Algorithm 2 AO-Based Two-layer-Three-Stages Optimization

1: Initialize: Convergence tolerance ε, iteration index j = 0, initial energy efficiency ${\rho }^{\left(0\right)}=0$. 2: Initialize: Feasible task scheduling S, power allocation P, and trajectory L. 3: repeat                                                                                                                   $▷$ Outer Loop 4. Set inner iteration index i = 0 5: Initialize inner variables: $\left\{{\mathbf{S}}^{\left(0\right)}, {\mathbf{P}}^{\left(0\right)}, {\mathbf{L}}^{\left(0\right)}\right\}⟵\{\mathbf{S},\mathbf{P},\mathbf{L}\}$ 6: while not converged do                                                             $▷$ Inner Loop: BCD Method 7: i $⟵$ i + 1 8: Fix ${\mathbf{P}}^{(i-1)}, {\mathbf{L}}^{(i-1)}$ and ${\rho }^{\left(j\right)}$, solve (P2.1) to obtain ${\mathbf{S}}^{\left(i\right)}$ 9: Fix ${\mathbf{S}}^i$, ${\mathbf{L}}^{(i-1)}$ and ${\rho }^{\left(j\right)}$, solve (P2.2) to obtain ${\mathbf{P}}^{\left(i\right)}$ 10: Fix ${\mathbf{S}}^i$, ${\mathbf{P}}^{\left(i\right)}$ and ${\rho }^{\left(j\right)}$, solve (P2.3) to obtain ${\mathbf{L}}^{\left(i\right)}$ 11: end while 12: Update variables: $\left\{\mathbf{S},\mathbf{P},\mathbf{L}\right\}⟵\left\{{\mathbf{S}}^{\left(i\right)}, {\mathbf{P}}^{\left(i\right)}, {\mathbf{L}}^{\left(i\right)}\right\}$ 13: Update energy efficiency: $${\eta }_{SEE}= {\displaystyle \frac{{\sum }_{n=1}^N{\sum }_{k=1}^K{\omega }_k\left[n\right]R_{k,rad}\left[n\right]}{E_t+E_f}}$$ 14: j ⟵ j + 1 15: until $\left|{\rho }^{\left(j\right)}- {\rho }^{\left(j-1\right)}\right| \le \epsilon$ 16: Output: Optimal solution ${\mathbf{S}}^{\ast }, {\mathbf{P}}^{\ast }, {\mathbf{L}}^{\ast }$ and max EE.