Skip to content

Increased Failure Rate of Gazebo Service Calls During Long Simulations #564

@paoloelle

Description

@paoloelle

Environment

  • OS Version: Ubuntu 22.04
  • Source installation tested with Gazebo Fortress and with Gazebo Harmonic (using Apptainer containerization)

Description

I'm running an evolutionary algorithm using ROS 2 Humble and Gazebo Fortress (as I said before, I also tested it with ROS 2 Jazzy and Gazebo Harmonic via Apptainer containerization). To do this, I need to call Gazebo services to stop/start the simulation and move various objects in the environment, including my robot. What I noticed is that after calling these commands, the simulations' performance slows down, especially because the service call often fails. However, the real-time factor doesn't decrease.

To test this undesired behavior, I implemented a simple simulation scenario, without any evolutionary algorithm on top of it. This simulation consists of just stopping the simulation, moving the robot and the objects, and restarting the simulation for 10 seconds. An example is shown in the video below.

simple_sim-2024-12-20_09.56.47.mp4

I let this simple simulation run for 24 hours and plotted the related results.
These show the total number of service calls for each hour of simulation, the service calls that didn't give back a successful answer, and the percentage of service calls that gave a successful answer. Please ignore the hour 25.

Total Service Calls Per Hour

Failed Service Calls Per Hour

Success Percentage of Service Calls Per Hour

Metadata

Metadata

Assignees

Labels

bugSomething isn't working

Type

No type

Projects

Status

In progress

Milestone

No milestone

Relationships

None yet

Development

No branches or pull requests

Issue actions