Abstract
—The coherent joint transmission (CJT) serving mode, which can achieve full coherent gain among the cooperative access points, has been considered as an optimal solution for cell-free massive MIMO systems under ideal assumptions. However, considering the practical deployment constraints, for instance limited fronthaul, it has been shown that non-coherent joint transmission (NCJT) is likely to be superior to CJT, due to the fact that NCJT has much lower fronthaul requirement. In other words, any single transmission mode (CJT or NCJT) cannot adapt the capacity limited fronthaul perfectly. To further explore the performance potential under limited fronthaul constraint by harnessing the merits of CJT and NCJT simultaneously, we propose a CJT-NCJT hybrid serving mode architecture, under which each user equipment (UE) is allowed to operate in either CJT or NCJT serving mode. The key challenge in the hybrid serving mode is to design a fronthaul-capacity-aware serving mode allocation algorithm obeying the newly established signal model. Specifically, we formulate a mixed-integer programming problem to design a low-complexity matching theory-assisted UE serving mode allocation algorithm. Furthermore, we develop a successive convex approximation-based power allocation algorithm to maximize the system's sum-rate with given serving mode. Index Terms—Cell-free massive MIMO, coherent joint transmission , non-coherent joint transmission, limited fronthaul, hybrid serving mode.