No acceleration limiting in Gazebo diff-drive simulation

[147258369ymc]

Description

The TurtleBot3 Gazebo simulation allows physically impossible accelerations.
The XL430-W250 Dynamixel motor has a maximum achievable linear acceleration of
approximately 5 m/s² (derived from stall torque and robot mass). However, in
simulation, step changes in cmd_vel produce accelerations exceeding 50 m/s².

This is caused by:

1. max_wheel_torque in the SDF model set to 20 Nm (actual XL430 stall torque: ~1.4 Nm)
2. No acceleration rate limiter in the diff-drive plugin

Steps to Reproduce

ros2 launch turtlebot3_gazebo empty_world.launch.py

Send alternating velocity commands rapidly:

ros2 topic pub -r 3 /cmd_vel geometry_msgs/msg/Twist "{linear: {x: 0.22}}"

Then immediately:

ros2 topic pub -r 3 /cmd_vel geometry_msgs/msg/Twist "{linear: {x: -0.22}}"

Monitor odometry and compute acceleration from velocity differences

Observed

Linear acceleration: 55.14 m/s² (11x physical limit)
Angular acceleration: 7194 rad/s² (240x physical limit)

Suggested Fix

1. Set max_wheel_torque to ~1.4 Nm (XL430-W250 stall torque)
2. Add acceleration rate limiting in the diff-drive plugin or SDF joint parameters

[sungjinjung2806]

Thanks for reporting this.
I tested a similar velocity-switching scenario in my environment and observed that the acceleration was high, but lower than the values reported here.
Environment:
Ubuntu 24.04
ROS 2 Jazzy
TurtleBot3 Burger

In my tests, the estimated linear acceleration was around 22 m/s², and angular acceleration was around 47 rad/s².
This suggests that acceleration behavior may vary depending on simulation setup, Gazebo version, or controller configuration.
That said, this still seems like a limitation of simulation fidelity, and improving acceleration constraints or motor dynamics could help make the simulation more physically realistic.