Joint power allocation and blocklength assignment for reliability optimization in CA-enabled HetNets

Heterogeneous cellular networks (HetNets) are widely recognized as representing the future development trend of networks and provide architectural support for the emergence of many new applications. For applications that are sensitive to reliability, carrier aggregation (CA) plays a crucial role in optimizing system performance by intelligently combining and aggregating idle resources. In this paper, we aim to address the resource allocation problem in order to optimize reliability in CA-enabled HetNets. Traditional optimization methods often use much computing time, whereas reinforcement learning (RL) techniques can produce real-time responses. We employ an innovative dual Markov decision model (MDP) to describe the relationship between the two agents. Then, an iterative framework is proposed to develop the power allocation and blocklength assignment strategy. With this framework, we solve the problem of action space explosion and learning efficiency reduction that traditional RL always encounters. The simulation results demonstrate the effectiveness of our method in solving resource allocation problems, and the framework shows excellent convergence for the three algorithms we implement.