...
详细信息
Electrical distribution networks are built as interconnected, meshed networks, while in the operation they are arranged into radial, tree structures. The network configuration problem is to find a radial operating structure that optimises the network performance while satisfying operating constraints. In fact, this problem can be viewed as the problem of determining an `optimal' tree of the given graph. The problem is usually formulated as a constrained multi-objective combinatorial problem that belongs to the class of very large non-linear mixed-integer problems. In this paper a fuzzy co-ordination technique is used to identify overall degree of satisfaction, while GA is employed to maximize it. The major difficulty in the application of GA to this problem is that crossovers normally generate infeasible solutions, as random combination of parts of different trees of the same graph do not create a new three of that graph. The constraints are enforced by adjusting the new string to the `nearest' three which, is in terms of GA, adequate with performing `mutation' at each crossover. The results obtained by the proposed algorithm are compared with the previously developed `greedy' algorithms on a slightly changed real distribution system, showing that GA has a potential to address some other planning and operation problems in electrical distribution systems.
暂无评论