Comparing The Effects Of Latency And Jitter in Distributed Furuta Pendulum Control : Latency and Jitter Effects in Distributed Furuta Pendulum Control

Sammanfattning

Industrial control systems are increasingly distributed across communication networks, which introducesdelays into the feedback loop and can degrade or destabilise fast, unstable plants. This is especially criticalfor systems such as the Furuta pendulum, where successful control depends on rapid and accurate feedback.Understanding how different delay-handling strategies perform under such conditions is thereforeimportant for the design of distributed control systems.This thesis investigates how three control strategies handle delayed feedback on a Furuta pendulum: abaseline linear quadratic regulator (LQR), LQR combined with a Kalman filter (KF), and LQR combinedwith a Smith predictor (SP). The strategies were evaluated experimentally under controlled constant latencyand jitter conditions using the same underlying LQR gain, so that differences in behaviour reflected thedelay-handling mechanism rather than controller tuning. All three strategies were tested over nominaldelays from 0 to 20 ms with and without ±5 ms jitter, using five runs per condition.The baseline LQR failed at every added delay level and had a delay margin of 0 ms. The KF improveddelay tolerance and remained stable up to 10 ms under both constant latency and jitter, while also reducingarm-angle motion at 0 ms relative to the baseline. The SP extended the stable range further and remainedstable up to 15 ms under both constant latency and ±5 ms jitter, although it failed at 20 ms and showed asmall residual pendulum-angle offset at zero delay attributable to model-plant mismatch in the predictor.These results show a clear ranking in delay tolerance, with the SP providing the strongest robustness underthe delay conditions examined in this study.