ntuck@neu

Gazebo

  • Physical robots can be inconvenient
  • Especially, they tend to be both expensive and fragile
  • This is really obvious with rockets, but it's also true for less dangerous robots like the Ranger.
  • Again, the solution is simulation.
  • To simulate wheeled robots, we'll be using Gazebo
  • Gazebo is an open source robot simulator that's commonly used in academia and sometimes used in industry.
  • It's normally used with the ROS software stack, but we're just going to use it directly.

Demo and HW01

  • In HW01, you'll be setting up your computer to run Gazebo and then writing a simple control program for a simulated robot.
  • Gazebo supports running on Linux. Officially, they require a recent release of Ubuntu.
  • I'm running Debian Linux (very similar to Ubuntu) to develop the assignments.
  • The two supported configurations will be Debian 10 or Ubuntu 20.04 directly installed (e.g. dual boot) on your personal computer.
  • If you can get it to work some other way, that's fine, but if it doesn't the course staff can't help you with your unsupported setup.
  • Gazebo does OpenGL rendering, which is likely to break several common alternate setups (e.g. running in a normal VM).

Show the solution code running.

  • Once you have Linux and Gazebo installed, you're ready for HW01.
  • Download the starter code from Inkfish.
  • Run Gazebo with ./sim.sh
  • Goal: Write a control program that drives the robot car from start to checkerboard.
  • Show the brain directory.
  • "Write a control program".
  • Run the solution brain program.
  • Sensors: Laser scanner, pose sensor.
  • Actuators: Turn wheels, steering.

Submitting the Homework

  • Create a "brain-hw01" directory.
  • Edit the README, add a video (the HW01 solution video)
  • make clean
  • Create a tarball
  • Submit on Inkfish