A packet-based dual-rate PID control strategy for a slow-rate sensing Networked Control System

TL;DR

This paper proposes a packet-based dual-rate PID control strategy for networked control systems that effectively manages delays, dropouts, and disorder, while enabling energy-efficient slow-rate sensing without sacrificing control performance.

Contribution

It introduces a dual-rate PID control approach with remote prediction and local fast-rate control to maintain performance under network uncertainties.

Findings

Achieves nominal control performance despite network delays and dropouts.

Outperforms previous gain scheduling strategies in real-time robot control.

Enables energy savings through slow-rate sensing without degrading control quality.

Abstract

This paper introduces a packet-based dual-rate control strategy to face time-varying network-induced delays, packet dropouts and packet disorder in a Networked Control System. Slow-rate sensing enables to achieve energy saving by reducing network load. In addition, choosing a slower sensing period than the longest round-trip time delay can avoid packet disorder. On the other hand, a slow-rate sensing usually degrades control performance in a conventional control framework. Therefore, including dual-rate control techniques can be useful to maintain the desired performance, since the controller is able to generate a fast-rate control signal from a slow-rate sensing signal. A dual-rate PID controller is used, which can be split into two parts: a slow-rate PI controller is located at the remote side (with no permanent communication to the plant) and a fast-rate PD controller, at the local side (close to the plant, sensor, and inside the actuator, which can offer a low computation power). In addition, at the remote side, where a powerful computation device is located, a prediction stage is included in order to generate the packet of future, estimated slow-rate control actions to be sent to the local side. At this side, these actions are converted to fast-rate ones and used when a packet does not arrive due to the network-induced delay or due to occurring dropouts. The control proposal is able to reach the nominal (no-dropout, no-delay) performance despite the existence of time-varying delays and packet dropouts. Via real-time control for a Cartesian robot, results clearly reveal the superiority of the control approach compared to a previous authors\' proposal, where the time-varying delays are faced by means of a gain scheduling control strategy.