“…In the sequel, we present how to update real-time primal solution such that the computation and signaling delay δT f can be substantially decreased. In practice, the BS is able to acquire the channel gain {g s,r n (ℓ), g s,d n (ℓ), g r,d n (ℓ)} N n=1 through channel prediction before Frame ℓ starts, if the wireless channel varies slowly across the frames [36]. Given the channel gain {g s,r n (ℓ), g s,d n (ℓ), g r,d n (ℓ)}, the BS can compute the ratios P (1) s,n (ω ℓ ) θ(1) m (ω ℓ ) , P (2) s,n (ω ℓ ) θ( 2 Once the source and relay nodes obtain the sensing outcomes x m (ℓ) and y m (ℓ), they can compute θ(1) m (ω ℓ ) and θ(2) m (ω ℓ ) according to the closed-form solutions (30), (32), (50), and (51).…”