This research project was part of my studies as M.Sc. in Electrical an Information Engineering at the University of Stuttgart.
Abstract of the reasearch project:
Choosing the activity variants of a project such that the total project duration and cost is minimized is a problem well known in literature as the discrete time-cost trade-off problem (DTCTP). This problem is extended and generalized by adding stochastic events and activity durations and allowing a continuous-valued schedule compression of activities, making it a mixed (categorical and continuous), probabilistic and nonlinear optimization problem. This thesis has two contributions: First it shows that algorithms for solving this generalized time-cost trade-off problem perform much better than humans and heuristics do and compares the performance of Bayesian Optimization, Genetic and Actor-Critic algorithms on these kinds of problems. Second it provides a framework for modelling complex project management problems including stochastic activity durations, stochastic events and continuous-valued schedule compression which is used for providing a simulation model of three example projects based on project management business games.