Energy-Efficient Joint AP Selection and Power Control in Cell-Free Massive MIMO Systems: A Hybrid Action Space-DRL Approach

In this letter, we investigate the problem of joint access point (AP) selection and power control in Cell-free massive MIMO (CF-mMIMO) networks, aiming at maximizing energy efficiency (EE) under a sum spectral efficiency (SSE) requirement. Given the limitations of separate optimizations due to strong couplings between AP selection and power control, a novel deep reinforcement learning (DRL) approach with hybrid action space is proposed. Simulation results show that our proposed algorithm achieves low-dimensional actions by embedding parameters and can achieve higher EE with lower complexity.