Construction planners face the decisions of selecting appropriate resources, including crew sizes, equipment, methods, and technologies, to perform the tasks of a construction project. In general there is a tradeoff between time and cost the less expensive the resources, the longer it takes. Using CPM techniques, the overall project cost can be reduced by using less expensive resources for noncritical activities without impacting the duration. This project concerns a lower-bounding technique to generate a high-quality estimate of the overall time-cost trade-off curve of a project.

Structural engineers at Caterpillar use finite-element analysis and fatigue simulation packages on structures to determine loads, deflections, stresses, and fatigue damage. There are many individual packages that handle each of the specific design tasks. Each design package presents the data as a separate output that may be useful in the decision-making process. However, due to the analysis process used at Caterpillar, all of the relevant data may not be available when necessary. Our proposal is to create ``tools'' for decision making that would increase the information density presented to the design engineers, and thus improve the design process. This will be accomplished by transforming the data from the various software packages to an integrated data format.

In one recent case study, messy genetic algorithms (mGAs) were theoretically and empirically shown to find solutions to boundedly deceptive problems in times that grew as a subquadratic number of function evaluations. This study generalizes this finding to problems that may be difficult by having isolated optima, misleading optima, noisy functions, covariant building blocks (crosstalk), or many false peaks. In particular, niching and speciation techniques are investigated to permit the simultaneous solution of multiple optima and to make mGA phasing less time critical. The mGA is then extended to classification and direct machine learning codings.

This project funds travel and ancillary expenses to permit the principal investigator to visit the Indian Institute of Technology, Madras, and collaborate with C. S. Krishnamoorthy of the Department of Civil Engineering. The goal of the project is combine Illinois foundational and theoretical knowledge of genetic algorithms with IIT structural and GA applications experience to create a structural optimization system capable of solving difficult, large-scale structural problems.

This project builds on the fast, accurate, reliable results obtained using fast messy genetic algorithms and investigates four means of obtaining even faster results: (1) spatial efficiencies, (2) temporal efficiencies, (3) sampling efficiencies and evaluation relaxations, and (4) systematic convergence criterion relaxations. Primary among these are spatial and temporal efficiencies, and the efficient use of space and time are critical to obtaining fast results on both serial and parallel machines. Bounding analytical studies show how to do the appropriate decomposition and empirical studies to verify the speedup. Although the work is intended for the speedup of messy GAs in particular, the study generalizes to almost all evolutionary computations with little or no modification.

Traditionally, performance analysis and routing algorithms for connection-oriented telecommunication networks were studied for only one type of connection, i.e., voice calls. However, there are two main features that distinguish emerging networks from voice networks: general network topology and multiple traffic classes having widely different arrival and departure characteristics. The main thrust of this research is to address the myriad challenges that arise due to these characteristics of integrated services networks. Specifically, we focus on developing optimal admission control and routing schemes and fast, accurate analytic and simulation techniques to evaluate the performance of these schemes.