Let $Y$ be the quotient of $X$ by the action of $G$, defined in the following way.
Define an equivalence $\sim_G$ relation on $X$ by
$$x_1\sim_G x_2 \iff \exists g\in G : x_2=g.x_1$$
and set $Y:=X/\sim_G$. (When the action is clear from context, the notation $X/G$ is also used).
For each $y\in Y$, set $O_y:= \mathrm{orb}_G(x)$ for any $x$ with $[x]=y$, where $[x]$ denotes the quotient projection (the class of $x$ in $Y$). This definition is well-posed by definition of $\sim_G$, i.e. if $x'$ is such that $[x']=y$, then $\mathrm{orb}_G(x')=\mathrm{orb}_G(x)$.
Now we have
$$X=\bigcup_Y O_y\quad.$$
In words, $Y$ replaces $R$ in that we do not need to choose one representative $x$ for each class $y=[x]$, and can rather directly choose as indexing set the set of classes. The awkwardness in using $R$ comes precisely from the arbitrariness of the choice of a representative for each class, but we in fact we do not need to make such a choice in the first place.