Deflection Compensation in Lightweight Robots Using a Digital Twin with Extended Dynamics as a Virtual Sensor

Lightweight, special-purpose robots are increasingly employed to automate tasks that standard industrial robots cannot handle for reasons of cost or suitability. However, their lightweight design makes them more prone to defection under operational loads, affecting tool-center-point (TCP) positioning accuracy. Although model-based compensation methods can mitigate these effects, they often require substantial engineering expertise. This paper presents a novel approach to improve the TCP positioning accuracy by implementing a virtual sensor that leverages digital twins with extended dynamics, in this case an elastic multibody system (EMBS). To reduce engineering effort, we reuse and expand the model from virtual commissioning (VC). We validate our method on a lightweight food-palletizing robot, deploying the digital twin on the control platform to actively compensate for TCP deflection. Its effectiveness is validated through external laser measurements.