查看文章

In this work, beamforming design and resource allocation for overload non-orthogonal multiple access (NOMA) systems is investigated. Based on this, a pure-NOMA framework is proposed to benefit from the non-orthogonal resources. In addition, an energy efficiency (EE) maximizing problem is formulated by jointly designing the beamforming, user grouping, as well as power allocation of users. Reinforcement learning (RL) has been shown to be very effective in tackling this kind of joint optimization in wireless communication networks. However, the high dimensionality and coupling non-convex mixed integer nonlinear programming (MINLP) problem makes conventional RL methods hard to get an ideal reward in practice. Given this challenge, a curiosity-driven solution is proposed to meet the MINLP challenge. Numerical results indicate that: 1) the pure-NOMA framework provides extra room for DoF and capacity …