Using curriculum learning to train the Kuka robot

The first step before actually kicking off some training is to customize the Kuka class as well as the KukaGymEnv to make them work with the curriculum learning parameters we described above. So, let's do that next.

Customizing the environment for curriculum learning

First, we start with creating a CustomKuka class which inherits the original Kuka class of PyBullet. Here is how we do it:

Chapter14/custom_kuka.py

1. We first need to create the new class, and accept an additional argument, jp_override dictionary, which stands for joint position override.

   class CustomKuka(Kuka):
       def __init__(self, *args, 
                    jp_override=None, **kwargs):
           self.jp_override = jp_override
           super(CustomKuka, self).__init__(*args...