Representation for an ARMA(1, 1) state process

[ashaffer]

I'm trying to figure out how to represent an ARMA(1,1) state process, and I think I have a way of doing it that should work in principle, but it's erroring out with this error:

https://github.com/atsa-es/MARSS/blob/master/R/utility_functions.R#L243

The method I was trying to use to represent the model is:

iter = 1
control = list(maxit=iter, minit=iter, trace = 1, safe = 1)
model <- list(
    B = matrix(c(0.4, 0, 0.2, 0), nrow=2),
    G = matrix(c(1, 1, 0, 0), nrow=2),
    Q = matrix(c('q1', 0, 0, 0), nrow=2),
    A = matrix(c(0), 1),
    Z = matrix(c(1, 0), 1, 2),
    R = matrix(c(1)),
    V0 = matrix(c(1, 0, 0, 0), 2),
    L = matrix(c(1, 1, 0, 0), 2),
    U = matrix(c(0, 0), 2, 1),
    x0 = matrix(c(0, 0), nrow=2),
    tinitx = 0
)

fit.0 <- MARSS(y, model = model, control=control, fun.kf="MARSSkfss")

The idea being to use the G matrix to copy the error term into the second state component, and then use it in the next cycle. Is there another way to represent this concept, or is it just not representable?

EDIT: Also, my understanding from the derivation paper is that I should be able to make Q 1x1 in this circumstance, and allow G to expand it to 2x2, but the library doesn't seem to like that. Is that intentional?