Skip to content
ARAKAWA, Naoya edited this page Dec 2, 2019 · 11 revisions

PyLIS (Life in Silico with PyBullet) is a 3D simulation environment for research in human level artificial intelligence (HLAI) and artificial general intelligence (AGI)[15][4]. PyLIS includes agent (robot) models with wheels for locomotion and a manipulator; more specifically:

  1. Software library for:

    a. reading robot models into the virtual 3D space with a physics engine

    b. obtaining color images and depth images from the head camera

    c. moving by wheels

    d. manipulating objects with a manipulator

    e. detecting collision with other objects

  2. A sample environment and agents to perform reinforcement learning with the library.

PyLIS can be used in research involving both manipulation and locomotion capabilities. The combination of manipulation and locomotion in tasks makes them much more complex than cases where those capabilities are tested separately (because of its expanded search space).

PyLIS can also be used for research in social interaction with object manipulation/operation with more than one agent, including research in acquisition of language involving object expressions.

PyLIS_Manual2019Fig1

Table of Contents

The Environment and Robot Models

Software Library Specifications

Deep reinforcement learning with PyLIS

How to install

Example runs

Available environments

References

Clone this wiki locally