Event-triggered scheduling for stochastic multi-loop networked control systems with packet dropouts

2016 IEEE 55th Conference on Decision and Control (CDC)

Kneissl

Hirche

2016

Self Cite

Abstract-In practice often multiple control applications share a communication channel, requiring a smart and scalable scheduling mechanism to coordinate the access to the capacitylimited communication medium. In this paper, we propose a decentralized event-triggered medium access control (MAC) for multiple feedback control loops which are coupled through a capacity-limited communication medium. The individual control loops are assumed to be linear time-invariant (LTI) with stochastic heterogeneous plants. Noisy state measurements from local sensors are transmitted through a shared communication medium to their respective control units. Due to capacity limitations in the shared communication channel, not all sensors are allowed to transmit simultaneously. To allocate the scarce resources, a decentralized MAC which prioritizes the channel access according to a real-time error-dependent measure, is introduced. The prioritization is orchestrated via a combined deterministic and probabilistic mechanism aiming at the efficient allocation of the limited capacity. We study stability of the described multi-loop NCS under the proposed MAC design in terms of Lyapunov stability in probability (LSP). It is demonstrated that the collision rate remains low by properly tuning the MAC parameters. Numerical results show that the proposed MAC design significantly outperforms conventional time-triggered and random access schemes, while its performance closely follows the centralized TOD approach.

Section: Numerical Resultsmentioning

confidence: 99%

“…Stochastic prioritybased protocols [14]- [17] are shown to improve scalability and may also cope with communication unreliability. However, the designs are either centralized [14], or rely on having access to perfect state information, i.e. they assume noiseless sensors measure state values perfectly at every time-step [8].…”

Section: Introductionmentioning

confidence: 99%

Decentralized event-triggered medium access control for networked control systems

2016 IEEE 55th Conference on Decision and Control (CDC)

Kneissl

Hirche

2016

Self Cite

“…Each node is characterized by only one of the above cases during transition from k to k+N −1. We apply Theorem 1 to the cases c 1 − c 3 employing the drift (13) and Lyapunov function (12). Remember that we consider local error of each node.…”

Section: Stability Analysismentioning

confidence: 99%

Event-based data scheduling for a class of interconnected networked control systems

2015 54th IEEE Conference on Decision and Control (CDC)

Deroo

Hirche

2015

Self Cite

Abstract-In this paper we address the problem of eventbased data scheduling for a class of physically interconnected networked control systems which compete for limited communication resources. The overall interconnected system consists of multiple heterogeneous LTI sub-systems with the physical interconnection being modeled by a directed acyclic graph (DAG). The sub-systems are controlled by a networked controller through a shared communication channel. In order to cope with the limited channel capacity, we introduce a bi-character deterministic-probabilistic scheduling mechanism which dynamically assigns access priorities to each sub-system at each time-step according to an error-dependent priority measure. The sub-systems which are granted channel access then transmit their state information through the communication network. Given an emulation-based control strategy, we prove stability of such networked systems under the proposed scheduling law in terms of f -ergodicity of overall network-induced error. Simulation results illustrate the proposed approach and show a reduction in the error variance compared to standard TDMA and uniform random-access scheduling policies.

“…We assume that the transmitted packets are not lost due to any external cause (see [17] where the problem of data packet loss in a similar NCS setup is addressed). Hence, if data packets do not collide in the channel, they will certainly be received by their corresponding control units.…”

Section: Problem Statementmentioning

confidence: 99%

Prioritized contention resolution for random access networked control systems

2017 IEEE 56th Annual Conference on Decision and Control (CDC)

Vilgelm

Kellerer

et al. 2017

Self Cite

Abstract-Decentralized scheduling schemes offer low-cost, easy-to-install, and scalable resource allocation mechanisms suitable for networked control systems (NCS) with a large number of loops. However, contentions are unavoidable as collisions randomly occur in the communication link which lowers channel throughput and consequently compromises control performance. In this paper, we introduce a novel deterministic state-dependent prioritization policy to resolve the contention in a multi-loop NCS with random access medium. The proposed prioritization scheme can be realized within most of the shared medium technologies, e.g. ad-hoc and bus networks, aiming to decrease the collision probability. The priority for an individual control loop depends on the current local network-induced errors with a larger error leading to higher priority. We show stability of the described NCS under the proposed collision reduction mechanism according to stochastic Lyapunovbased techniques. It is demonstrated that the proposed statedependent contention resolution is considerably effective in lowering the collision rate, though incapable of eliminating it due to the decentralized nature of the medium access control.