Tactile Cyber-Physical Systems: A Testbed, A Performance Metric and Network Resource Allocation Protocols

Abstract

Tactile Cyber-Physical Systems (TCPS) use haptic, audio and video modalities to facilitate humans to perform real-time physical interaction of remote objects over ultra-low latency and ultra-reliable networks referred to as the Tactile Internet. TCPS has applications in automation, education, entertainment, gaming, healthcare, and industrial transportation. TCPS requires ultra-low latency of a few milliseconds and ultra-reliability of 99.99999% to prevent control-loop instabilities and operator-side cybersickness. The thesis focuses on design and development of a testbed, a performance metric and network resource allocation protocols for TCPS.

We first design and develop TCPSbed, a modular testbed for TCPS. TCPSbed facilitates integrating different components, both real and simulated, to realize different TCPS applications and evaluate their latency and stability. TCPSbed supports edge intelligence modules that predict command and feedback signals at the operator and teleoperator ends, allowing TCPS applications to perform well in adverse network conditions.

Our second work deals with design of an evaluation method and a metric, called QoC, to characterize TCPS. For characterization, we adopt step response analysis, a well-known control-theoretic method. The adoption entails replacing the TCPS operator with a controller with known characteristics and analyzing its response to slave-side step disturbance. From the step response curves, we derive QoC, which captures the characteristics of both networking and non-networking components in a TCPS.

We then focus on network slicing protocols for TCPS. In TCPS, the requirement of network resources, such as latency and bandwidth, varies over time. We propose a mechanism to dynamically create, destroy and switch network slices on a per flow basis depending on instantaneous resource requirements of the flows. Our solution consists of two main components: (a) a clustering algorithm to determine the slices and their specifications to support TCPS flows (b) on-the-fly provisioning and switching of these slices using Software-Defined Networking and P4-programmable switches.

We finally develop decentralized dynamic switch configuration protocols for IEEE 802.1 Time Sensitive Networks (TSN) to support TCPS. Unlike traditional best-effort networks, TSN can guarantee low packet latencies, jitter, and loss for time-critical flows by isolating them from external traffic. Our protocols supports plug-and-play operation of TCPS terminals with guaranteed minimal packet latencies in the presence of external traffic. We develop PYTSN, an open-source discrete-event network simulator containing models of TSN network components to evaluate the proposed protocols.