Huisouan
Huisouan
During the reinforcement learning training process using SKRL, my action variable often becomes all NaN at a fixed iteration (around 67). What could be causing this issue? 
I got this error when trying to render the  way to reprocude: 
在genesis生成复杂地形的时候,convert_heightfield_to_watertight_trimesh函数里面使用了open3d的simplify_quadric_decimation对mesh进行简化。然而这个函数在计算上可能有一定的问题,请看如下链接:[simplify_quadric_decimation works horrible #2581 ](https://github.com/isl-org/Open3D/issues/2581) 是否可以先暂时不要使用这个简化,直接使用sdf_mesh。 ` def convert_heightfield_to_watertight_trimesh(height_field_raw, horizontal_scale, vertical_scale, slope_threshold=None): """ Adapted from Issac Gym's `convert_heightfield_to_trimesh` function. Convert a heightfield array to a triangle mesh represented by vertices...