Reinforcement Learning Agent Design and Optimization with Bandwidth Allocation Model

  • 2022-11-23 14:48:25
  • Rafael F. Reale, Joberto S. B. Martins
  • 0


Reinforcement learning (RL) is currently used in various real-lifeapplications. RL-based solutions have the potential to generically addressproblems, including the ones that are difficult to solve with heuristics andmeta-heuristics and, in addition, the set of problems and issues where someintelligent or cognitive approach is required. However, reinforcement learningagents require a not straightforward design and have important design issues.RL agent design issues include the target problem modeling, state-spaceexplosion, the training process, and agent efficiency. Research currentlyaddresses these issues aiming to foster RL dissemination. A BAM model, insummary, allocates and shares resources with users. There are three basic BAMmodels and several hybrids that differ in how they allocate and share resourcesamong users. This paper addresses the issue of an RL agent design andefficiency. The RL agent's objective is to allocate and share resources amongusers. The paper investigates how a BAM model can contribute to the RL agentdesign and efficiency. The AllocTC-Sharing (ATCS) model is analyticallydescribed and simulated to evaluate how it mimics the RL agent operation andhow the ATCS can offload computational tasks from the RL agent. The essentialargument researched is whether algorithms integrated with the RL agent designand operation have the potential to facilitate agent design and optimize itsexecution. The ATCS analytical model and simulation presented demonstrate thata BAM model offloads agent tasks and assists the agent's design andoptimization.


Quick Read (beta)

loading the full paper ...