TUM's delay-aware PMSM control cuts sampling needs while keeping bandwidth guarantees

Modern robotic joints require high-bandwidth current control across wide speed, acceleration, and torque ranges, but conventional PMSM controllers are designed in continuous time and then discretized. That forces engineers to heuristically pick sampling frequencies and validate iteratively, without explicitly accounting for communication and computation delays. Researchers at TU Munich introduce a task-aware, delay-extended discrete-time joint model that explicitly incorporates those delays. This enables direct synthesis of a discrete PI current controller with prescribed bandwidth and delay guarantees throughout the operating envelope. The framework analytically computes the minimum required sampling frequency, controller gains, and DC-link voltage needed to satisfy motor and joint performance specs — removing guesswork from the design process.

Simulations across a range of dynamic requirements show that this delay-aware approach substantially reduces sampling-frequency and DC-link-voltage requirements compared to standard continuous-time-based design. The team also built a custom robotic joint and validated the method under real embedded implementation conditions, confirming the theoretical guarantees. By tying controller design directly to hardware constraints, this work makes high-bandwidth joint control more predictable and efficient. Robot manufacturers can use it to shrink power electronics, lower cost, and speed up development of next-generation actuators.