#TASK COACH OFFLINE FULL#

The two road configurations we include are shown above: a single-lane ring environment and a highway merge intersection. The Flow environment involves controlling the acceleration of autonomous vehicles (1 in ring, up to 5 in merge) in order to maximize traffic flow.

All trajectories now timeout at 1000 steps.

metadata/policy/*: Contains the weights of the policy used to collect the data.

infos/qpos and infos/qvel: State information.

infos/action_log_probs: The log-probabilities for actions in the dataset.

However, the *-v2 datasets have improved metadata: The *-v0 datasets were used to generate the results reported in our whitepaper and are included for backwards compatibility.

Medium-Expert uses a 50-50 split of medium and expert data (slightly less than 2M samples total).

Timeouts in this dataset are not always marked when the agent reaches the max trajectory length, but rather when 1000 timesteps have been sampled for a particular training iteration.

Medium-Replay uses the replay buffer of a policy trained up to the performance of the medium agent.Medium uses 1M samples from a policy trained to approximately 1/3 the performance of the expert.Expert uses 1M samples from a policy trained to completion with SAC.Random uses 1M samples from a randomly initialized policy.The *-human tasks now have timeouts set when the demonstrations end, rather than at the maximum horizon of the environment.

#TASK COACH OFFLINE FULL#

Assorted information providing full environment state ( infos/qpos, infos/qvel, and task-specific fields).However, the *-v1 datasets have improved metadata: Expert uses 5000 trajectories sampled from an expert that solves the task, provided in the DAPG repository.The demonstration trajectories are copied to match the number of behavioral cloned trajectories. Cloned uses a 50-50 split between demonstration data and 2500 trajectories sampled from a behavioral cloned policy on the demonstrations.Human uses the 25 human demonstrations provided in the DAPG repository.

There are 3 datasets for each environment. Clockwise from the top left, they are pen (aligning a pen with a target orientation), door (opening a door), relocate (move a ball to a target position), and hammer (hammer a nail into a board). There are 4 tasks, from the hand_dapg repository. The Adroit domain involves controlling a 24-DoF robotic hand. Two datasets are provided: minigrid-fourrooms-v0, which is generated by a controller that randomly samples goal locations and navigates to them, and minigrid-fourrooms-random-v0, which samples actions uniformly at random. The Minigrid domain is a discrete analog of Maze2D. The 'play' dataset is generated by commanding specific hand-picked goal locations from hand-picked initial positions. The dataset in 'antmaze-umaze-v0' is generated by commanding a fixed goal location from a fixed starting location (these are the opposite sides of the wall in the umaze).įor harder tasks, the 'diverse' dataset is generated by commanding random goal locations in the maze and navigating the ant to them. The AntMaze domain uses the same umaze, medium, and large mazes from the Maze2D domain, but replaces the agent with the 'Ant' robot from the OpenAI Gym MuJoCo benchmark. The four environments maze2d-open-v0, maze2d-umaze-v0, maze2d-medium-v0, maze2d-large-v0 use a sparse reward which is has a value of 1.0 when the agent (light green ball) is within a 0.5 unit radius of the target (light red ball).Įach environment has a dense reward version, which instead uses the negative exponentiated distance as the reward. The four maze layouts are shown below (from left to right: open, umaze, medium large): However, for the purposes of being able to split the trajectory into smaller subtrajectories, we use the timeout field to denote instances when the randomly selected navigation goal has been reached. The dataset consists of one continuous trajectory of the agent navigating to random goal locations, and thus has no terminal states. The observation consists of the (x, y) location and velocities. The Maze2D domain involves moving force-actuated ball (along the X and Y axis) to a fixed target location.

Task coach offline