Search-Based Synthesis of Probabilistic Models for Quality-of-Service Software Engineering (T)

Gerasimou, Simos; Tamburrelli, Giordano; Călinescu, Radu

doi:10.1109/ase.2015.22

Cited by 42 publications

(62 citation statements)

References 47 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…We realised the functionality of CANDIDATEDESIGNS using the jMetal [29] Java framework for multi-objective optimisation with metaheuristics. Our Robust DEsign Synthesis (RODES) tool operates with pCTMCs expressed in the high-level modelling language of PRISM [9] extended with the following constructs (adopted from [13]) for specifying the parameters k ∈ K and d ∈ D from Definition 1: …”

Section: Methodsmentioning

confidence: 99%

“…RELATED WORK In previous work [13], we proposed a purely search-based engineering approach that uses evolutionary algorithms to synthesise probabilistic models that satisfy multi-objective specifications. However, the designs generated by this approach are non-parameteric probabilistic models, and thus cannot support sensitivity analysis like our new method.…”

Section: Case Studiesmentioning

confidence: 99%

“…Precise parameter synthesis [12] identifies transition rates that enable continuous Markov models to satisfy a quality requirement or to optimise a quality attribute of the modelled system. Finally, probabilistic model synthesis [13] starts from a design template that captures alternative system designs, and uses multiobjective optimisation to generate the Pareto-optimal set of Markov models associated with the quality requirements of the system. However, these promising approaches unrealistically assume that the parameters of the real system (e.g.…”

Section: Introductionmentioning

confidence: 99%

“…To the best of our knowledge, our work is the first to integrate design synthesis and sensitivity analysis into a single endto-end method -existing research efforts have tackled the challenges associated with design synthesis (e.g. [13], [15]) and sensitivity analysis (e.g. [16]- [20]) in isolation.…”

Section: Introductionmentioning

confidence: 99%

See 3 more Smart Citations

Designing Robust Software Systems through Parametric Markov Chain Synthesis

Călinescu

Češka

Gerasimou

et al. 2017

2017 IEEE International Conference on Software Architecture (ICSA)

Self Cite

View full text Add to dashboard Cite

Abstract-We present a method for the synthesis of software system designs that satisfy strict quality requirements, are Paretooptimal with respect to a set of quality optimisation criteria, and are robust to variations in the system parameters. To this end, we model the design space of the system under development as a parametric continuous-time Markov chain (pCTMC) with discrete and continuous parameters that correspond to alternative system architectures and to the ranges of possible values for configuration parameters, respectively. Given this pCTMC and required tolerance levels for the configuration parameters, our method produces a sensitivity-aware Pareto-optimal set of designs, which allows the modeller to inspect the ranges of quality attributes induced by these tolerances, thus enabling the effective selection of robust designs. Through application to two systems from different domains, we demonstrate the ability of our method to synthesise robust designs with a wide spectrum of useful tradeoffs between quality attributes and sensitivity.Keywords-software performance and reliability engineering; probabilistic model synthesis; multi-objective optimisation I. INTRODUCTION Evaluating the performance, reliability and other quality attributes of alternative designs is essential for the cost-effective engineering of software [1], [2]. Delaying this evaluation until integration or system testing can greatly increase engineering costs, as defects identified late in the development lifecycle require much more effort to fix [3]. A common method to avoid this delay uses model-based simulation [4] or formal verification [5] to predict the quality attributes of alternative designs. Models that meet the quality requirements of the system under development are then used as a basis for its implementation. Models based on queueing networks [6], probabilistic models [2], [5] and timed automata [7] have been used for this purpose, together with tools for their simulation (e.g. Palladio [8]) and verification (e.g. PRISM [9]). Furthermore, recently proposed approaches automate the search for suitable designs. Probabilistic model repair [10], [11] automatically modifies the transition probabilities of Markov models that violate a quality requirement, generating new models that meet the requirement. Precise parameter synthesis [12] identifies transition rates that enable continuous Markov models to satisfy a quality requirement or to optimise a quality attribute of the modelled system. Finally, probabilistic model synthesis [13] starts from a design template that captures alternative system designs, and uses multiobjective optimisation to generate the Pareto-optimal set of Markov models associated with the quality requirements of the system.

show abstract

Section: Methodsmentioning

confidence: 99%

Section: Case Studiesmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

See 2 more Smart Citations

Designing Robust Software Systems through Parametric Markov Chain Synthesis

Călinescu

Češka

Gerasimou

et al. 2017

2017 IEEE International Conference on Software Architecture (ICSA)

Self Cite

View full text Add to dashboard Cite

show abstract

“…Our algorithm is not prescriptive about the method used to get new candidate policies. As such, the function GETCAN-DIDATEPOLICIES from line 4 can be implemented using a metaheuristic such as the genetic algorithm used to synthesise Markovain models in (Gerasimou et al, 2015), a simple heuristic like hill climbing, or just random search.…”

Section: The Arl Approachmentioning

confidence: 99%

Assured Reinforcement Learning with Formally Verified Abstract Policies

Călinescu¹,

Kudenko²,

Banks³

2017

Proceedings of the 9th International Conference on Agents and Artificial Intelligence

Self Cite

View full text Add to dashboard Cite

Markov Decision Processes. Abstract:We present a new reinforcement learning (RL) approach that enables an autonomous agent to solve decision making problems under constraints. Our assured reinforcement learning approach models the uncertain environment as a high-level, abstract Markov decision process (AMDP), and uses probabilistic model checking to establish AMDP policies that satisfy a set of constraints defined in probabilistic temporal logic. These formally verified abstract policies are then used to restrict the RL agent's exploration of the solution space so as to avoid constraint violations. We validate our RL approach by using it to develop autonomous agents for a flag-collection navigation task and an assisted-living planning problem.

show abstract