Hi everyone,
I am working on real-time detection of IRB 14050 arm-angle discontinuities and redundancy-resolution inconsistency using ABB Robot Web Services (RWS).
Why this specific topic?
For IRB 14050 / native 7-axis robots, arm angle is not just a theoretical quantity — it is part of how a robtarget is fully specified, and controller settings such as Arm-Angle Reference Direction can influence singularity behavior and motion consistency. That makes arm-angle continuity a practical monitoring target, not just a geometric curiosity.
>Figure. Replay benchmark of a read-only RWS-style probe for an IRB 14050-like 7-axis stream. The estimated TCP step remains nearly constant, while the NARH continuity score rises sharply within the same time window. Here, NARH (Non-Associative Residual Hypothesis) is used as an order-sensitive joint-space continuity metric to expose hidden redundancy-state discontinuities that may not be obvious from Cartesian monitoring alone. Replay benchmark only; not yet plant data.
I built a small Python audit probe called SIPA (Simulation Integrity & Physics Auditor).
It includes a NARH-based continuity metric, where NARH stands for Non-Associative Residual Hypothesis. In practical terms, I use it here as an order-sensitive joint-space continuity diagnostic for exposing hidden redundancy-state discontinuities that may remain invisible in Cartesian monitoring.
The current scope is intentionally conservative:
- read-only
- alarm-only
- no controller intervention
- RWS JSON input
RWS is a natural starting point because it allows non-intrusive controller-side observation before discussing anything lower-level.
Current replay benchmark
I now have a minimal RWS-style replay benchmark for an IRB 14050 / 7-axis stream. It demonstrates:
- warm-up handling
- joint-space associator / NARH score
- estimated TCP step tracking
- alarm-only behavior
Example output:
[RWS-DEBUG] t= 1.213s assoc= 0.000 tcp_step_mm= 1.304 alarms=none
[RWS-DEBUG] t= 1.414s assoc= 0.429 tcp_step_mm= 1.304 alarms=none
# CRITICAL MOMENT: Capturing Arm-Angle Discontinuity
[RWS-DEBUG] t= 2.243s assoc= 3090.785 tcp_step_mm= 1.302 alarms=ALERT:associator_peak=3090.785
[RWS-DEBUG] t= 2.444s assoc= 11224.296 tcp_step_mm= 1.302 alarms=ALERT:associator_peak=11224.296
In this benchmark, the estimated TCP step remains small and smooth, while the joint-space associator spikes sharply during a synthetic redundant-axis discontinuity.
So the point is not “the TCP looks bad.”
The point is: the TCP may still look acceptable while the redundancy state has already become unstable or discontinuous.
Important note: the 11,000+ value is not a joint-velocity error and not a TCP tracking error. It is a joint-space continuity / instability indicator intended to expose hidden redundancy-state discontinuities.
This replay benchmark is a diagnostic demonstration, not yet a claim about production controller behavior under real plant data.
Repository / benchmark / code / NARH Theory : https://github.com/ZC502/SIPA.git
What I am looking for
To move from replay benchmark to controller-grade validation, I would be very interested in:
-
Real IRB 14050 / native 7-axis RWS traces, especially for:
• high-speed arcs
• tight-space avoidance
• singularity-neighborhood motion
• redundancy-heavy assembly moves -
Discussion with ABB / RobotWare / EGM users
My current probe is RWS-only and read-only.
Future work would be to evaluate whether the same continuity metric could also be useful in EGM-related workflows.
My goal here is not to push another plugin, but to ask a technical question:
Can arm-angle continuity be monitored early enough to flag redundancy-switch risk before it becomes a shop-floor problem?
If anyone has relevant RWS traces, IRB 14050 / YuMi experience, or EGM experience, I would be very interested in discussing it.