maxspahn / gym_envs_urdf

URDF environments for gym

Home Page:https://maxspahn.github.io/gym_envs_urdf/

Geek Repo:Geek Repo

Github PK Tool:Github PK Tool

Goals do not induce a done state from the environment step function

Capsar opened this issue · comments

Hi,

How can we use the goals to end the environment?
I have a robot that is able to move into the goal region with epsilon given.
However, the environment does not terminate / the done boolean does not seem to update.

Furthermore, how can we define more complex goals, for instance where a manipulator has to be in a specific state.
Now it only seems to accept x,y,z coordinates.
Is there some documentation on how to define the goal / multiple goals?

Kind regards,
Caspar

Hi,

That is a correct observation. In order to be able to terminate, it is necessary to have access to a forward kinematics function, which is hard to have. The other option is to check the state of the specified link directly in bullet.

You can on the other hand specify joint space goals using StaticJointSpaceSubGoal. Multiple goals can be compose as GoalComposition.

Let me know, if you need further help on that.