Chance-constrained programming in activity networks: A critical evaluation

We review the chance-constrained programming (CCP) model for event realization in activity networks (ANs), and the more recent contribution to it by Kress, from the points of view of validity and accuracy. We present a classification scheme of stochastic programming and confirm that the CCP does not solve the problem of finding a point estimate that satisfies a required confidence level, and is of little help in determining the cumulative distribution function (c.d.f.) of the project completion time. We show that the CCP leads to an extremely weak lower bound (l.b.) on the exact c.d.f. A backtracking approach is proposed that improves this l.b. via the CCP, which is still too weak to be of any practical value. Finally, we utilize the CCP approach to derive a result due to Kress in a simpler and more direct way. We also demonstrate that the result itself, unfortunately, is of questionable validity.