Trojanowski K. in  analyze the efficiency of two mutation operators ap- plied in a clonal selection based optimization algorithm (AIIA) for non-stationary tasks. Both operators use a α−stable random number generator. The author ar- gues that appropriate tuning of the α parameter allows to outperform the results of algorithms with the traditional operators. The algorithms were tested with six environments generated with two test-benchmarks: a Test Case Generator and a Moving Peaks Benchmark.
In previous works the ability of CPS-MCPC (an evolutionary, co-operative, population search method with multiple crossovers per couple) to build well delineated Pareto fronts in diverse multiobjective optimization problems (MOOPs) was demonstrated. To test the potential of the novel method when dealing with the Job Shop Scheduling Problem (JSSP), regular and non-regular objectives functions were chosen. They were the makespan and the mean absolute deviation (of job completion times from a common due date, an earliness/tardiness related problem). Diverse representations such as priority list representation (PLR), job-based representation (JBR) and operation-based representation (OBR) among others were implemented and tested. The latter showed to be the best one. As a good parameter setting can enhance the behaviour of an evolutionary algorithm distinct parameters combinations were imple- mented and their influence studied. Multiple crossovers on multiple parents (MCMP), a powerful mul- tirecombination method showed some enhancement in single objective optimization when compared with MCPC.
13 Even though a simultaneous solution of the two research directions outlined above is highly desirable, we can find only a few contributions with this aim, e.g. [37-39]. Particularly in , a meta-heuristic method based on a greedy randomized adaptive search procedure (GRASP) is used to design off-grid electrification systems with distributed generation. How- ever, a non-trivial computational effort is demanded as the complexity of the system in- creases. In , a genetic algorithm-based tool is tested to solve a dynamic multistage plan- ning that aims at sizing and locating substations in distribution networks. This algorithm gen- erates satisfactory results, as long as a set of plausible substation locations and branch inter- connections are provided a priori. In , a model for active distribution systems expansion planning based on genetic algorithms is presented, where DG integration is considered to- gether with network reconfiguration. The possible drawback of this model is that only con- siders the minimization of a single objective function based on costs and cannot guarantee the network radiality and accomplishment of power quality parameters.
Given the variability of the objects and conditions in the images, it is recom- mended to use a global search method to be able to reach global correct solutions in all the possible situations. There are several global minimization approaches in literature such as the branch and bound algorithms , simulated annealing , tabu search  or evolutionary strategies . The branch and bound algorithms consist of a systematic enumeration of the whole set of candidate solutions by using super and lower estimated bounds of the quantity to optimize. Typically, these tech- niques rely on some a priori structural knowledge about the problem. The simulated annealing approach considers, at each step, some neighbors of the current state and decides, given a probability, if the system is moved to the new state or not. It allows non optimal states in the search. This strategy finds a good approximation of the global minimum, but it does not guarantee the global optimum. The tabu search for- bids states already visited in the search space at least for the upcoming few events.
In this paper we have proposed an analytical model that accurately predicts the acoustic power maps gener- ated by the SRP-PHAT algorithm in both anechoic and non- anechoic conditions. It is based on reasonable assumptions about sound propagation and its interaction with the environment. Our model predicts that SRP-PHAT (and the corresponding GCC-PHAT functions from which SRP- PHAT is calculated) depends on the topology of the array, room's geometry and signal bandwidth, but not on the spectral content of the signal. This last property is very important in speaker localization as the speech signal is unknown. Our model allows us to discuss the influence of all these factors in the localization accuracy, specially in reverberant scenarios. The model has been thoroughly validated using simulated and real data for a wide range of conditions (speaker positions, bandwidth and array topology considerations). In the synthetic case we show that our model predictions are very close to that provided by the image method, a standard room acoustics simulator. We also tested our model with real data from a publicly available dataset. Our results are reproducible and verify empirically that the model is able to reproduce SRP-PHAT power maps with high fidelity in a real case.
Previous works on MEP have focused, for instan- ce, on the construction of the controller itself, dif- ferent parameters for an elevator system, and re- optimization. For example, the construction of con- trollers itself, Ho et al.  propose a combination of Petri Nets with Neural Networks to learn the best possible scheduling policy, Kojima et al.  apply ADN-Computation to minimize the waiting times. More recently, some evolutionary approaches have been proposed, e.g., Markon et al.  developed a Genetic Network Programming to evolve a controler for a multiple elevator system. Other approaches ap- ply fuzzy controllers, expert systems, neural networks and several combinations of them [1, 4, 5, 6]. With respect to the techniques towards to the optimization of certain parameters for a controller, it is assumed that a control unit takes into account a number of pa- rameter values to make the more appropriate decision about which elevator should be assigned when the ele- vators are called. Accordingly, the objective in this context is determine the best parameter values in or- der to optimize a number of criteria, e.g., minimiza- tion of waiting times, reduction of the crowding fac- tor, minimization of the riding time, etc. In this di- rection, Fujino et al. [11, 3] show the application of genetic algorithms to find the best parameter setting used in the controller for a multiple elevator system.
Abstract. Many distributed systems (task scheduling, moving priori- ties, changing mobile environments, ...) can be linked as Dynamic Opti- mization Problems (DOPs), since they require to pursue an optimal value that changes over time. Consequently, we have focused on the utilization of Distributed Genetic Algorithms (dGAs), one of the domains still to be investigated for DOPs. A dGA essentially decentralizes the population in islands which cooperate through migrations of individuals. In this ar- ticle, we analyze the effect of the migrants selection and replacement on the performance of the dGA for DOPs. Quality and distance based cri- teria are tested using a comprehensive set of benchmarks. Results show the benefits and drawbacks of each setting in dynamic optimization.
Non–stationary, or dynamic, problems change over time. There exist a variety of forms of dynamism. The concept of dynamic environmentsin the context of this paper means that the fitness landscape changes during the run of an evolutionary algorithm. Genetic diversity is crucial to provide the necessary adaptability of the algorithm to changes. Two mechanism of macromutation are incorporated to the algorithm to maintain genetic diversity in the population. The algorithm was tested on a set of dynamic testing functions provided by a dynamic fitness problem generator. The main goal was to determinate the algorithm’s ability to reacting to changes of optimum values that alter their locations, so that the optimum value can still be tracked when dimensional and multimodal scalability in the functions is adjusted. The effectiveness and limitations of the proposed algorithm is discussed from results empirically obtained.
In general the conditions of an optimization problem changes by one of the following reasons or a combination of both : 1) The objective function changes itself, 2) The constraints change. A change in the objective function appears when the purpose of the problem changes. Here conditions which were considered desirable before can turn out to be undesirable now and vice versa. Changes in constraints, which modify feasibility of solutions, are related to resources and their availability. Changes can be small or big, soft or abrupt, chaotic, etc. When changes are big, abrupt or chaotic the similarity between solutions found so far and the new ones can be worthless. Even under these hard environmentsEvolutionary Computation (EC) offers advantages, which are absent in other heuristics when searching for solutions to non-stationary problems. The main advantage relies in the fact that Evolutionary Algorithms (EAs) keep a population of solutions. Consequently, facing the change, they allow moving from a solution to another one to determine if any of them are of merit to continue the search from them instead of from scratch . Goldberg and Smith , Cobb and Grefenstette  initiated the research related to the behaviour of EAS on dynamic fitness functions between 1987 and 1992. Recently the interest in this area was dramatically incremented , , , , , , , , , and . The following sections are organized as follows. Section 2 presents a definition of dynamic environments studied in this work. Section 3 describes the dynamic test functions used. Section 4 describes the EA. In section 5 the experiments performed are described. In section 6 results are discussed and finally this document shows our conclusions, current and future work.
➢ Frequency hopping (FH): Frequency agile radar transmission, either on a pulse to pulse basis or on bursts of pulses. Additionally, PRI can be made agile. Pulse to pulse agility gives ECM protection which is proportional to the agile bandwidth, defeating the repeater jammer. It works under the control of pseudo- noise (PN) code. It’s one of the favored phase modulation technique for generating spread-spectrum waveform in which the transmitted RF bandwidth is controlled directly by the PN code clock rate. As reference fast FH techniques used more than 500 hops/s, and medium hop-rate FH systems used between 50-500 hops/s .
ity of the proposal (for a very standard set of parame- ters) as a simplified algorithm to be implemented on an FPGA, where performance is sacrificed up to an extent. Further tests for a wider variation of all the parameters in the algorithm shall be done in order to improve it while keeping the required computing resources. Besides, different operators for the ES shall be tested, as for example, an even simpler mutation scheme, the two-point rule . But special care should be taken here since it has been observed during the several runs of the algorithm that it is somehow prone to stagnation at local optima. Therefore, as suggested in the literature of ES, a boundary rule to avoid this behavior shall be observed. It is also worth to try a different approach for the recombination of both the strategy and object parameters, mixing intermediate with discrete recombination, as suggested in . As for the evaluation phase, a different method will try to be developed to avoid the stagnation of the evolution. The methodology followed allows these modifications to be more easily tested than if directly implemented in hardware.
principles are related to the function of the product, so they can be applied more directly. The absence of direct functional information about the overall characteristics of the design, during the optimization process, could reduce the possibilities of a valid solution. On complex geometries or time consuming optimizations where manufacturing constraints play an important role, more restricted TRIZ interpretations could be the best suited . TRIZ concepts could prove useful in suggesting modification possibilities based on some of the TRIZ innovation principles. Such principles can be identified by generalizing existing design contradictions in the given part to obtain suggestions from a predefined contradiction table. These suggestions should enunciate the modification principle along with a design-oriented example depicted by CAD models. Other inventive principles are of a rather topological nature and therefore may be implemented in CAD systems' assembly modules. In other cases the principles are of a mechanical or physical nature, which also involves the effect of time and other physical parameters such as velocity, force, acceleration, temperature, etc. and may be implemented using multi-body systems. Normally a graphic description along with a picture or drawing depicting the given suggestion is also provided. With these examples, the user would have a much better idea about where and how the shape modification process should be focused. However, the designer has to implement the required modifications on his or her own by editing step by step the actual shapes and topologies based on how he or she understands the recommendation. This is commonly a time-consuming task that avoids the search for better solutions.
7. CONCLUSION AND FUTURE WORKS We have modeled the signal strength of indoor and outdoor WLAN environments, by studying the propagation characteristics of both environments, considering fading effects as well as obstacles. In tackling these issues, we took measurements of QoS parameters over a defined distance and compute the pathloss for both environments. An optimized model that predicts the signal quality of WLANs over various distances was then derived from the experiment. Experimental data obtained from the field were used to validate the models. In both environments, we could predict the distance at which wireless signal quality received was optimal. We observed that the link quality (SNR) degrades with distance and other environmental factors. The fading phenomenon was also approached with the aim of proffering a practical solution. This approach was accomplished by monitoring the signal quality coverage (at various distances). As an outlook, we shall investigate indoor localization and the problem of interference.
ABSTRACT. When nutrients become scarce E. coli cells enter into a non-growth phase known as stationary and develop a multiple-stress resistance state analogue to sporulation in B. subtilis. Morphological changes are observed, including rounded shape, loss of flagella and thickening of the cell wall. General metabolism is re-directed, macro- molecular degradation is increased, and storage and osmoprotection compounds are synthesized. The reorganization of the nucleoid is ac- companied by an overall repression of gene expression, but a subset of genes required for starvation survival become transcribed in a manner dependent on the stationary phase-specific subunit of RNA polymerase (RpoS or σ s ). The regulatory function of σ s seems to be
As EAs are blind search methods our new variant (MCMP-SRSI) , proposes to insert problem-specific knowledge by recombining potential solutions (individuals of the evolving population) with seeds, which are solutions provided by other heuristics specifically designed to solve the scheduling problems under study. In MCMP-SRSI, the process for creating offspring is similar to that of MCMP-SRI, except that the mating pool contains also seed immigrants. In this way, the evolutionary algorithm incorporates problem-specific knowledge supplied by the specific heuristic. Figure 1 displays these processes.
The mathematical origin of the problem. This optimization question, we are dealing with, comes from a central problem in the field of the symbolic com- putation of algebraic curves (see  for further details), appears in many compu- tational aspects of the practical applications of curves and is, to our knowledge, not solved. Let us first motivate the problem: In many practical applications, such as in computed aided geometric design, in physics, etc., one deals with parametric representations of a curve. For instance, if we have to compute a line integral along an arc of the curve of equation y 3 = x 2 , we might use the parametric representation
We used a steady-state evolutionary algorithm, which objective is to maximize the fitness. That is to maximize the total value of cut pieces by applying a penalty function. This algorithm uses: binary solution representation , the uniform crossover and big-creep mutation. The created child ever should replace an individual from the population; the individual to be replaced is selected randomly from the ten worst individuals. Two diferent methods to select parents are used, one of them is the Binary Tournament and another is the Roulette Wheel method. For using the last method is necessary scaling the fitness because sometimes the fitness is negative. The used technique for scaling is the Normalizing technique, which was proposed by Cheng and Gen in . For a maximization problem, it takes the following form:
By means of a stylized example of an inventory management problem that is relevant for practice we study whether SDP is a suitable method if a minimal service level constraint applies and demand is non-stationary. The example has a fixed time horizon is six periods. We started to solve a deterministic problem, where the variable demand equals the expected demand in the stochastic cases. Introducing uncertainty through a Uniform distribution and requiring that all demand has to be fulfilled gives a 75% cost increase. Subsequently we introduced an α-service level constraint and a fill rate constraint per period, being minimal service level constraints.