A function field tower is a sequence of function fields $$\mathcal{F}_0 \subset \mathcal{F}_1 \subset \mathcal{F}_2 \dots \subset \mathcal{F}_{n} \subset \mathcal{F}_{n+1} \subset \dots $$ over a base field (in our case the finite field $F_{q}$ say) such that each extension is finite and separable, and the genus tends to infinity with $n$. (Such towers are constructed and studied with the aim of ensuring that the number of rational points also grows with the genus (or better still, are optimal), and such towers have applications in codes and cryptography.) Simple constructions of towers are "recursive" in the sense that we use one equation of the form $f(Y) = h(X)$ as the defining equation of the tower, and then $\mathcal{F}_{n+1}$ would be obtained from $\mathcal{F}_{n}$ as $f(X_{n+1}) = h(X_{n})$, with $\mathcal{F}_0 = F_{q}(X_0)$. (We would need to choose the defining equation appropriately so that this defines a valid tower with good properties).
My question is whether there exist any known tower constructions with a larger "depth" or "memory", i.e. where $X_{n+1}$ depends not just on $X_n$ but also $X_{n-1}$ or more. This would involve a defining equation which is not just bivariate.
Thaks.