1: \begin{definition}[{\cite[Theorem 5.3]{brownianwebandnet}}] \label{def:martingaleproblem} A stochastic process $\vec{X}(t)=(X_1(t),...,X_n(t))$ started from $\vec X(0)$ will be called $n$-point sticky Brownian motions if it solves the following martingale problem called the Howitt-Warren martingale problem with drift $\beta$ and characteristic measure $\nu$. 

2: \begin{itemize}

3: \item{(i)} $\vec{X}$ is a continuous, square integrable martingale.

4: \item{(ii)} The processes $X_i$ and $X_j$ have covariation process

5: $$\langle X_i, X_j \rangle(t)=\int_0^t \mathds{1}_{X_i(s)=X_j(s)} ds, \qquad \text{for $t \geq 0$, $i,j=1,...,n$}.$$

6: \item{(iii)} Consider any $\Delta \subset \{1,...,n\}$. For $\vec{x} \in \mathbb{R}^n$, let

7: $$f_{\Delta}(\vec{x}):=\max_{i \in \Delta} \{ x_i\} \qquad \text{and} \qquad g_{\Delta}(\vec{x}):=|\{i \in \Delta: x_i=f_{\Delta}(\vec{x})\}|,$$

8: where $|S|$ denotes the number of elements in a set $S$. Then

9: $$f_{\Delta}(\vec{X}(t))-\int_0^t \beta_+(g_{\Delta}(\vec{X}(t)) ds$$

10: is a martingale with respect to the filtration generated by $\vec{X}$, where

11: $$\beta_+(1):=\beta \qquad and \qquad \beta_+(m):= \beta+2 \int \nu(dy) \sum_{k=0}^{m-2} (1-y)^k = \beta + 2\sum_{k=1}^{m-1} \theta(1,k).$$

12: \end{itemize}

13: \end{definition}