Online testing with reinforcement learning

  • Authors:
  • Margus Veanes;Pritam Roy;Colin Campbell

  • Affiliations:
  • Microsoft Research, Redmond, WA;University of California, Santa Cruz;Microsoft Research, Redmond, WA

  • Venue:
  • FATES'06/RV'06 Proceedings of the First combined international conference on Formal Approaches to Software Testing and Runtime Verification
  • Year:
  • 2006

Quantified Score

Hi-index 0.00

Visualization

Abstract

Online testing is a practical technique where test derivation and test execution are combined into a single algorithm. In this paper we describe a new online testing algorithm that optimizes the choice of test actions using Reinforcement Learning (RL) techniques. This provides an advantage in covering system behaviors in less time than with a purely random choice of test actions. Online testing with conformance checking is modeled as a 1$\frac{1}{2}$-player game, or Markov Decision Process (MDP), between the tester as one player and the implementation under test (IUT) as the opponent. Our approach has been implemented in C#, and benchmark results are presented in the paper. The specifications that generate the tests are written as model programs in any .NET language such as C# or VB.