The system state (the length of the queue)
\(X_t\in \mathbb {R}^+\). Consequently, the policies of the players are defined on
\(\mathbb {R}^+\). The customers arrive at the queue according to a fluid process with rate
\({\uplambda }\). As each of them uses some policy
\(\pi _k\), the real incoming rate at time
t is
\(\overline{{\varvec{{\pi }}}}(X_t){\uplambda }\) where
\(\overline{{\varvec{{\pi }}}}(X_t)\) is the average strategy of the arriving users. Each of them stays in the queue an exponentially distributed time with parameter
\(\mu \), and so the outflow is according to a fluid process with rate
\(\mu X_t\). This can be described as the following ODE:
$$\begin{aligned} {\left\{ \begin{array}{ll} \overset{.}{X}_t({\varvec{{\pi }}})=\overline{{\varvec{{\pi }}}}(X_t){\uplambda }-\mu X_t({\varvec{{\pi }}}),\quad \forall t\ge 0\\ X_0=x_0 \end{array}\right. } \end{aligned}$$
(3)
Since there are infinitely many players in the game now, we encounter problems with defining the multi-policies. For that reason, we assume that in multi-policy
\({\varvec{{\pi }}}\) all the players use the same policy
\(\pi \). If we want to write that only one player, say player
k changes his policy to some
\(\pi '_k\), we write that players apply policy
\([{\varvec{{\pi }}}^{-k},\pi '_k]\), meaning that each player uses policy
\(\pi \) except player
k. Also note that the game is symmetric since each player has the same payoff function and strategy space, and thus, it is very difficult to implement an asymmetric NE—we elucidate the inherent complications considering only two players: If in an NE
\(\pi _{1}^{*}\ne \pi _2^{*}\), then, by the symmetric nature of the game,
\((\pi _2^{*},\pi _1^{*})\) is also an NE. If player 2 knows that player 1 selects
\(\pi ^{*}\) (
\(\pi _2^{*}\), respectively), then the optimal response for player 2 is to select
\(\pi _2^{*}\) (
\(\pi _1^{*}\), respectively), but player 2 cannot know the selection of player 1 due to the non-cooperation between them. Under symmetric NE, all players select the same strategy and thus the above complication is somewhat alleviated. Moreover,
\(\overline{{\varvec{{\pi }}}}\) is not always well defined, but in such a case
\(\overline{{\varvec{{\pi }}}}(X_t)\equiv \pi (X_t)\). Also with these assumptions, both the cost and the equilibrium can be defined as in the discrete model.