Quadruped Tasks#

These are tasks where a quadruped robot is used for locomotion and/or manipulation. This cateogry primarily uses robots with four legs like the ANYmal or Unitree go robots. The document here has both a high-level overview/list of all tasks in a table as well as detailed task cards with video demonstrations after.

Task Table#

Table of all tasks/environments in this category. Task column is the environment ID, Preview is a thumbnail pair of the first and last frames of an example success demonstration. Max steps is the task’s default max episode steps, generally tuned for RL workflows.

Task

Preview

Dense Reward

Success/Fail Conditions

Demos

Max Episode Steps

AnymalC-Reach-v1
AnymalC-Reach-v1 AnymalC-Reach-v1

✅

✅

✅

200

UnitreeGo2-Reach-v1

✅

✅

❌

200

AnymalC-Spin-v1
AnymalC-Spin-v1 AnymalC-Spin-v1

✅

✅

❌

200

AnymalC-Reach-v1#

dense-reward sparse-reward demos

Task Card

Task Description: Control the AnymalC robot to reach a target location in front of it. Note the current reward function works but more needs to be added to constrain the learned quadruped gait looks more natural

Randomizations:

  • Robot is initialized in a stable rest/standing position

  • The goal for the robot to reach is initialized 2.5 +/- 0.5 meters in front, and +/- 1 meters to either side

Success Conditions:

  • If the robot position is within 0.35 meters of the goal

Fail Conditions:

  • If the robot has fallen over, which is considered True when the main body (the center part) hits the ground

Goal Specification:

  • The 2D goal position in the XY-plane

AnymalC-Spin-v1#

dense-reward sparse-reward

Task Card

Task Description: Control the AnymalC robot to spin around in place as fast as possible and is rewarded by its angular velocity.

Randomizations:

  • Robot is initialized in a stable rest/standing position

Fail Conditions:

  • If the robot has fallen over, which is considered True when the main body (the center part) hits the ground