MARVEL: Enabling controller load balancing in software-defined networks with multi-agent reinforcement learning

Dynamic Controller Assignment in Software Defined Internet of Vehicles Through Multi-Agent Deep Reinforcement Learning

“…Most related work on controller assignment focuses on wired networks, including data centers [16], [17] and widearea networks [23]- [25]. Related work typically considers different metrics and goals, e.g., distance between forwarding devices and controllers [25], load balancing across controllers [24], and controllers' response delays [16], [17], [23]. More recently, some proposals have considered controller assignment in wireless mobile networks, where node mobility and frequent topology changes impose unique challenges to the controller assignment problem.…”

“…To solve for the dynamic controller assignment problem, different classes of algorithms can be considered, including: (1) Optimization algorithms, such as linear programming [18], [26] and integer quadratic programming [12]; (2) Heuristic algorithms, such as matching-based algorithms [16], [27]; 3Greedy algorithms, such as [28]; (4) Goal-oriented learning algorithms, such as DRL [24]. Although DRL has been used in the context of Intelligent Transport Systems [2] and SDNs for routing and resource allocation [29], it has not yet been explored to perform dynamic controller assignment in decentralized control plane architectures, especially in the context of IoVs.…”

5GhNet: an intelligent QoE aware RAT selection framework for 5G-enabled healthcare network

“…(iv) Software-defined wireless networking layer: This layer simply implements a distributed control plane modelled as multi-agent system to address the extensibility issue present in the single controller system (Sun et al 2020). Amidst the multiple controllers present in the SDWN layer, each controller manages a subset of RATs leading relays optimal network selection decision to the considered environment through southbound interface on the basis of real-time network status and patient demand.…”

Reinforcement Learning for Caching with Space-Time Popularity Dynamics

“…To capture the complicated interactions between caching decisions of parent and leaf nodes along with the space-time evolution of file requests, we develop a scalable deep reinforcement learning approach based on a hyper deep Q-network (DQN) implementation developed in our prior works [46,47]. Indeed, deep reinforcement learning has demonstrated state-of-the-art performance in diverse domains, including e.g., video games [48], data centers [49,50], smart grid [47], and software-defined networking [51,52]. The objective here is to endow the caching policy of the parent node with the capability of adapting itself to local policies of leaf nodes and space-time evolution of file requests.…”