Super Agents of AI

What is It

RLBench is an ambitious large-scale benchmark and learning environment featuring 100 unique, hand-design tasks, tailored to facilitate research in a number of vision-guided manipulation research areas, including: reinforcement learning, imitation learning, multi-task learning, geometric computer vision, and in particular, few-shot learning. — From the official website

Paper and Code

Features

1. 100 completely unique, hand-designed tasks:
   1. Ranging in difficulty, from simple target reaching and door opening, to longer multi-stage tasks, such as opening an oven and placing a tray in it
   2. Each task comes with a number of textual descriptions and an infinite set of demonstrations using waypoint-based motion planning
2. Observations:
   1. Proprioceptive
      1. Joint angles
      2. Joint velocities
      3. Torques
      4. The gripper pose
   2. Visual: an over-the-shoulder stereo camera and an eye-in-hand monocular camera
      1. rgb
      2. depth
      3. segmentation masks

The scene in the V-REP environment:

Important Notes

RLBench employs 3 keys terms: Task, Variation, and Episode.

1. Each task consists of one or more variations
2. From each variation, an infinite number of episodes can be drawn
3. Each variation of a task comes with a list of textual descriptions that verbally summaries this variation of the task, which could prove useful for human robot interaction (HRI) and natural language processing (NLP) research

Each task has a completely sparse reward of +1 which is given only on task completion.

Users have a wide variety of action spaces at their disposal, which include:

• Absolute or delta joint velocities
• Absolute or delta joint positions
• Absolute or delta joint torque
• Absolute or delta end-effector velocities
• Absolute or delta end-effector poses

Building New Tasks

Each task has 2 associated files: a V-REP model file (.ttm), which holds all of the scene information and demo waypoints, and a python (.py) file, which is responsible for wiring the scene objects to the RLBench backend, applying variations, defining success criteria, and adding other more complex task behaviors.

Note: Cover Picture