billtubbs / gym-cartpole-bt-v0 Goto Github PK

I.e.

d(k) = C(z)/D(z)e(k)

where C(z) and D(z) are polynomials that can be specified and e(k) is the current random disturbance variable added to the angle state.

x3(k) = x3(k) + d(k)

Pendulum angle (theta) is the reverse of how it is defined in CartPole-v1. In CartPole-v1, theta = 0 is the pendulum in the vertical-up position which make sense, whereas here theta = 0 is the pendulum in the vertical-down (stable) position.

Consider changing it for consistency.

(This will also affect how we deal with angle wrap-around and the calculation of rewards).

At the moment, you should use the following command to run the unit tests:

python -m unittest test/test_cartpole_bt_env.py

If you run the usual

python -m unittest

It also runs the scripts that start with 'test_'!

In this variant, the pendulum would start in the down position and the goal is to swing it up to the vertical up position. To make this even more challenging the weight of the pendulum could be adjusted so that multiple swings are needed to generate enough energy in the system for the final swing-up.

This will be good for testing observers. Propose to use the labels '-mL' and '-mH' for this (low and high measurement noise).