The engineering of high-quality software requirements generally relies on properties and assumptions about the environment in which the software-to-be has to operate. Such properties and assumptions, referred to as environment conditions in this paper, are highly subject to change over time or from one software variant to another. As a consequence, the requirements engineered for a specific set of environment conditions may no longer be adequate, complete and consistent for another set.

Software systems are deployed in environments that keep changing over time. They should therefore adapt to changing conditions to meet their requirements. The satisfaction rate of these requirements depends on the rate at which adverse conditions prevent their satisfaction. Obstacle analysis is a goal-oriented form of risk analysis for requirements engineering (RE), whereby obstacles to system goals are identified, assessed, and resolved through countermeasures. The selection of effective countermeasures relies on environment assumptions and on the assessed likelihood and criticality of the corresponding obstacles. Those various factors estimated at RE time may, however, evolve at system runtime. To meet the system’s goals under changing conditions, this article proposes to defer obstacle resolution to system runtime. Techniques are presented for monitoring obstacle satisfaction rates; deciding when adaptation should be triggered; and adapting the system on-the-fly to countermeasures that are more effective. The approach relies on a model where goals and obstacles are refined and specified in a probabilistic linear temporal logic. The techniques allow for monitoring the satisfaction rate of probabilistic leaf obstacles; determining the severity of obstacle consequences on goal satisfaction rates computed from the monitored obstacle satisfaction rates; and shifting to countermeasures that better meet the required goal satisfaction rates. Our approach is evaluated on fragments of an ambulance dispatching system.

Requirements engineers are faced with multiple sources of uncertainty. In particular, the extent to which the identified software requirements and environment assumptions are adequate and sufficiently complete is uncertain; the extent to which they will be satisfied in the system-to-be is uncertain; and the extent to which obstacles to their satisfaction will occur is uncertain. The resolution of such domain-level uncertainty requires estimations of the likelihood that those different types of situations may or may not occur. However, the extent to which the resulting estimates are accurate is uncertain as well. This meta-level uncertainty limits current risk-based methods for requirements engineering. The paper introduces a quantitative approach for managing it. An earlier formal framework for probabilistic goals and obstacles is extended to explicitly cope with uncertainties about estimates of likelihoods of fine-grained obstacles to goal satisfaction. Such estimates are elicited from multiple sources and combined in order to reduce their uncertainty margins. The combined estimates and their uncertainties are up-propagated through obstacle refinement trees and then through the system's goal model. Two metrics are introduced for measuring problematic uncertainties. When applied to the probability distributions obtained by up-propagation to the top-level goals, the metrics allow critical leaf obstacles with most problematic uncertainty margins to be highlighted. The proposed approach is evaluated on excerpts from a real ambulance dispatching system.

Missing requirements are known to be among the major sources of software failure. Incompleteness often results from poor anticipation of what could go wrong with an over-ideal system. Obstacle analysis is a model-based, goal-anchored form of risk analysis aimed at identifying, assessing and resolving exceptional conditions that may obstruct the behavioral goals of the target system. The obstacle resolution step is obviously crucial as it should result in more adequate and more complete requirements. In contrast with obstacle identification and assessment, however, this step has little support beyond a palette of resolution operators encoding tactics for producing isolated countermeasures to single risks. In particular, there is no single clue to date as to where and how such countermeasures should be integrated within a more robust goal model. To address this problem, the paper describes a systematic technique for integrating obstacle resolutions as countermeasure goals into goal models. The technique is shown to guarantee progress towards a complete goal model; it preserves the correctness of refinements in the overall model; and keeps the original, ideal model visible to avoid cluttering the latter with a combinatorial blow-up of exceptional cases. To allow for this, the goal specification language is slightly extended in order to capture exceptions to goals seperately and distinguish normal situations from exceptional ones. The proposed technique is evaluated on a non-trivial ambulance dispatching system.

Requirements completeness is among the most critical and difficult software engineering challenges. Missing requirements often result from poor risk analysis at requirements engineering time. Obstacle analysis is a goal-oriented form of risk analysis aimed at anticipating exceptional conditions in which the software should behave adequately. In the identify-assess-control cycles of such analysis, the assessment step is not well supported by current techniques. This step is concerned with evaluating how likely the obstacles to goals are and how likely and severe their consequences are. Those key factors drive the selection of most appropriate countermeasures to be integrated in the system goal model for increased completeness. Moreover, obstacles to probabilistic goals are currently not supported; such goals prescribe that some corresponding target property should be satisfied in at least X% of the cases. The paper presents a probabilistic framework for goal specification and obstacle assessment. The specification language for goals and obstacles is extended with a probabilistic layer where probabilities have a precise semantics grounded on system-specific phenomena. The probability of a root obstacle to a goal is thereby computed by up-propagation of probabilities of finer-grained obstacles through the obstacle refinement tree. The probability and severity of obstacle consequences is in turn computed by up-propagation from the obstructed leaf goals through the goal refinement graph. The paper shows how the computed information can be used to prioritize obstacles for countermeasure selection towards a more complete and robust goal model. The framework is evaluated on a non-trivial carpooling support system.