Discontinuity issues with cpp examples

[manavkulshrestha]

Hi,

I'm running into some discontinuity issues when I run the c++ example (examples/generate_joint_position_motion_external_control_loop.cpp). I was getting similar issues on python (franky library) and was told it is likely that the slow speed of python is causing command packets to be dropped and causing discontinuity (slower speed often works for predefined movements).

Is there some other example I should be using as a reference which is not going to cause discontinuities? Is there something else going wrong? Any guidance for how/what to debug would be useful

I'm on robot firmware 8 (control 5.7.1), gripper firmware 3, and libfranka version 0.14.

talha@ideas-orin:~/mk/cpp$ g++ franka_server.cpp examples_common.cpp   -I/home/talha/miniconda3/envs/mkf311/include   -I/usr/include/eigen3   -lfranka -lpthread && ./a.out
WARNING: This example will move the robot! Please make sure to have the user stop button at hand!
Press Enter to continue...

Finished moving to initial joint configuration.
libfranka: Move command aborted: motion aborted by reflex! ["joint_motion_generator_velocity_discontinuity", "joint_motion_generator_acceleration_discontinuity"]

It seems inconsistent. Sometimes I get this error and other times it happens to work.

Context: My overall goal is to control the franka robot with teleop. To that end, I've managed to hook up a controller to simulation to get validated/collision free joint angles in real time. I just want to now command the robot without discontinuity errors.

EDIT: I changed lines 73-74 to this since I'm on an older library version

auto upper_velocity_limits = franka::computeUpperLimitsJointVelocity(reference_joint_position);
auto lower_velocity_limits = franka::computeLowerLimitsJointVelocity(reference_joint_position);

ping also seems fine

PING 192.168.131.41 (192.168.131.41) 56(84) bytes of data.
64 bytes from 192.168.131.41: icmp_seq=1 ttl=64 time=0.705 ms
64 bytes from 192.168.131.41: icmp_seq=2 ttl=64 time=0.801 ms
64 bytes from 192.168.131.41: icmp_seq=3 ttl=64 time=0.959 ms
64 bytes from 192.168.131.41: icmp_seq=4 ttl=64 time=0.759 ms

[AndreasKuhner]

Hi @manavkulshrestha ,
unfortunately, your ping is really bad: 64 bytes from 192.168.131.41: icmp_seq=4 ttl=64 time=0.759 ms -> ~0.8ms round trip time 😅

You would need to get that to around 0.2 ms to get a stable connection. Around 0.5 ms/0.6 ms it starts to get flaky. 0.8 ms is well beyond.

Are you using a good network card? Is your switch good? Do you have a direct connection to the robot?

Cheers,
Andreas

[manavkulshrestha]

I have made a couple changes:

1. updated from 5.7.1 to 5.9.2 (we are using clearpath ridgeback with 2 franka arms and that was the firmware it was shipped with). this made most of the difference. examples work now
2. directly connected the NUC to the robot controller instead of a switch (ordered a better Ethernet cable as well since we're at 0.6 ms, hasn't arrived yet). communication_test.cpp passes now, with about 95%

I'll try with a better cable. But, Another thing I want to know: what is the intended way to do teleoperation with the robot? if I have end effector poses (O_T_EE), is generate_cartesian_pose_motion_external_control_loop.cpp a good file to adapt or there's an alternate recommended path?

[AndreasKuhner]

The teleop question I would forward to @ignacio-pm

[ignacio-franka]

Hi @manavkulshrestha
We do not have teleop examples in libfranka directly but you can take inspiration from our ROS 2 controllers:

• FR3 to FR3: https://github.com/frankarobotics/franka_ros2_teleop/blob/main/src/teleop_follower_controller.cpp
• Joint-based teleop device: https://github.com/wuphilipp/gello_software/blob/main/ros2/src/franka_fr3_arm_controllers/src/joint_impedance_controller.cpp
• Cartesian-based teleop device: https://github.com/frankarobotics/franka_spacemouse/blob/main/src/franka_arm_controllers/src/joint_impedance_ik_controller.cpp