Source author record

Florian Jaumann

Florian Jaumann appears in the imported research catalog. Authorship, coauthor and topic links are available while profile ownership is still unclaimed.

ResearcherUnclaimed source record

Catalog footprint

What is connected

1works
2topics
2close collaborators

Actions

Connect this record

Log in to claim

Research graph

See the researcher in context

Open full explorer

Inspect adjacent papers, topics, institutions and collaborators without losing the researcher page.

Building this map preview

BZPEER is loading the nearby papers, people, topics and institutions for this page.

Published work

1 published item(s)

preprint2020arXiv

Saving Tokens in Rollout Control with Token Bucket Specification

We consider a communication network over which transmissions must fulfill the so-called token bucket traffic specification, with a rollout (i.e., predictive) controller that both schedules transmissions and computes the corresponding control values. In the token bucket specification, a transmission is allowed if the current level of tokens is above a certain threshold. Recently, it has been shown that having a full bucket at the time of a set point change significantly improves the control performance as compared to when the bucket level is low. In this work, we develop mechanisms that guarantee that the bucket fills up after the controlled plant has converged to a set point. To do this, we consider two different setups. First, we consider that all transmissions over the network must fulfill the token bucket specification and show convergence to the upper sector of the bucket by adding a slight terminal cost on the bucket level. Afterwards, we consider a modified network which additionally features a direct link over which transmissions need not fulfill the token bucket specification. In this setup, we prove convergence of the bucket level exactly to the upper rim. These mechanisms enable a similar level of flexibility as event-triggered control: In converged state, little communication is used while in precarious operating conditions, a burst of transmissions is possible. Other than event-triggered approaches, the proposed methods allow to specify the network traffic beforehand by means of the token bucket. Lastly, we validate the proposed approaches in a numerical example.