search

AutoRally / autorally

Software for the AutoRally platform
http://autorally.github.io
727 stars 228 forks source link

Simulation Instructions / Tutorial #34

Closed ShawnSchaerer closed 6 years ago

ShawnSchaerer commented 8 years ago

I think there is a need to have instructions on how to run the system in Gazebo. I can run the simulator after installing as per the document, but its not clear on how to launch the rest of the system to run in simulation mode.

I tried to launch the autorally_platform but there are too many dependencies on physical hardware and the simulation does not work.

Is there a simulation tutorial somewhere? Thanks

bgoldfai commented 8 years ago

Yeah we have some tutorials on the wiki to run the system autonomously that work in simulation and with a physical platform. If you go through them in the following order (one tutorial for each node involved) you should have the vehicle driving autonomously in simulation:

Would something like an overall tutorial or better section at the end of the README that lays that procedure out explicitly help? Or is there something larger I'm missing that we should add tutorial wise?

ShawnSchaerer commented 8 years ago

Thanks, I will go through and let you know what I think. It was not clear to me on how to proceed base on the ##What's Next on the bottom of the readme.

Maybe just a better section at the end of the readme for now.

bgoldfai commented 8 years ago

OK I'll add to that section.

ShawnSchaerer commented 8 years ago

I tried to run those tutorials in order but I could not get the car to move in gazebo (ROS Kinetic on Ubuntu 16.04). I did get it to move when I set the speed manually using rqt_graph. Perhaps I have not configured everything correctly or there is a bug with Kinetic? I will try again tomorrow.

bgoldfai commented 8 years ago

On my 16.04 test setup it looks like the simulated GPS and IMU aren't publishing data looking at topics /gpsRoverStatus and /imu/imu (I don't even see these topics). Without those the state estimator and the waypoint follower won't function. I'll continue to debug, this may just be a topic remapping issue as I can see /autorally_platform/imu and /autorally_plaform/gpsRoverStatus.

The constant speed controller appears to control the throttle, as it only relies on the /wheelSpeeds topic. If you are using a gamepad, make sure motionEnabled in the /runstop message coming from the joystick node is true. You can toggle it with any of the buttons next to the right stick.

If you aren't using a gamepad:

ShawnSchaerer commented 8 years ago

Thanks, I will try the constant speed controller and see what happens.

bgoldfai commented 8 years ago

OK so I think I solved the issue by publishing the simulated IMU and GPS on the expected topics directly instead of remapping them. The most concise procedure to get the platform driving autonomously once the simulator is up and running with the robot positioned where it starts, and you have verified that /runstop enableMotion is true from your /runstop source:

Start state estimator:

Start waypoint follower:

Start constant speed controller:

If the robot turns and hits the barrier it's probably because the state estimator is not converged, so its orientation estimate is incorrect. Just select the track barriers and move them up to allow the robot to continue driving, and the estimator should converge and the vehicle will return to within the barriers.

ShawnSchaerer commented 8 years ago

Ok thanks. I will try it out and post results

On Wed, Aug 10, 2016 at 11:06 AM, Brian Goldfain notifications@github.com wrote:

OK so I think I solved the issue by publishing the simulated IMU and GPS on the expected topics directly instead of remapping them. The most concise procedure to get the platform driving autonomously once the simulator is up and running with the robot positioned where it starts, and you have verified that /runstop enableMotion is true from your /runstop source:

Start state estimator:

  • in autorally_core/launch/state_estimator.launch change InvertY and InvertZ to false
  • rosparam set /gps_imu/FixedInitialPose true
  • roslaunch autorally_core state_estimator.launch

Start waypoint follower:

  • roslaunch autorally_control waypointFollower.launch

Start constant speed controller:

  • roslaunch autorally_control constantSpeedController.launch
  • rosrun rqt_publisher rqt_publisher -- publish topic constantSpeedController/speedCommand of type std_msgs_Float64 at rate 10 with value of 3. This is the target velocity in m/s and as soon as you do this the platform should move, assuming motion is enabled.

If the robot turns and hits the barrier it's probably because the state estimator is not converged, so its orientation estimate is incorrect. Just select the track barriers and move them up to allow the robot to continue driving, and the estimator should converge and the vehicle will return to within the barriers.

— You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub https://github.com/AutoRally/autorally/issues/34#issuecomment-238916262, or mute the thread https://github.com/notifications/unsubscribe-auth/ANrTFYlrX7XSCV5SY3GPWhRP7D4pr0Glks5qefcMgaJpZM4Jgtdl .

ShawnSchaerer commented 8 years ago

I verified that this procedure works.
Its good enough for me to learn the code base now. :)

I would suggest that we put that procedure for the sim in the readme.

bgoldfai commented 8 years ago

Awesome, thanks for taking the time to work through all this with me. I'll add that procedure to the readme then refer to the wiki for more information about each of the nodes for the curious.

ShawnSchaerer commented 8 years ago

no problem. I am going to be adapting parts of the system to a new platform and trying to see what happens if we swap out the GTSAM with Robot_localization.

great work so far :)

On Thu, Aug 11, 2016 at 7:39 AM, Brian Goldfain notifications@github.com wrote:

Awesome, thanks for taking the time to work through all this with me. I'll add that procedure to the readme then refer to the wiki for more information about each of the nodes for the curious.

— You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub https://github.com/AutoRally/autorally/issues/34#issuecomment-239148990, or mute the thread https://github.com/notifications/unsubscribe-auth/ANrTFTP9OZRriEv5u6orZFiO-2hxmFS8ks5qexgDgaJpZM4Jgtdl .

dstoyer commented 7 years ago

This is awesome work, so my following comment is intended to hopefully improve the installation experience.

I have installed the autorally software platform in Ubuntu 14.04 and Unbuntu 16.04 (using the kinetic-devel branch). In both cases I had to clone imu_3dm_gx4 into catkin_ws/src. Installation went smoothly by following the current instructions with these steps included: After cloning the appropriate autorally into catkin_ws/src

  1. clone imu_3dm_gx4 (or download and unzip) into the same directory
  2. run "rosdep install --from-path src --ignore-src -y" from catkin_ws Continue by running catkin_make from catkin_ws/

I don't know if imu_3dm_gx4 is supposed to be installed as a dependency automatically with the "rosdep install ..." line, but it currently is not.

I have tested the install an 3 different systems and VMware workstation. The install is successful on all of them when adding those two steps (for Gazebo in VMware, OpenGL needs to be set to 2.0: export SVGA_VGPU10=0).

Again, you guys are doing awesome work, hopefully this will either clarify the installation instructions, or point out a bug in the installation implementation.

Cheers, Dan

bgoldfai commented 7 years ago

Hi, thanks for the comment and the enthusiasm for the project!

To make sure I understand correctly, you are working with the simulated AutoRally platform in Gazebo and not on a physical AutoRally platform, and the installation/compilation process fails if the imu_3dm_gx4 repository is not cloned into your catkin workspace using the 2 steps you describe?

I ask because the instructions used to include cloning imu_3dm_gx4 exactly where you are suggesting, but that step was moved into the first section of the platform configuration instructions because that repository is (in theory) only needed if you are installing on a physical AutoRally platform with a 3DM-GX4-25 IMU. The simulated IMU comes from hector_gazebo_plugins which should automatically install in the rosdep install step.

dstoyer commented 7 years ago

To make sure I understand correctly, you are working with the simulated AutoRally platform in Gazebo and not on a physical AutoRally platform, and the installation/compilation process fails if the imu_3dm_gx4 repository is not cloned into your catkin workspace using the 2 steps you describe?

Correct.

the instructions used to include cloning imu_3dm_gx4 exactly where you are suggesting, but that step was moved into the first section of the platform configuration instructions

Admittedly, I did not go through the platform installation instructions. I did skim over it but moved on to the software instructions since I don't have any of the hardware and saw that the software could be run in simulation.

that repository is (in theory) only needed if you are installing on a physical AutoRally platform with a 3DM-GX4-25 IMU.

Specifics on issues I ran into without imu_3dm_gx4 present:

Hope that helps,

Dan

bgoldfai commented 7 years ago

This is all great and thanks for taking the time to put this information together!

I added your suggested step back into the instructions since the state estimator does need a custom message from the imu repo looking at the error message you provided and the code (I didn't think we had that dependency and didn't check compilation on a machine without the imu code). That change is in the devel branch now, and should make its way into the kinetic-devel next week.

Strictly speaking, GTSAM is not required for the simulation world to run. GTSAM is required to compile and run the state estimator, which provides a fused state estimate for the robot and is required to run the waypoint follower.

Let us know if you have any other questions or issues.

ksoltani commented 6 years ago

Hi all, Thanks for the discussions. I have been trying to get the car execute wayPointFollower. I set the speed using the constantSpeedController and through rqt_publisher. The car does move but it never steers. I suspect somehow wayPointFollower is not actually changing the state of the car. Have you tried this? Did you have similar issues?

bgoldfai commented 6 years ago

What OS, ROS version, and autorally branch are you working in, and are any errors generated on the command line that you can provide for debugging? I think there is a known issue in the kinetic_devel branch where the waypoint follower crashes on launch that I'll look into if I can replicate it on my machine.

Another thing to check is to look at the steeringCommander in the /chassisStatus topic to see what is in control of the steering to verify that commands from the waypoint follower are being received.

ksoltani commented 6 years ago

Hi Brian, I was using grady-kinetic-devel branch. I am running Ubuntu 16.04, ROS Kinetic. Once I used kinetic-devel branch I was able to get it to execute wayPointFollower control using throttle setting in ocs. I did not test with speed controller.

I also remember I had to change OpenCV version requirement to 3 instead of 2 before it could build the code.

bgoldfai commented 6 years ago

I'm glad kinetic-devel worked, thanks for letting us know. We've encountered the OpenCV version problem as well and plan to fix it in the next release and remove the OpenCV exact version requirement in the CMakeLists file.

bgoldfai commented 6 years ago

The suggestions in this issue have been incorporated into the current kinetic-devel branch, everything seems to be working, so I'm closing.