Genetic fuzzy-based steering wheel controller using a mass-produced car

Intelligent Transportation Systems (ITS) cover a broad range of methods and technologies that provide answers to many problems of transportation. Unmanned control of the steering wheel is one of the most important challenges facing researchers in this area. This paper presents a method to adjust automatically a fuzzy controller to manage the steering wheel of a mass-produced vehicle to reproduce the steering of a human driver. To this end, information is recorded about the car's state while being driven by human drivers and used to obtain, via genetic algorithms, appropriate fuzzy controllers that can drive the car in the way that humans do. These controllers have satisfy two main objectives: to reproduce the human behavior, and to provide smooth actions to ensure comfortable driving. Finally, the results of automated driving on a test circuit are presented, showing both good route tracking (similar to the performance obtained by persons in the same task) and smooth driving.

The experiences of learner drivers, provisional drivers and supervisors with graduated driver licensing in two Australian jurisdictions

Newly licensed drivers on a provisional or intermediate licence have the highest crash risk when compared with any other group of drivers. In comparison, learner drivers have the lowest crash risk. Graduated driver licensing is one countermeasure that has been demonstrated to effectively reduce the crashes of novice drivers. This thesis examined the graduated driver licensing systems in two Australian states in order to better understand the behaviour of learner drivers, provisional drivers and the supervisors of learner drivers. By doing this, the thesis investigated the personal, social and environmental influences on novice driver behaviour as well as providing effective baseline data against which to measure subsequent changes to the licensing systems. In the first study, conducted prior to the changes to the graduated driver licensing system introduced in mid-2007, drivers who had recently obtained their provisional licence in Queensland and New South Wales were interviewed by telephone regarding their experiences while driving on their learner licence. Of the 687 eligible people approached to participate at driver licensing centres, 392 completed the study representing a response rate of 57.1 per cent. At the time the data was collected, New South Wales represented a more extensive graduated driver licensing system when compared with Queensland. The results suggested that requiring learners to complete a mandated number of hours of supervised practice impacts on the amount of hours that learners report completing. While most learners from New South Wales reported meeting the requirement to complete 50 hours of practice, it appears that many stopped practising soon after this goal was achieved. In contrast, learners from Queensland, who were not required to complete a specific number of hours at the time of the survey, tended to fall into three groups. The first group appeared to complete the minimum number of hours required to pass the test (less than 26 hours), the second group completed 26 to 50 hours of supervised practice while the third group completed significantly more practice than the first two groups (over 100 hours of supervised practice). Learner drivers in both states reported generally complying with the road laws and were unlikely to report that they had been caught breaking the road rules. They also indicated that they planned to obey the road laws once they obtained their provisional licence. However, they were less likely to intend to comply with recommended actions to reduce crash risk such as limiting their driving at night. This study also identified that there were relatively low levels of unaccompanied driving (approximately 15 per cent of the sample), very few driving offences committed (five per cent of the sample) and that learner drivers tended to use a mix of private and professional supervisors (although the majority of practice is undertaken with private supervisors). Consistent with the international literature, this study identified that very few learner drivers had experienced a crash (six per cent) while on their learner licence. The second study was also conducted prior to changes to the graduated driver licensing system and involved follow up interviews with the participants of the first study after they had approximately 21 months driving experience on their provisional licence. Of the 392 participants that completed the first study, 233 participants completed the second interview (representing a response rate of 59.4 per cent). As with the first study, at the time the data was collected, New South Wales had a more extensive graduated driver licensing system than Queensland. For instance, novice drivers from New South Wales were required to progress through two provisional licence phases (P1 and P2) while there was only one provisional licence phase in Queensland. Among the participants in this second study, almost all provisional drivers (97.9 per cent) owned or had access to a vehicle for regular driving. They reported that they were unlikely to break road rules, such as driving after a couple of drinks, but were also unlikely to comply with recommended actions, such as limiting their driving at night. When their provisional driving behaviour was compared to the stated intentions from the first study, the results suggested that their intentions were not a strong predictor of their subsequent behaviour. Their perception of risk associated with driving declined from when they first obtained their learner licence to when they had acquired provisional driving experience. Just over 25 per cent of participants in study two reported that they had been caught committing driving offences while on their provisional licence. Nearly one-third of participants had crashed while driving on a provisional licence, although few of these crashes resulted in injuries or hospitalisations. To complement the first two studies, the third study examined the experiences of supervisors of learner drivers, as well as their perceptions of their learner’s experiences. This study was undertaken after the introduction of the new graduated driver licensing systems in Queensland and New South Wales in mid- 2007, providing insights into the impacts of these changes from the perspective of supervisors. The third study involved an internet survey of 552 supervisors of learner drivers. Within the sample, approximately 50 per cent of participants supervised their own child. Other supervisors of the learner drivers included other parents or stepparents, professional driving instructors and siblings. For two-thirds of the sample, this was the first learner driver that they had supervised. Participants had provided an average of 54.82 hours (sd = 67.19) of supervision. Seventy-three per cent of participants indicated that their learners’ logbooks were accurate or very accurate in most cases, although parents were more likely than non-parents to report that their learners’ logbook was accurate (F (1,546) = 7.74, p = .006). There was no difference between parents and non-parents regarding whether they believed the log book system was effective (F (1,546) = .01, p = .913). The majority of the sample reported that their learner driver had had some professional driving lessons. Notwithstanding this, a significant proportion (72.5 per cent) believed that parents should be either very involved or involved in teaching their child to drive, with parents being more likely than non-parents to hold this belief. In the post mid-2007 graduated driver licensing system, Queensland learner drivers are able to record three hours of supervised practice in their log book for every hour that is completed with a professional driving instructor, up to a total of ten hours. Despite this, there was no difference identified between Queensland and New South Wales participants regarding the amount of time that they reported their learners spent with professional driving instructors (X2(1) = 2.56, p = .110). Supervisors from New South Wales were more likely to ensure that their learner driver complied with the road laws. Additionally, with the exception of drug driving laws, New South Wales supervisors believed it was more important to teach safety-related behaviours such as remaining within the speed limit, car control and hazard perception than those from Queensland. This may be indicative of more intensive road safety educational efforts in New South Wales or the longer time that graduated driver licensing has operated in that jurisdiction. However, other factors may have contributed to these findings and further research is required to explore the issue. In addition, supervisors reported that their learner driver was involved in very few crashes (3.4 per cent) and offences (2.7 per cent). This relatively low reported crash rate is similar to that identified in the first study. Most of the graduated driver licensing research to date has been applied in nature and lacked a strong theoretical foundation. These studies used Akers’ social learning theory to explore the self-reported behaviour of novice drivers and their supervisors. This theory was selected as it has previously been found to provide a relatively comprehensive framework for explaining a range of driver behaviours including novice driver behaviour. Sensation seeking was also used in the first two studies to complement the non-social rewards component of Akers’ social learning theory. This program of research identified that both Akers’ social learning theory and sensation seeking were useful in predicting the behaviour of learner and provisional drivers over and above socio-demographic factors. Within the first study, Akers’ social learning theory accounted for an additional 22 per cent of the variance in learner driver compliance with the law, over and above a range of socio-demographic factors such as age, gender and income. The two constructs within Akers’ theory which were significant predictors of learner driver compliance were the behavioural dimension of differential association relating to friends, and anticipated rewards. Sensation seeking predicted an additional six per cent of the variance in learner driver compliance with the law. When considering a learner driver’s intention to comply with the law while driving on a provisional licence, Akers’ social learning theory accounted for an additional 10 per cent of the variance above socio-demographic factors with anticipated rewards being a significant predictor. Sensation seeking predicted an additional four per cent of the variance. The results suggest that the more rewards individuals anticipate for complying with the law, the more likely they are to obey the road rules. Further research is needed to identify which specific rewards are most likely to encourage novice drivers’ compliance with the law. In the second study, Akers’ social learning theory predicted an additional 40 per cent of the variance in self-reported compliance with road rules over and above socio-demographic factors while sensation seeking accounted for an additional five per cent of the variance. A number of Aker’s social learning theory constructs significantly predicted provisional driver compliance with the law, including the behavioural dimension of differential association for friends, the normative dimension of differential association, personal attitudes and anticipated punishments. The consistent prediction of additional variance by sensation seeking over and above the variables within Akers’ social learning theory in both studies one and two suggests that sensation seeking is not fully captured within the non social rewards dimension of Akers’ social learning theory, at least for novice drivers. It appears that novice drivers are strongly influenced by the desire to engage in new and intense experiences. While socio-demographic factors and the perception of risk associated with driving had an important role in predicting the behaviour of the supervisors of learner drivers, Akers’ social learning theory provided further levels of prediction over and above these factors. The Akers’ social learning theory variables predicted an additional 14 per cent of the variance in the extent to which supervisors ensured that their learners complied with the law and an additional eight per cent of the variance in the supervisors’ provision of a range of practice experiences. The normative dimension of differential association, personal attitudes towards the use of professional driving instructors and anticipated rewards were significant predictors for supervisors ensuring that their learner complied with the road laws, while the normative dimension was important for range of practice. This suggests that supervisors who engage with other supervisors who ensure their learner complies with the road laws and provide a range of practice to their own learners are more likely to also engage in these behaviours. Within this program of research, there were several limitations including the method of recruitment of participants within the first study, the lower participation rate in the second study, an inability to calculate a response rate for study three and the use of self-report data for all three studies. Within the first study, participants were only recruited from larger driver licensing centres to ensure that there was a sufficient throughput of drivers to approach. This may have biased the results due to the possible differences in learners that obtain their licences in locations with smaller licensing centres. Only 59.4 per cent of the sample in the first study completed the second study. This may be a limitation if there was a common reason why those not participating were unable to complete the interview leading to a systematic impact on the results. The third study used a combination of a convenience and snowball sampling which meant that it was not possible to calculate a response rate. All three studies used self-report data which, in many cases, is considered a limitation. However, self-report data may be the only method that can be used to obtain some information. This program of research has a number of implications for countermeasures in both the learner licence phase and the provisional licence phase. During the learner phase, licensing authorities need to carefully consider the number of hours that they mandate learner drivers must complete before they obtain their provisional driving licence. If they mandate an insufficient number of hours, there may be inadvertent negative effects as a result of setting too low a limit. This research suggests that logbooks may be a useful tool for learners and their supervisors in recording and structuring their supervised practice. However, it would appear that the usage rates for logbooks will remain low if they remain voluntary. One strategy for achieving larger amounts of supervised practice is for learner drivers and their supervisors to make supervised practice part of their everyday activities. As well as assisting the learner driver to accumulate the required number of hours of supervised practice, it would ensure that they gain experience in the types of environments that they will probably encounter when driving unaccompanied in the future, such as to and from education or work commitments. There is also a need for policy processes to ensure that parents and professional driving instructors communicate effectively regarding the learner driver’s progress. This is required as most learners spend at least some time with a professional instructor despite receiving significant amounts of practice with a private supervisor. However, many supervisors did not discuss their learner’s progress with the driving instructor. During the provisional phase, there is a need to strengthen countermeasures to address the high crash risk of these drivers. Although many of these crashes are minor, most involve at least one other vehicle. Therefore, there are social and economic benefits to reducing these crashes. If the new, post-2007 graduated driver licensing systems do not significantly reduce crash risk, there may be a need to introduce further provisional licence restrictions such as separate night driving and peer passenger restrictions (as opposed to the hybrid version of these two restrictions operating in both Queensland and New South Wales). Provisional drivers appear to be more likely to obey some provisional licence laws, such as lower blood alcohol content limits, than others such as speed limits. Therefore, there may be a need to introduce countermeasures to encourage provisional drivers to comply with specific restrictions. When combined, these studies provided significant information regarding graduated driver licensing programs. This program of research has investigated graduated driver licensing utilising a cross-sectional and longitudinal design in order to develop our understanding of the experiences of novice drivers that progress through the system in order to help reduce crash risk once novice drivers commence driving by themselves.

Construction of Neural Network Classification Expert Systems Using Switching Theory Algorithms

A new family of neural network architectures is presented. This family of architectures solves the problem of constructing and training minimal neural network classification expert systems by using switching theory. The primary insight that leads to the use of switching theory is that the problem of minimizing the number of rules and the number of IF statements (antecedents) per rule in a neural network expert system can be recast into the problem of minimizing the number of digital gates and the number of connections between digital gates in a Very Large Scale Integrated (VLSI) circuit. The rules that the neural network generates to perform a task are readily extractable from the network's weights and topology. Analysis and simulations on the Mushroom database illustrate the system's performance.

Evolving Distributed Algorithms with Genetic Programming

Distributed systems are one of the most vital components of the economy. The most prominent example is probably the internet, a constituent element of our knowledge society. During the recent years, the number of novel network types has steadily increased. Amongst others, sensor networks, distributed systems composed of tiny computational devices with scarce resources, have emerged. The further development and heterogeneous connection of such systems imposes new requirements on the software development process. Mobile and wireless networks, for instance, have to organize themselves autonomously and must be able to react to changes in the environment and to failing nodes alike. Researching new approaches for the design of distributed algorithms may lead to methods with which these requirements can be met efficiently. In this thesis, one such method is developed, tested, and discussed in respect of its practical utility. Our new design approach for distributed algorithms is based on Genetic Programming, a member of the family of evolutionary algorithms. Evolutionary algorithms are metaheuristic optimization methods which copy principles from natural evolution. They use a population of solution candidates which they try to refine step by step in order to attain optimal values for predefined objective functions. The synthesis of an algorithm with our approach starts with an analysis step in which the wanted global behavior of the distributed system is specified. From this specification, objective functions are derived which steer a Genetic Programming process where the solution candidates are distributed programs. The objective functions rate how close these programs approximate the goal behavior in multiple randomized network simulations. The evolutionary process step by step selects the most promising solution candidates and modifies and combines them with mutation and crossover operators. This way, a description of the global behavior of a distributed system is translated automatically to programs which, if executed locally on the nodes of the system, exhibit this behavior. In our work, we test six different ways for representing distributed programs, comprising adaptations and extensions of well-known Genetic Programming methods (SGP, eSGP, and LGP), one bio-inspired approach (Fraglets), and two new program representations called Rule-based Genetic Programming (RBGP, eRBGP) designed by us. We breed programs in these representations for three well-known example problems in distributed systems: election algorithms, the distributed mutual exclusion at a critical section, and the distributed computation of the greatest common divisor of a set of numbers. Synthesizing distributed programs the evolutionary way does not necessarily lead to the envisaged results. In a detailed analysis, we discuss the problematic features which make this form of Genetic Programming particularly hard. The two Rule-based Genetic Programming approaches have been developed especially in order to mitigate these difficulties. In our experiments, at least one of them (eRBGP) turned out to be a very efficient approach and in most cases, was superior to the other representations.

Los conceptos de salud financiera, riesgo y epidemiología en los estados financieros de las compañías del sector de extracción de petróleo crudo y gas natural en Colombia

El presente proyecto tiene como objeto identificar cuáles son los conceptos de salud, enfermedad, epidemiología y riesgo aplicables a las empresas del sector de extracción de petróleo y gas natural en Colombia. Dado, el bajo nivel de predicción de los análisis financieros tradicionales y su insuficiencia, en términos de inversión y toma de decisiones a largo plazo, además de no considerar variables como el riesgo y las expectativas de futuro, surge la necesidad de abordar diferentes perspectivas y modelos integradores. Esta apreciación es pertinente dentro del sector de extracción de petróleo y gas natural, debido a la creciente inversión extranjera que ha reportado, US$2.862 millones en el 2010, cifra mayor a diez veces su valor en el año 2003. Así pues, se podrían desarrollar modelos multi-dimensional, con base en los conceptos de salud financiera, epidemiológicos y estadísticos. El termino de salud y su adopción en el sector empresarial, resulta útil y mantiene una coherencia conceptual, evidenciando una presencia de diferentes subsistemas o factores interactuantes e interconectados. Es necesario mencionar también, que un modelo multidimensional (multi-stage) debe tener en cuenta el riesgo y el análisis epidemiológico ha demostrado ser útil al momento de determinarlo e integrarlo en el sistema junto a otros conceptos, como la razón de riesgo y riesgo relativo. Esto se analizará mediante un estudio teórico-conceptual, que complementa un estudio previo, para contribuir al proyecto de finanzas corporativas de la línea de investigación en Gerencia.

Mapping and controlling molecular processes: The application of control theory and system identification algorithms to ASOPS and other fast pulse spectroscopies

Asynchronous Optical Sampling (ASOPS) [1,2] and frequency comb spectrometry [3] based on dual Ti:saphire resonators operated in a master/slave mode have the potential to improve signal to noise ratio in THz transient and IR sperctrometry. The multimode Brownian oscillator time-domain response function described by state-space models is a mathematically robust framework that can be used to describe the dispersive phenomena governed by Lorentzian, Debye and Drude responses. In addition, the optical properties of an arbitrary medium can be expressed as a linear combination of simple multimode Brownian oscillator functions. The suitability of a range of signal processing schemes adopted from the Systems Identification and Control Theory community for further processing the recorded THz transients in the time or frequency domain will be outlined [4,5]. Since a femtosecond duration pulse is capable of persistent excitation of the medium within which it propagates, such approach is perfectly justifiable. Several de-noising routines based on system identification will be shown. Furthermore, specifically developed apodization structures will be discussed. These are necessary because due to dispersion issues, the time-domain background and sample interferograms are non-symmetrical [6-8]. These procedures can lead to a more precise estimation of the complex insertion loss function. The algorithms are applicable to femtosecond spectroscopies across the EM spectrum. Finally, a methodology for femtosecond pulse shaping using genetic algorithms aiming to map and control molecular relaxation processes will be mentioned.

Defects location in power systems components through a dedicated genetic algorithm

In this work the problem of defects location in power systems is formulated through a binary linear programming (BLP) model based on alarms historical database of control and protection devices from the system control center, sets theory of minimal coverage (AI) and protection philosophy adopted by the electric utility. In this model, circuit breaker operations are compared to their expected states in a strictly mathematical manner. For solving this BLP problem, which presents a great number of decision variables, a dedicated Genetic Algorithm (GA), is proposed. Control parameters of the GA, such as crossing over and mutation rates, population size, iterations number and population diversification, are calibrated in order to obtain efficiency and robustness. Results for a test system found in literature, are presented and discussed. © 2004 IEEE.

Data Processing Systems Manual: Highway Noise Prediction, 1972

The Data Processing Department of ISHC has developed coding forms to be used for the data to be entered into the program. The Highway Planning and Programming and the Design Departments are responsible for coding and submitting the necessary data forms to Data Processing for the noise prediction on the highway sections.

'Explicit Learning: an Effort towards Human Scheduling Algorithms'

Abstract Scheduling problems are generally NP-hard combinatorial problems, and a lot of research has been done to solve these problems heuristically. However, most of the previous approaches are problem-specific and research into the development of a general scheduling algorithm is still in its infancy. Mimicking the natural evolutionary process of the survival of the fittest, Genetic Algorithms (GAs) have attracted much attention in solving difficult scheduling problems in recent years. Some obstacles exist when using GAs: there is no canonical mechanism to deal with constraints, which are commonly met in most real-world scheduling problems, and small changes to a solution are difficult. To overcome both difficulties, indirect approaches have been presented (in [1] and [2]) for nurse scheduling and driver scheduling, where GAs are used by mapping the solution space, and separate decoding routines then build solutions to the original problem. In our previous indirect GAs, learning is implicit and is restricted to the efficient adjustment of weights for a set of rules that are used to construct schedules. The major limitation of those approaches is that they learn in a non-human way: like most existing construction algorithms, once the best weight combination is found, the rules used in the construction process are fixed at each iteration. However, normally a long sequence of moves is needed to construct a schedule and using fixed rules at each move is thus unreasonable and not coherent with human learning processes. When a human scheduler is working, he normally builds a schedule step by step following a set of rules. After much practice, the scheduler gradually masters the knowledge of which solution parts go well with others. He can identify good parts and is aware of the solution quality even if the scheduling process is not completed yet, thus having the ability to finish a schedule by using flexible, rather than fixed, rules. In this research we intend to design more human-like scheduling algorithms, by using ideas derived from Bayesian Optimization Algorithms (BOA) and Learning Classifier Systems (LCS) to implement explicit learning from past solutions. BOA can be applied to learn to identify good partial solutions and to complete them by building a Bayesian network of the joint distribution of solutions [3]. A Bayesian network is a directed acyclic graph with each node corresponding to one variable, and each variable corresponding to individual rule by which a schedule will be constructed step by step. The conditional probabilities are computed according to an initial set of promising solutions. Subsequently, each new instance for each node is generated by using the corresponding conditional probabilities, until values for all nodes have been generated. Another set of rule strings will be generated in this way, some of which will replace previous strings based on fitness selection. If stopping conditions are not met, the Bayesian network is updated again using the current set of good rule strings. The algorithm thereby tries to explicitly identify and mix promising building blocks. It should be noted that for most scheduling problems the structure of the network model is known and all the variables are fully observed. In this case, the goal of learning is to find the rule values that maximize the likelihood of the training data. Thus learning can amount to 'counting' in the case of multinomial distributions. In the LCS approach, each rule has its strength showing its current usefulness in the system, and this strength is constantly assessed [4]. To implement sophisticated learning based on previous solutions, an improved LCS-based algorithm is designed, which consists of the following three steps. The initialization step is to assign each rule at each stage a constant initial strength. Then rules are selected by using the Roulette Wheel strategy. The next step is to reinforce the strengths of the rules used in the previous solution, keeping the strength of unused rules unchanged. The selection step is to select fitter rules for the next generation. It is envisaged that the LCS part of the algorithm will be used as a hill climber to the BOA algorithm. This is exciting and ambitious research, which might provide the stepping-stone for a new class of scheduling algorithms. Data sets from nurse scheduling and mall problems will be used as test-beds. It is envisaged that once the concept has been proven successful, it will be implemented into general scheduling algorithms. It is also hoped that this research will give some preliminary answers about how to include human-like learning into scheduling algorithms and may therefore be of interest to researchers and practitioners in areas of scheduling and evolutionary computation. References 1. Aickelin, U. and Dowsland, K. (2003) 'Indirect Genetic Algorithm for a Nurse Scheduling Problem', Computer & Operational Research (in print). 2. Li, J. and Kwan, R.S.K. (2003), 'Fuzzy Genetic Algorithm for Driver Scheduling', European Journal of Operational Research 147(2): 334-344. 3. Pelikan, M., Goldberg, D. and Cantu-Paz, E. (1999) 'BOA: The Bayesian Optimization Algorithm', IlliGAL Report No 99003, University of Illinois. 4. Wilson, S. (1994) 'ZCS: A Zeroth-level Classifier System', Evolutionary Computation 2(1), pp 1-18.

'Explicit Learning: an Effort towards Human Scheduling Algorithms'

Abstract Scheduling problems are generally NP-hard combinatorial problems, and a lot of research has been done to solve these problems heuristically. However, most of the previous approaches are problem-specific and research into the development of a general scheduling algorithm is still in its infancy. Mimicking the natural evolutionary process of the survival of the fittest, Genetic Algorithms (GAs) have attracted much attention in solving difficult scheduling problems in recent years. Some obstacles exist when using GAs: there is no canonical mechanism to deal with constraints, which are commonly met in most real-world scheduling problems, and small changes to a solution are difficult. To overcome both difficulties, indirect approaches have been presented (in [1] and [2]) for nurse scheduling and driver scheduling, where GAs are used by mapping the solution space, and separate decoding routines then build solutions to the original problem. In our previous indirect GAs, learning is implicit and is restricted to the efficient adjustment of weights for a set of rules that are used to construct schedules. The major limitation of those approaches is that they learn in a non-human way: like most existing construction algorithms, once the best weight combination is found, the rules used in the construction process are fixed at each iteration. However, normally a long sequence of moves is needed to construct a schedule and using fixed rules at each move is thus unreasonable and not coherent with human learning processes. When a human scheduler is working, he normally builds a schedule step by step following a set of rules. After much practice, the scheduler gradually masters the knowledge of which solution parts go well with others. He can identify good parts and is aware of the solution quality even if the scheduling process is not completed yet, thus having the ability to finish a schedule by using flexible, rather than fixed, rules. In this research we intend to design more human-like scheduling algorithms, by using ideas derived from Bayesian Optimization Algorithms (BOA) and Learning Classifier Systems (LCS) to implement explicit learning from past solutions. BOA can be applied to learn to identify good partial solutions and to complete them by building a Bayesian network of the joint distribution of solutions [3]. A Bayesian network is a directed acyclic graph with each node corresponding to one variable, and each variable corresponding to individual rule by which a schedule will be constructed step by step. The conditional probabilities are computed according to an initial set of promising solutions. Subsequently, each new instance for each node is generated by using the corresponding conditional probabilities, until values for all nodes have been generated. Another set of rule strings will be generated in this way, some of which will replace previous strings based on fitness selection. If stopping conditions are not met, the Bayesian network is updated again using the current set of good rule strings. The algorithm thereby tries to explicitly identify and mix promising building blocks. It should be noted that for most scheduling problems the structure of the network model is known and all the variables are fully observed. In this case, the goal of learning is to find the rule values that maximize the likelihood of the training data. Thus learning can amount to 'counting' in the case of multinomial distributions. In the LCS approach, each rule has its strength showing its current usefulness in the system, and this strength is constantly assessed [4]. To implement sophisticated learning based on previous solutions, an improved LCS-based algorithm is designed, which consists of the following three steps. The initialization step is to assign each rule at each stage a constant initial strength. Then rules are selected by using the Roulette Wheel strategy. The next step is to reinforce the strengths of the rules used in the previous solution, keeping the strength of unused rules unchanged. The selection step is to select fitter rules for the next generation. It is envisaged that the LCS part of the algorithm will be used as a hill climber to the BOA algorithm. This is exciting and ambitious research, which might provide the stepping-stone for a new class of scheduling algorithms. Data sets from nurse scheduling and mall problems will be used as test-beds. It is envisaged that once the concept has been proven successful, it will be implemented into general scheduling algorithms. It is also hoped that this research will give some preliminary answers about how to include human-like learning into scheduling algorithms and may therefore be of interest to researchers and practitioners in areas of scheduling and evolutionary computation. References 1. Aickelin, U. and Dowsland, K. (2003) 'Indirect Genetic Algorithm for a Nurse Scheduling Problem', Computer & Operational Research (in print). 2. Li, J. and Kwan, R.S.K. (2003), 'Fuzzy Genetic Algorithm for Driver Scheduling', European Journal of Operational Research 147(2): 334-344. 3. Pelikan, M., Goldberg, D. and Cantu-Paz, E. (1999) 'BOA: The Bayesian Optimization Algorithm', IlliGAL Report No 99003, University of Illinois. 4. Wilson, S. (1994) 'ZCS: A Zeroth-level Classifier System', Evolutionary Computation 2(1), pp 1-18.

Identifying the knowledge management structures in enterprise systems projects

Enterprise System (ES) implementation and management are knowledge intensive tasks that inevitably draw upon the experience of a wide range of people with diverse knowledge capabilities. Knowledge Management (KM) has been identified as a critical success factor in ES projects. Despite the recognized importance of managing knowledge for ES benefits realization, systematic attempts to conceptualize KM-structures have been few. Where the adequacy of KM-structures is assessed, the process and measures are typically idiosyncratic and lack credibility. Using the ‘KM-process’, itself based in sociology of knowledge, this paper conceptualizes four main constructs to measure the adequacy of KM-structures. The SEM model is tested using 310 responses gathered from 27 ES installations that had implemented SAP R/3. The findings reveal six constructs for KM-structure. Furthermore, the paper demonstrates the application of KM-structures in the context of ES using the Adaptive Structuration Theory. The results demonstrate that having adequate KM-structures in place, while necessary, is not sufficient. These rules and resources must be appropriated to have greater positive influence on the Enterprise System. Furthermore, the study provides empirical support for knowledge-based theory by illustrating the importance of knowledge use/re-use (vs. knowledge creation) as the most important driver in the process of KM.