Interface for visualizing Q_risk for all navigation environments.

Applies the action and steps simulation :param action: action at time-step :return: obs dict where: -each k

Renders the enviornment. Implements custom rendering support. If mode is: - dual: renders both left and main cameras - left:

Resets this controller (clears previous solution, calls all update functions). Returns: None

Arguments: states, actions --> list of states and list of corresponding actions to get Q_risk values for

Train task policy and associated Q function with experience in replay buffer (memory)

Creates a MjSim from passed in model_path :param model_path: Absolute path to model file :return: None

:return: mean of the distances between all objects and goals

Get action from current task policy

Trains the internal model of this controller. Once trained, this controller switches from applying random actions to using MPC. Argum

samples a random action from the action space.

Creates controller configuration.

Returns the action that this controller would take at time t given observation obs. Arguments: obs: The current observation

:return: mean of the distances between all objects and goals

Construct experiment map for this experiment