Case Study: feedback control

[bchasnov]

To be expanded

Basically take a system with state evolving linearly as s(t+1) = a*s(t) + b1*u1(t) + b2*u2(t) and parameterize the actions ui(t) = ki*x(t) and perform gradient descent on the infinite time horizon cost on parameters (k1, k2) and see what the equilibria are.

The allowed parameters (k1,k2) such that |a+b1*k1+b2*k2|<1.