Hi and thanks for your great work on this project — it's an impressive integration of Robot manipulation!
While going through the codebase, I had a couple of questions regarding the implementation in plan/callback.py:
1.
In the update() function, there's a line:
spec_score = self.spec(traj_tensor)
However, I couldn't find where self.spec is explicitly initialized in the Updater class. Could you clarify where this attribute is set, or whether it's assumed to be passed from a higher-level module?
2.
Is spec_score intended to represent the task-level reward as described in the paper, used to optimize the planning module according to LTL specifications?
Thanks again for the open-source release — it’s been very helpful for my research.
Best regards,
Hi and thanks for your great work on this project — it's an impressive integration of Robot manipulation!
While going through the codebase, I had a couple of questions regarding the implementation in plan/callback.py:
1.
In the update() function, there's a line:
spec_score = self.spec(traj_tensor)
However, I couldn't find where self.spec is explicitly initialized in the Updater class. Could you clarify where this attribute is set, or whether it's assumed to be passed from a higher-level module?
2.
Is spec_score intended to represent the task-level reward as described in the paper, used to optimize the planning module according to LTL specifications?
Thanks again for the open-source release — it’s been very helpful for my research.
Best regards,