kputtann / gms

A Generalized H-infinity Mixed Sensitivity Convex Approach to Multivariable Control Design Subject to Simultaneous Output and Input Loop-Breaking Specifications

gms

A Generalized H-infinity Mixed Sensitivity Convex Approach to Multivariable Control Design Subject to Simultaneous Output and Input Loop-Breaking Specifications

Computes a H-infinity based Feedback Controller based on multiobjective constrained convex optimization.

Outline of steps for GMS problem setup:

• Form the design plant:

  • Define the original plant
  • Integrator augmentation if needed
  • Bilinear transformation values if needed

• Select weighting functions:

  • Tradeoff param rho
  • W for obj
  • W for constraint

• Select optimization params:

  • LB and UB
  • Init point
  • Maximum number of iterations

• Select Youla/Zames parametrization:

  • Select Youla or Zames
  • Initial controller

• Finite Dimensionality

  • Basis params

• Objective function: - sum/max/stacking

• Find initial controller (Ko, F, L)

• Youla parameterization

• Find Initial Q parameter using initial controller (Ko, F, L)

• Extract required data from problem setup

• Vectorize the optimization problem

• Optimization process

  • define how subgradient is picked based on sum/max/stacking

• form Q using the optimized variables and bases

• form Controller K using the obtained Q

• Inverse bilinear transformation if needed

• Inverse of integrator augmentation if needed

• Compute OL and CL maps

gms's Issues

Computation time for individual optimization iterations