designed data driven gaussian process controllers for 6 different gaussian process methods:
1,2. affine dense kernel and corresponding affine dense random fourier features 3,4. affine dot product kernel and corresponding affine dot product random fourier features 5,6. vanilla kernel method and vanilla random fourier features
plots and experiments implemented for all controllers on inverted pendulum.
work in progress, more to come!
To initialize the submodule dynamics library:
git submodule update --init --recursive
Then to setup
cd core; python setup.py develop
cd fast_control; python -e setup.py
episodic usage
gps_names options are ad_kernel, adp_kernel, ad_rf, adp_rf