Hi Chris,

I've been developing the Teensy 4.1 firmware further and am currently working on a dynamics observer, with the longer-term goal of dynamic/gravity compensation.

I've already implemented numerical Jacobian/singularity monitoring and am now looking to populate a rigid-body dynamics model of the AR4 Mk1. Do you have any of the following mechanical/dynamic parameters available from the AR4 CAD or original design calculations?

For each moving link / joint assembly J1–J6:

• Mass of each link/assembly (kg)
• Centre of mass location for each link (XYZ), and which coordinate frame the COM is referenced to
• Moment/inertia tensor about the link COM: Ixx, Iyy, Izz and, if available, Ixy, Ixz, Iyz
• Confirmation of the link coordinate frames used for those values

For the drivetrain:

• Gear reduction for J1–J6
• Any belt/pulley reductions in addition to the gearbox reduction
• Motor rotor inertia for each motor, if known
• Gearbox/output inertia or reflected inertia, if available
• Approximate gearbox efficiency for each joint, if known
• Any known static/Coulomb friction or viscous-friction values

For the kinematic model:

• The definitive AR4 DH parameters / link geometry used by the controller
• Confirmation of joint-axis directions/sign conventions
• Any mechanical zero offsets that aren't represented directly in the DH table

For the wrist/tool:

• Mass and COM of the standard J6/flange assembly
• Rated/default payload
• Payload COM assumptions used when the AR4 was designed
• Any maximum joint torque figures or calculated torque requirements for J1–J6

I'm particularly interested in whether you already have a mass-properties export from the CAD. If you have the original CAD assemblies with materials assigned, even a table/export of mass, COM and inertia tensor for each moving assembly would be ideal.

I'm initially running the dynamics as an observer only — calculating gravity, inertial and Coriolis/centrifugal loading without changing motion — so I can validate the model against encoder following error before implementing any active compensation.

Cheers,
Stefan

Hi everyone,

Has anyone here performed ISO 9283 performance tests on the Annin Robotics AR4 robot, particularly for pose accuracy, pose repeatability, path accuracy, or path repeatability?

I’m currently planning an experimental evaluation of the AR4 according to ISO 9283 and would be interested to know if anyone has already conducted similar tests.

If so, I’d really appreciate hearing about your experience, especially regarding the measurement equipment, test setup, test positions/trajectories, and methodology you used.

I’m also interested in any papers, reports, datasets, videos, or other references involving ISO 9283 testing.

video for reference https://www.youtube.com/watch?v=hYueiQNrM_M&pp=ygUbaXNvOTI4MyBwZXJmb3JtYW5jZSB0ZXN0aW5n

Thanks!

Hey all - for the AR4 MKII and later models (where the stepper drives are mounted closely together in the J1 base), has anyone ever had any overheating issues with the fan off?

For my desk arm, I want to disable the fan just for noise purposes but don't want to damage the stepper drives.

For the production arms, I think leaving the fan running would be a good idea.

idea is to scale it down ,like 1/3 or even 1/4 ,for these 1/20 loader trucks

to have ar4 like hand for loading ,and to have ultra sonic sensors on hand to look for objects (just simple triger like open/sensor see it pick something - go pick ,closed /dont see it ,sit on 0 place)

is it possible to build it just with 3d prints ?

i mean construction only , not rest parts electronics screws bearings wires etc

Hello. Since StepperOnline motors are expensive, I wanted to ask if it's okay to use alternatives from AliExpress. Among Nema23 motors, is the link below okay? Can you also recommend other motors?

1.2N.m 네마 23 기어드 스테퍼 모터, 4, 5, 10, 20, 50:1 감속비 정밀 유성 기어박스 세트, 저소음, 고토크, CNC용 - AliExpress

This includes a complete Python API for the robot, new features for the virtual offline robot as well as a few minor bug fixes on the Modbus communication and kinematics.

Please checkout the new software and let me know if you have any questions or issues. Reach out to me anytime at info@anninrobotics.com

Hi everyone,

I’m building an AR4 MK4 and machining some of the parts myself. I’m also making a few custom changes to the original CAD, but the files don’t include the machining tolerances I need.

For those who have already machined the parts, what tolerances/fits did you use for bearing seats, shaft fits, pulley bores, and critical alignment faces?

Did you follow standard fits like H7/h6, H7/m5, etc., or did you define your own?

I’m new to mechatronics, so any advice or lessons learned would help a lot.

Thanks!

Hi all,

I am looking to continue supporting getting the ROS2 drivers into the ROS index. Currently only the docs are on the index, https://index.ros.org/r/ar4_ros_driver/#jazzy and https://github.com/ros/rosdistro/blob/master/jazzy/distribution.yaml#L546. I was wondering if it would be possible to update it to be the Annin-Robotics github src and then go from there.

There's also the second ros github, https://anninrobotics.com/ros2/, for isaac sim. I was thinking maybe they could all be combined? I would love to contribute, and help out with this effort!

I'd like to try using steppers or servos with closed loop drives. I'm wondering if the encoder feedback requirement for the AR4 Teensy code can be commented out and ignored? The object is to see if running smoother motors will achieve smoother movement for when using the robot with a moving image camera? Would this effect the repeatability/accuracy of the system in any way? (give that closed loop steppers and servos are very reliable already, and do not really need monitoring. I've used them in CNC and other motion control applications, never had to worry about losing position when motor size and gears are matched accordingly... Thanks

ROBODK AR4 Direct Control and Model Drawing using G-Code, ROBODK Home Return and Initial Position Control using Python

I want to issue commands using X, Y, Z, Rz, Ry, and Rx coordinates in other software, but I encounter issues with calculation formulas. Additionally, no other robots utilize the UR coordinate system for machining, and the generated code does not match the actual movements, making it difficult to use for machining.

It is difficult to use for machining that requires only precise movements. Currently, I am processing all movements using joint commands.