zipping-suger / parameter_synthesis_pMC

parameter synthesis problem

README

Prequisite

python 3.8 (tested)

Gurobi 9.5 with python package gurobipy installed

PRISIM 4.6 (optional)

Description

Knuth_Yao_pMC.py

The main running file Knuth_Yao_pMC.py offers four different methods to solve parameter synthesis for the Knuth -Yao algorithm pMC model. (see reprot.pdf)

• Function original_qcqp solves the original nonconvex QCQP problem using a nonlinear Gurobi optimizer.
• Function penalty_ccp uses the convex-concave procedure to iteratively solve the convexified QCQP problem.
• Function penalty_scp uses sequential convex programming to iteratively solve an approximated LP problem.
• Function iteration_LP substitutes parts of the probability variables with results obtained by model checking, and iteratively solves the approximated LP problem.

GRB_optimize.py

The file GRB_optimize.py contains the formulation of the subproblem in each iteration, which are solved by the Gurobi optimizer (e.g. LP, QCP, nonlinear optimizer).

model_check.py

The file model_check.py offers two methods for probabilistic model checking.

The function model_check_PRISM uses the probabilistic model checker PRISM for model checking. If PRISM 4.6 is not installed, this function is not recommended.

And the default function model_check_LE check the probability by solving a linear equation system (see report.pdf).