The definition of the coefficients $a_{r,k}$ is given by theorem III.17 on page 53 of "Elliptic curves in cryptography" by I.F. Blake, G. Seroussi, N.P. Smart (Cambridge University Press, 1999).

With $v,s,l$ as in the question, define $f(\tau) = \left( \dfrac {\eta(\tau)} {\eta(l\tau)} \right)^{2s}$, where $\eta$ is Dedekind's $\eta$ function. If $j$ is the invariant function (see page 47 of the quoted book), then there exist numbers $a_{r,k} \in \mathbb Z$ such that

$$\sum _{r=0} ^{l+1} \sum _{k=0} ^v a_{r,k} f(\tau)^r j(lt)^k =0 \ .$$

Now just formally replace $f(\tau)$ with $x$ and $j(l\tau)$ with $y$ to get $G_l(x,y)$.

The definition of the coefficients $a_{r,k}$ is given by theorem III.17 on page 53 of "Elliptic curves in cryptography" by I.F. Blake, G. Seroussi, N.P. Smart (Cambridge University Press, 1999) (also on Google Books).

With $v,s,l$ as in the question, define $f(\tau) = \left( \dfrac {\eta(\tau)} {\eta(l\tau)} \right)^{2s}$, where $\eta$ is Dedekind's $\eta$ function. If $j$ is the invariant function (see page 47 of the quoted book), then there exist numbers $a_{r,k} \in \mathbb Z$ such that

$$\sum _{r=0} ^{l+1} \sum _{k=0} ^v a_{r,k} f(\tau)^r j(lt)^k =0 \ .$$

Now just formally replace $f(\tau)$ with $x$ and $j(l\tau)$ with $y$ to get $G_l(x,y)$.

Brief descriptions of the actual computation of these coefficients are given on page 54, and in "Counting the number of points on elliptic curves over finite fields of characteristic greater than three" by F. Lehmann, M. Maurer, V. Müller, V. Shoup.