What can we get out of MuJoco
?
env = gym.make('Reacher-v2')
obs,info = env.reset()
for tick in range(1000):
env.render()
action = policy(obs)
obs,reward,done,_ = env.step(action)
For those who have run the code above, you are already running MuJoCo
under the hood. However, MuJoCo
is not just some physics engines that simulates some robots. In this repository, we focus on the core functionalities of MuJoCo
(or any other proper simulators) and how we can leverage such information in robot learning tasks through the lens of a Roboticist.
In particular, we will distinguish kinematic
and dynamic
simulations (e.g., forward/inverse kinematics/dynamcis).
Contact: [email protected]