Time in general-purpose control systems: the Control Time Protocol and an experimental evaluation

TLDR

Distributed control applications require accurate timing, yet no two clocks generally agree, necessitating message passing to interpret distributed clocks. The authors present the Control Time Protocol, an architecturally clean algorithm for providing timing information in distributed control. The protocol was implemented on a testbed in the Convergence Laboratory at the University of Illinois, demonstrating its practical feasibility. The study reveals a fundamental indeterminacy in time estimation with asymmetric delays, shows that the protocol adds no extra dependencies and improves system reliability, and presents experimental results demonstrating enhanced stability, performance, and measured network latencies.

Abstract

No two clocks generally agree. Distributed control applications, however, require accurate timing information. This necessitates some form of message passing for interpretation of distributed clocks. We exhibit a fundamental indeterminacy in estimating time when delays are unequal in the two directions, necessitating an assumption of symmetry. We present the Control Time Protocol as an architecturally clean algorithm for providing timing information in the domain of distributed control. The protocol introduces no additional dependencies beyond those already present in the very control loops, thus enhancing the reliability of the systems. It has been implemented on the testbed in the Convergence Laboratory at the University of Illinois. Experimental results demonstrating how time stamping can be used to enhance stability and performance, as well as measurements of network latencies, are provided.