A Self-Organizing Distributed Reinforcement Learning Algorithm to Achieve Fair Bandwidth Allocation for Priority-Based Bus Communication

Ziermann, Tobias; Mühleis, Nina; Wildermann, Stefan; Teich, Jürgen

doi:10.1109/isorcw.2010.18

Cited by 2 publications

(5 citation statements)

References 9 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…In the following section, we will present PLA first introduced in [16]. In the following section, we will present PLA first introduced in [16].…”

Section: Penalty Learning Algorithm (Pla)mentioning

confidence: 99%

“…Ziermann et al [16] propose to limit the sending probability to be between 0 and 1, because on first glance it does not make sense to send more than every time and less than never. First, the bus load of the last monitoring interval has to be calculated (line 1) by dividing the number of rounds in which a player has sent a message by the total number of rounds of the monitoring interval.…”

Section: Functional Principlementioning

confidence: 99%

“…For this reason, we chose to develop our own game-specific algorithm-the PLA. In the following section, we will present PLA first introduced in [16]. It is based on the enhanced priority-based medium access game.…”

Section: Penalty Learning Algorithm (Pla)mentioning

confidence: 99%

“…Ziermann et al . 16 propose to limit the sending probability to be between 0 and 1, because on first glance it does not make sense to send more than every time and less than never. However, experiments show that this restriction in some cases may lead to unfair convergence if ε is set too small.…”

Section: Penalty Learning Algorithm (Pla)mentioning

confidence: 99%

“…In order to test the PLA, we need to evaluate all five degrees of freedom of our simulation: starting probability of each player , free bandwidth ε, intervalLength , learning rate η and number of players n . These degrees have been extensively evaluated in 16. For the sake of completeness, we will outline the results.…”

Section: Penalty Learning Algorithm (Pla)mentioning

confidence: 99%

See 4 more Smart Citations

Distributed self‐organizing bandwidth allocation for priority‐based bus communication

Ziermann

Wildermann

Mühleis

et al. 2011

Concurrency and Computation

Self Cite

View full text Add to dashboard Cite

The raising complexity in distributed embedded systems makes it necessary that the communication of such systems organizes itself automatically. In this paper, we tackle the problem of sharing bandwidth on priority-based buses. Based on a game theoretic model, reinforcement learning algorithms are proposed that use simple local rules to establish bandwidth sharing. The algorithms require little computational effort and no additional communication. Extensive experiments show that the proposed algorithms establish the desired properties without global knowledge or a priori information. It is proven that communication nodes using these algorithms can co-exist with nodes using other scheduling techniques. Finally, we propose a procedure that helps to set the learning parameters according to the desired behavior. Therefore, we prefer to approach the goal of a working communication system from a different perspective. Starting with minimal constraints on the individual communication nodes, the nodes themselves should be able to establish communication. This should be done by using simple local rules with as less global information as possible. As the first step in this development process, we want to take a look at bandwidth-constrained communication, where the quality of the function varies with the used amount of bandwidth. In future work, these constraints will be extended to cover a larger field of applications.The remainder of the paper is organized as follows. In Section 2 the problem is defined and placed into the context-related work. In Section 3, the priority-based medium access game is formulated, which represents the fundamentals for the proposed methods. Section 4 presents the distributed reinforcement learning algorithm to establish fair bandwidth distribution. The parameters are analyzed and it is proven that the algorithm also works together with other scheduling methods. In Section 5, the period access scheme is introduced and analyzed, while Section 6 concludes our work and gives outlook on future work.

show abstract

“…In the following section, we will present PLA first introduced in [16]. In the following section, we will present PLA first introduced in [16].…”

Section: Penalty Learning Algorithm (Pla)mentioning

confidence: 99%

Section: Functional Principlementioning

confidence: 99%

Section: Penalty Learning Algorithm (Pla)mentioning

confidence: 99%

Section: Penalty Learning Algorithm (Pla)mentioning

confidence: 99%

Section: Penalty Learning Algorithm (Pla)mentioning

confidence: 99%