07-16-2026, 02:58 AM (This post was last modified: 07-24-2026, 04:10 PM by Albert.)
I have seen that for UR robots, you can open a URP using UR_LoadControllerKinematics and ticking "use controller kinematics". This allows the unique kinematic values of each UR robot to accurately simulate the robot.
Could a similar feature please be added for Techman robots, so the Techman factory calibration is reflected in RoboDK and I can program offline more accurately?
We tried adding this feature for Techman robots a few years ago but it is not straightforward. We noticed that TM robots have a non linear coupling for the wrist axes. For example, axes 4,5 and/or 6 do not have a constant gear ratio. This makes it very complicated for us to model.
Do you know if this changed? If the nominal model is a pure DH or DHM table we should be able to model it in RoboDK without issues.
I have been talking to Techman and they have given me what they say is required to implement the correct forward kinematics, therefore obtaining an accurate simulation.
The details are as follows:
"However, the kinematic model used inside the controller is not only the nominal DH model.
During factory calibration, each robot is individually measured and calibrated. The controller internally applies calibration parameters to compensate for manufacturing tolerances, including:
Joint angle offsets (Δθ)
Link length errors (Δa)
Link offset errors (Δd)
Link twist errors (Δα)
Additional axis orientation corrections (Δβ)
These compensation parameters improve the robot's absolute positioning accuracy while maintaining the same repeatability. Similar modified kinematic models with an additional β rotation are commonly used in industrial robot calibration to better represent slight axis misalignment.
At each joint, the controller applies the calibrated geometric parameters before calculating the transformation to the next link.
This is also why an offline simulation software (such as RoboDK) using only the nominal robot model may produce a different TCP pose even when the joint angles are identical.
The detailed implementation of the controller's internal kinematic algorithm and calibration process is proprietary to Techman Robot. To perform the forward kinematics calculation, you simply need the DH and Delta-DH parameters (which has been shared at July 16th )"
And an image was attached of the DH transformation matrix.
Is this enough to prove it would be worth adding a factory calibration option for Techman robots? I reckon the process of calibration could look like this:
0. User has TM robot in station
1. Connect to robot -> run "get kinematic signature" command (get Δθ, Δa, Δd, Δα, Δβ)
2. Robot parameters panel -> tick "use calibrated model" or select the new kinematic solver
Or what you guys think is best, of course :)
This information from Techman seems a little simple, hardly accounting for non-constant gear ratios, so let me know if Techman perhaps hasn't given me all the parts of the puzzle and I can ask them more questions. We buy lots of robots from them so they seem to reply a bit quicker! :P
I probably won't make custom kinematics for the TM7S as it wouldn't be the most cost-effective use of company time. Right now I am using RoboDK to make the path using the Techman_PathNode postprocessor and importing it into my TMFlow program, which is a fine workaround.
However, I would really encourage you to work with Techman to fully support their robots - I think it could be a great partnership and they already have RoboDK in their catalog as "plug 'n' play" compatible (see attached).
I was really disappointed when I wasn't able to use RoboDK's amazing features with my Techman robot - especially the incredible new collision-free path planner! Please take my most sincere recommendation to get in touch with Techman and organise a joint effort to support their robots in RoboDK - maybe their software engineers have a simple implementation ready to use.
Please let me know if this happens - for then I can more confidently recommend RoboDK to every one of my employers, making RoboDK bucketloads in sales!
Thank you for your feedback Luke! We appreciate your comments.
We would love to integrate proper kinematics for TM but this is challenging as we have put reasonable efforts and we got close enough to the point were we were accurate to the order of 100 microns (numerically speaking, we should be at 0 though). The rotation around Y was puzzling.
We are accounting for this feedback and we'll add it to our pipeline of improvements, probably as a custom DLL. This may take some time though.
Would it be possible for you to share real data from a real TM controller and their corresponding model? Ideally a few joint values with their corresponding Cartesian values (flange with respect to the robot base). Or put us in touch with someone at TM who can help. It is better if you contact us by email via info@robodk.com and mention this thread.
