Evaluating and reproducing real-world robot manipulation policies (e.g., RT-1, RT-1-X, Octo) in simulation under common setups (e.g., Google Robot, WidowX+Bridge)
All codes are set up with the introduction in README. vulkaninfo and python -m sapien.example.offscreen works fine, seems that there are no rendering issues.
Segmentation fault is located at:
ManiSkill2_real2sim/mani_skill2_real2sim/agents/controllers/pd_ee_pose.py line 61:
result, success, error = self.pmodel.compute_inverse_kinematics(
self.ee_link_idx,
target_pose,
initial_qpos=self.articulation.get_qpos(),
active_qmask=self.qmask,
max_iterations=max_iterations,
)
Another person #6 has pointed out that pinocchio encounters issue when numpy>=2.0.
Installing numpy==1.24.4 (numpy<2.0) will resolve this issue.
I've updated the readme.
System: UBUNTU 20.04 Python: 3.10.14
All codes are set up with the introduction in README. vulkaninfo and python -m sapien.example.offscreen works fine, seems that there are no rendering issues.
Segmentation fault is located at: ManiSkill2_real2sim/mani_skill2_real2sim/agents/controllers/pd_ee_pose.py line 61: result, success, error = self.pmodel.compute_inverse_kinematics( self.ee_link_idx, target_pose, initial_qpos=self.articulation.get_qpos(), active_qmask=self.qmask, max_iterations=max_iterations, )
How can I solve this problem?