Finding robust solutions in requirements models

Abstract

Abstract. Solutions to non-linear requirements engineering problems may be "brittle"; i.e. small changes may dramatically alter solution effectiveness. Hence, it is not enough to just generate solutions to requirements problems-we must also assess solution robustness. The KEYS2 algorithm can generate decision ordering diagrams. Once generated, these diagrams can assess solution robustness in linear time. In experiments with real-world requirements engineering models, we show that KEYS2 can generate decision ordering diagrams in O(N 2 ). When assessed in terms of terms of (a) reducing inference times, (b) increasing solution quality, and (c) decreasing the variance of the generated solution, KEYS2 out-performs other search algorithms (simulated annealing, ASTAR, MaxWalkSat).

Not going to take this anymore: Multi-objective overtime planning for Software Engineering projects

“…Complex multi-objective decision problems with competing and conflicting constraints such as this are well suited to Search Based Software Engineering (SBSE) [11], which has proved able to provide decision support for other early-stage development activities, notably requirements engineering [12], [13], [14]. We believe that this is the first time that an approach has been introduced to provide decision support for software engineers attempting to reconcile these complex and difficult problems.…”

Computational alignment of goals and scenarios for complex systems

“…From artificial intelligence, multiobjective evolutionary optimization algorithms have been applied to requirements selection in release planning [6], but these search-based techniques are limited to validating small numbers of predefined quality goals such as cost and time-todeliver. Likewise, the KEYS2 Bayesian-based method has been combined with decision ordering diagrams to explore requirements-led changes [7] on single qualities such as robustness rather than a broader set of system and organizational goals. What we need are new applications of artificial intelligence to analyse more complete sets of system behaviours and qualities, as well as enable human understanding of those behaviours and qualities.…”

Automatically finding the control variables for complex system behavior

“…Furthermore, the algorithms that we use are unconstrained (constrained algorithms work towards a pre-determined number of possible solutions while unconstrained methods are allowed to adjust to the goal space). Simulated Annealing has been used to optimize similar design models in our own previous work [19].…”

“…Gay and Menzies recently conducted a similar treatment learning exercise on NASA Jet Propulsion Lab projects [19]. These projects were encoded in the Defect Detection & Prevention format [13,17], which is a compiled model representing the requirements of a module, the risks that could compromise those requirements, and mitigations that can allay these risks.…”