“…Note that δ = 1 indicates no PSL constraint and the optimization relies soley on the CRLB (this yields d δ=1 = [0, 0.5, 1, 9, 9.5, 10] for M = 6). In addition, a ULA setup (e.g., a linear array with d ula = [0, 1,2,3,4,5] for M = 6) and a best-scenario CRLB (e.g., the CRLB of d δ=1 for M = 6) are also used as benchmarks. We assume a coarse DOA estimate is available (e.g., from current connected subarray, or from IMUs and GPS), and the maximum DOA estimation error ∆u is chosen as 0.1 for generating û.…”