1

我正在使用 mesa（基于代理的建模）来实现景观模型。为了生成景观，我使用了几个 numpy 数组 (800 x 800)，每个数组都包含特定信息（例如海拔、土壤湿度等）。为了初始化模型，Mesa 遍历用户指定的网格大小（在本例中为 800 x 800）并将“代理”分配给网格上的每个“补丁”：

    def coord_iter(self):
    """ An iterator that returns coordinates as well as cell contents. """
    for row in range(self.width):
        for col in range(self.height):
            yield self.grid[row][col], row, col  # agent, x, y

然后，使用此迭代器，我将 numpy 数组中的值分配给网格上的每个 x,y：

for agent, x, y in self.grid.coord_iter(self.landscape.elevation):
        # check that patch is land
        if self.landscape.elevation[x, y] != -9999:
            self.agents_created += 1
            self.create_landscape_patch(x, y)

            # set landscape_classes (i.e. 0 = bracken) if 5 then this is a deposition area
            if self.landscape.vegetation[x, y] != 5:
                self.create_vegetation_patch(x, y)

            # populate landscape patch with values
            self.add_landscape_patch(x, y)

        # if deposition site
        if self.landscape.vegetation[x, y] == 5:
            self.create_deposition_patch(x,y)

在 numpy 数组中，有一些我不需要分配给网格的 x、y 值，所以我目前使用一个简单的“if”语句来过滤掉这些值 ( if self.landscape.elevation[x, y] != -9999:)。显然，通过使用 coord_iter 函数，我不可避免地会执行 800 x 800 次循环，而当 numpy 数组中的值 == -9999 时，我可以避免迭代。下面给出了如何创建每个“代理”然后将其分配给台面网格的示例（景观补丁）。

def create_landscape_patch(self, x, y):

    self.elevation_xy = int(self.landscape.elevation[x, y])

    # calculate burn probabilities based on soil and temp
    self.burn_s_m_p = self.landscape.burn_s_m_p_array[x, y]
    self.burn_s_t_p = self.landscape.burn_s_t_p_array[x, y]

    # calculate succession probabilities based on soil and temp
    self.succ_s_m_p = self.landscape.succ_s_m_p_array[x, y]
    self.succ_s_t_p = self.landscape.succ_s_t_p_array[x, y]

    self.time_colonised_xy = self.landscape.time_colonised[x, y]
    self.is_patch_colonised_xy = self.landscape.colonised[x, y]

def add_landscape_patch(self, x, y):
    patch = Landscape(self.unique_id, (x, y), self, self.elevation_xy, self.burn_s_m_p, self.burn_s_t_p, self.vegetation_typ_xy,
                      False, self.time_colonised_xy, self.is_patch_colonised_xy, self.succ_s_m_p, self.succ_s_t_p, self.veg_this_patch)
    self.grid.place_agent(patch, (x, y))
    self.schedule.add(patch)
    self.unique_id += 1

目前我的代码有效，但我知道必须有更有效的方法来实现这个结果。“构建”景观所需的时间是几分钟，我希望能构建数千个！任何有关如何提高效率的帮助将不胜感激。谢谢

4

0 回答 0