33. Simulate What You Want and Not What You Don't with Gazebo, with Louise Poubel

2023-06-09 · 1:36:54
gazebo
machine learning
open-source
reinforcement learning
simulation

In this episode of the Sense Think Act Podcast, host Audrow Nash interviews Louise Poubel, former lead of the Gazebo project and current Simulation Lead of Zipline. The conversation covers a range of topics, including the role of Gazebo, an open-source 3D robotics simulator, its comparison to other simulators, and its efforts to be more accessible to beginners. They also discuss the use of simulation in Machine Learning and Reinforcement Learning, the challenges in the field of simulation, and its application with physical robots. Louise shares her insights on the future of robotics, the importance of open-source projects, and the lessons she's learned from her experiences.


Links

Comments

Outline

  • 0:00:00 - Start
  • 0:01:29 - Introducing Louise Poubel
  • 0:04:06 - Introducing Gazebo
  • 0:06:11 - Clearing up Gazebo naming
  • 0:10:42 - Gazebo’s Garden release
  • 0:20:24 - Prioritizing what goes in a Gazebo release + working with the community
  • 0:26:17 - How does Gazebo compare to other simulators?
  • 0:30:37 - Gazebo’s plugin architecture
  • 0:36:19 - Making Gazebo beginner friendly
  • 0:40:11 - Gazebo as a “Generic simulator”
  • 0:46:55 - Simulation for Machine Learning and Reinforcement Learning
  • 0:52:42 - Gazebo and the web
  • 0:59:24 - Hard Challenges in simulation
  • 1:08:51 - How simulation can be used with physical robots
  • 1:10:49 - Lessons learned from leading Gazebo
  • 1:14:56 - Why Open Source?
  • 1:16:15 - Moving to Zipline
  • 1:24:21 - How to learning Gazebo
  • 1:31:47 - Future of robotics
  • 1:35:49 - Trying Gazebo