See store_transition.

See store_transition.

Perform the training operation.

See update.

See update.

Return the action space.

Return the environment's full observation space.

Return if an object can be moved in the x-direction. Parameters ---------- movable : int move-ability of the contacting object

Return if an object can be moved in the y-direction. Parameters ---------- movable : int move-ability of the contacting object

Return if an object can be moved in the z-direction. Parameters ---------- movable : int move-ability of the contacting object

Ensure environment configurations were properly entered. This is done via a sequence of assertions. Parameters ---------- model_name

Clear internal memory that is used by the replay buffer.

Perform the sample collection operation over a single step. This method is responsible for executing a single step of the environment

See class definition.

Calculate the conjugate gradient of Ax = b. Based on https://epubs.siam.org/doi/book/10.1137/1.9781611971446 Demmel p 312. P

Construct the structure of the maze. Parameters ---------- maze_id : str type of environment. This dictates the structure the maz

Return the environment's context space.

Add the algorithm hyperparameters to the parser.

Add the feedforward policy hyperparameters to the parser.

Add the goal-conditioned policy hyperparameters to the parser.

Add the multi-agent policy hyperparameters to the parser.

Add the PPO hyperparameters to the parser.

Add the SAC hyperparameters to the parser.

Add the TD3 hyperparameters to the parser.

Add the TRPO hyperparameters to the parser.

See parent class.

See parent class.

Draw the trajectory when using the AntFourRooms environment. Parameters ---------- context : [float, float] the (x,y) coordinates

Draw the trajectory when using the AntGather environment. Parameters ---------- objects : list of (float, float, int) list of app

Draw the trajectory when using the AntMaze environment. Parameters ---------- context : [float, float] the (x,y) coordinates of t

Compute the bootstrapped/discounted returns. Parameters ---------- mb_rewards : array_like a minibatch of rewards from a given en

See parent class.

Call the actor methods to compute policy actions. Parameters ---------- obs : dict of array_like the observations

Collect the contextual term. None if it is not passed.

Return the current position.

Collect the algorithm-specific hyperparameters from the checkpoint. Parameters ---------- ckpt_path : str the path to the checkpo

Return the initial observation from the environment.

Return a processed lane number.

Return an end goal. Returns ------- array_like the end goal

Return the orientation of the humanoid.

Return the orientation.

Return egocentric range sensor observations of maze.

Return egocentric range sensor observations of maze.

Collect the spaces of the environments. Returns ------- gym.spaces.* the action space of the training environment

See parent class.

See parent class.

See class definition.

Get target update operations. Parameters ---------- _vars : list of tf.Tensor the initial variables target_vars : list of tf.

See parent class.

Return dict map for the summary (to be run in the algorithm).

Convert a batch to a td_map.

Convert a batch to a td_map.

Convert a batch to a td_map.

Return the x,y position of the agent.

Return the environment's time horizon.

Initialize the policy. This is used at the beginning of training by the algorithm, after the model parameters have been initialized.

Check whether a policy is a feedforward policy.

Iterate over arrays in batches. Must provide either num_batches or batch_size, the other must be None. Parameters ----------

Load model parameters from a checkpoint. Parameters ---------- load_path : str location of the checkpoint

Update the dataset with current step data. Parameters ---------- obs : array_like the current observation

Execute multiple training operations.

Execute multiple training operations.

Execute multiple training operations.

Execute multiple training operations.

Execute multiple training operations.

Create the actor variables. Parameters ---------- obs : tf.compat.v1.placeholder the input observation placeholde

Create an actor tensor. Parameters ---------- obs : tf.compat.v1.placeholder the input observation placeholder

Create the actor variables. Parameters ---------- obs : tf.compat.v1.placeholder the input observation placeholde

Create a critic tensor. Parameters ---------- obs : tf.compat.v1.placeholder the input observation placeholder

Return a session that will use <num_cpu> CPU's only. Parameters ---------- num_cpu : int number of CPUs to use for TensorFlow

Return the observation space.

Parse plotting options user can specify in command line.

Parse training options user can specify in command line. Returns ------- argparse.Namespace the output parser object

Return the distance between two points. Parameters ---------- p1 : (float, float) (x,y) values of point 1 p2 : (float, float)

Return the inverse of a quaternion.

Return the inverse of a quaternion.

Check ray intersection with segment. Check if the ray originated from (x, y) with direction theta intersects the line segment (x1, y1) -- (x2

Render the environment.

Render the environment.

See parent class.

Render the environment.

Reset the environment. If the environment is using the contextual setting, a new context is issued. Returns -------

See parent class.

Reset the environment.

See parent class.

Save, and potential plot, the trajectory data. Parameters ---------- fp : str the path to save the data to

Save the parameters of a tensorflow model. Parameters ---------- save_path : str Prefix of filenames created for

Set the goal position of the agent. Parameters ---------- goal : array_like the desired position of the agent

Update the goal position of the goal agent.

Create the graph, session, policy, and summary objects.

Create the environment and collect the initial observations. Parameters ---------- env : str the name of the envi

Advance the simulation by one step.