MPCPlannerBase

class torchrl.modules.MPCPlannerBase(*args, **kwargs)

MPCPlannerBase abstract Module.

This class inherits from SafeModule. Provided a TensorDict, this module will perform a Model Predictive Control (MPC) planning step. At the end of the planning step, the MPCPlanner will return a proposed action.

Parameters:

• env (EnvBase) – The environment to perform the planning step on (Can be ModelBasedEnvBase or EnvBase).

• action_key (str, optional) – The key that will point to the computed action.

forward(tensordict: TensorDictBase, tensordict_out: Optional[TensorDictBase] = None, **kwargs) → TensorDictBase

When the tensordict parameter is not set, kwargs are used to create an instance of TensorDict.

abstract planning(td: TensorDictBase) → Tensor

Performs the MPC planning step.

Parameters:

td (TensorDict) – The TensorDict to perform the planning step on.