haosulab
ValueError: Could not load resource ……/link0.obj.convex.stl
I tried to make a demo using my own codes，following the documetiation：
from mplib import Planner …… planner = Planner( urdf="/home/potato/workplace/IsaacGym_Preview_4_Package/isaacgym/assets/urdf/franka_description/robots/franka_panda.urdf", move_group="panda_hand", # srdf=None, # new_package_keyword="", use_convex=False, link_names=["panda_link0", "panda_link1", "panda_link2", "panda_link3", "panda_link4", "panda_link5", "panda_link6", "panda_link7", "panda_link8","panda_hand"], joint_names=["panda_joint1", "panda_joint2", "panda_joint3", "panda_joint4", "panda_joint5", "panda_joint6", "panda_joint7"], joint_vel_limits=None, joint_acc_limits=None, objects=[], verbose=False, ) …… status, q_goals = planner.IK( …… ） ‘ it printed errors：
Traceback (most recent call last):
File "/home/potato/workplace/test_sapien/get start/xbox_demo.py", line 171, in
@LittlePotatoChip Same problem here. Have you solved it?
no, I have switched to other tools,I find it's not necessary for my work
What version are you guys using? There might have been a convex bug before iirc. Can you please try the nightly wheels here?
@Lexseal Thanks for your prompt response. I'm using the 0.1.1 version since Maniskill only supports this version at the moment. I managed to use the following code to generate a convex.stl file from a stl file. Is it correct? Also, since I did not find the document for version 0.1.1, why must a convex hull be used? Could you please explain or refer me to some explanation links?
import numpy as np
import scipy.spatial
from stl import mesh
def generate_convex_stl(input_stl_path, output_stl_path):
original_mesh = mesh.Mesh.from_file(input_stl_path)
vertices = np.unique(original_mesh.vectors.reshape(-1, 3), axis=0)
hull = scipy.spatial.ConvexHull(vertices)
convex_vertices = vertices[hull.vertices]
convex_triangles = hull.simplices
convex_mesh = mesh.Mesh(np.zeros(len(convex_triangles), dtype=mesh.Mesh.dtype))
for i, simplex in enumerate(convex_triangles):
convex_mesh.vectors[i] = convex_vertices[simplex]
convex_mesh.save(output_stl_path)
print(f"Convex hull saved to {output_stl_path}")
def batch_convex_hull_generation(input_directory, output_directory):
import os
os.makedirs(output_directory, exist_ok=True)
for filename in os.listdir(input_directory):
if filename.lower().endswith('.stl'):
input_path = os.path.join(input_directory, filename)
output_path = os.path.join(output_directory, f'{filename}.convex.stl')
generate_convex_stl(input_path, output_path)
convex hull should work if your geometry is more or less convex to begin with. at worst your collision detection will be a little preemptive