Many practical and important controller design problems,
such as multi-objective controller design,
structured controller design,
and lower order controller design,
are bilinear matrix inequality (BMI) problems
and cannot be reduced to LMI problems.
Accordingly, methods to solve BMI problems are still important,
even though the BMI problems are known to be NP-hard.
The overbounding approximation method is a one of
the numerical methods to obtain approximate solutions to BMI problems.
This method successively approximates
BMI constraints to sufficient LMI constraints
and solve the approximated LMI problem iteratively.
This approach has following features.
- This approach only requires a general SDP solver.
- This approach can be applied to any types of BMI problems.
- The performance index to be optimized
is monotonically decreasing during the successive design procedure.