View on GitHub

Guanxiong Chen

just a CPEN dude

NSERC USRA Research Intern at UBC Lab for Computational Intelligence

May 2021 to August 2021

Volunteered from May 2020 to April 2021


Under Prof. Ian Mitchell’s supervision, I have been working on a project aimed for building an interface between ROS and AI Habitat. Research areas involved are physics-based simulation and robotics.

What is AI Habitat and What is ROS? Why Together? And what should the Interface Do?

What is AI Habitat?

AI Habitat is a reinforcement learning framework used for training and evaluating embodied agents on various tasks, such as visual navigation and language navigation. The framework consists of two parts: 1) Habitat Lab, an API for task definition, training, testing, and 2) Habitat Sim, a simulator.

What is ROS?

ROS is a robotics middleware which allows us to develop robotics software on top of. For example, we can construct a SLAM-based agent and simulate it within the Gazebo simulator under the ROS framework.

Why Together?

AI Habitat provides novel methods for defining tasks and RL-based agents, and a simulator better in photo-realism than Gazebo, the canonical choice of simulator under ROS. ROS on the other hand is where most traditional planners, drivers and visualization tools reside. So we aim to bridge the gap between the frameworks.


Coming soon!

Code Access

You can check our source code on our Github repo.