Robot Learning

Policy architectures for robot manipulation and imitation learning.

World Models and Model-Based RL

Modules for model-based reinforcement learning, including world models and dynamics models.

WorldModelWrapper(*args, **kwargs)	World model wrapper.
DreamerActor(out_features[, depth, ...])	Dreamer actor network.
ObsEncoder([channels, num_layers, ...])	Observation encoder network.
ObsDecoder([channels, num_layers, ...])	Observation decoder network.
RSSMPosterior([hidden_dim, state_dim, ...])	The posterior network of the RSSM.
RSSMPrior(action_spec[, hidden_dim, ...])	The prior network of the RSSM.
RSSMRollout(*args, **kwargs)	Rollout the RSSM network.

PILCO

Components for moment-matching model-based policy search (PILCO).

GPWorldModel(obs_dim, action_dim[, in_keys, ...])	Gaussian Process world model with moment-matching uncertainty propagation.
RBFController(input_dim, output_dim, max_action)	Radial Basis Function controller for moment-matching policy search.