This problem is a typical DP. we denote dp[n][i][j] as possible ways to be in node (i, j) as timestep n.
As we can easily discover that:
dp[n+1][i][j] = dp[n][i+1][j]+dp[n][i-1][j]+dp[n][i][j+1]+dp[n][i][j-1]And for every timestep, we need to check whether dp[n][i][j] can jump out of the boundary.
Note: carefully check the boundaries.