Adaptable bandwidth planning using reinforcement learning

Abstract

In order to improve the bandwidth allocation considering feedback of operational environment, adaptable bandwidth planning based on reinforcement learning is proposed. The approach is based on new constrained scheduling algorithms controlled by reinforcement learning techniques. Different constrained scheduling algorithms,, such as "conflict free scheduling with minimum duration", "partial displacement" and "pattern oriented scheduling" are defined and implemented. The scheduling algorithms are integrated into reinforcement learning strategies. These strategies include: - Q-learning for selection of optimal planning schedule using Q-values; - Informed Q-learning for exploitation and handling of prior-knowledge (patterns) of network behaviour; - Relational Q-learning for improving of bandwidth allocation policies dynamically in operational networks considering actual network performance data. Scenarios based on integration of the scheduling algorithms and reinforcement learning techniques in the experimental monitoring and bandwidth planning system called QORE (QoS and resource optimisation) are given. The proposed adaptable bandwidth planning is required for more efficient usage of network resources.

Keywords

• scheduling algorithm
• informed learning
• bandwidth planning
• reinforcement learning
• Q-learning