Energy-Efficient Base Station Deployment in HetNet Based on Traffic Load Distribution

Link Quality-Guaranteed Minimum-Cost Millimeter-Wave Base Station Deployment

“…However, the formulated INP is difficult to be solved directly. While this could motivate us to pursue suboptimal algorithms (e.g., [24]- [26], [32], [33], [35]- [38]), we instead show that the formulated INP can be decomposed into two separable subproblems, i.e., (i) cell coverage optimization problem and (ii) minimum subset BS selection problem, and optimally solved by sequentially solving these two subproblems. In particular, we show that the second subproblem (minimum subset BS selection problem), which is also INP, can be transformed into a linear equivalent form that can be efficiently solved via existing software.…”

“…Though these subproblem frameworks were largely benefited from the deterministic link models at sub-6GHz bands, such models and associated problem formulations cannot be extended to the mmWave. Although prior approaches [24]- [26], [35]- [38] deal with INP, they are different from our proposed approach in terms of the objective functions, associated constraints, and thereby, the developed algorithms.…”

“…About existing work, this aspect has some similarities to the well-investigated problems on optimizing BS sleeping and user association at sub-6GHz bands. The latter problems were often addressed by formulating INP with the objectives of maximizing network throughput [35], [36] or minimizing the power consumption [24]- [26], [37], [38]. Because large-scale INP is NP-hard and is generally very difficult to be optimally solved, devising suboptimal but efficient algorithms was the focus of these approaches, by using greedy heuristics [24]- [26], Lagrangian dual [36], [38], and sequential subproblem formulations [35], [37].…”

“…The latter problems were often addressed by formulating INP with the objectives of maximizing network throughput [35], [36] or minimizing the power consumption [24]- [26], [37], [38]. Because large-scale INP is NP-hard and is generally very difficult to be optimally solved, devising suboptimal but efficient algorithms was the focus of these approaches, by using greedy heuristics [24]- [26], Lagrangian dual [36], [38], and sequential subproblem formulations [35], [37]. Though these subproblem frameworks were largely benefited from the deterministic link models at sub-6GHz bands, such models and associated problem formulations cannot be extended to the mmWave.…”

“…More specifically, directly searching for the optimal solution needs to evaluate all 2 B×(G+1) combinations of (y, X), which is prohibitive for relatively large B and G (e.g., B ≥ 100 and G ≥ 100). In the next subsection, we address this challenge and propose a low-complexity approach to the problem (24), while optimally solving it.…”