Algorithm 1 DDPG-based Optimization Framework for MEC-enabled Blockchain IoT Systems.

1:for each GN $m\in {\mathsf{\Phi }}_G$ do 2:   Initialization: replay memory ${\mathcal{B}}_m$, critic network $Q(s,a|{\theta }_m^Q)$, actor network $\mu \left(s\right|{\theta }_m^{\mu })$ and corresponding target networks $Q^{\prime }$ and ${\mu }^{\prime }$ with weights ${\theta }_m^{{\mu }^{\prime }}\leftarrow {\theta }_m^{\mu }$ and ${\theta }_m^{Q^{\prime }}\leftarrow {\theta }_m^Q$; 3:end for 4:for each episode $\in \{1,2,\dots ,K_{max}\}$ do 5:   Initialization: state $s_{m,1}$ for each GN $m\in {\mathsf{\Phi }}_G$; 6:   for each decision epoch $n=1,2,\dots ,T_{max}$ do 7:     for each GN $m\in {\mathsf{\Phi }}_G$ do 8:        Select action $a_{m,n}=\mu \left(s_{m,n}\right|{\theta }_m^{\mu })+\Delta \mu$ based on the exploration noise $\Delta \mu$ to decide block interval and power allocation; 9:        Observe reward $r_{m,n}$ and next state $s_{m,n+1}$; 10:        Store transition data $(s_{m,n},a_{m,n},r_{m,n},$ $s_{m,n+1})$ into replay memory ${\mathcal{B}}_m$; 11:        Sample a mini-batch of Z transition tuples ${\left\{(s_z,a_z,r_z,s_z^{\prime })\right\}}_{z=1}^Z$ from memory ${\mathcal{B}}_m$ at random; 12:         Update critic network by minimizing the loss L: $$L=\frac{1}{Z}\sum _{z=1}^Z{\left(Q(s_z,a_z|{\theta }_m^Q)-{ϵ}_z\right)}^2;$$ 13:         Update actor policy based on the sampled policy gradient: $${\nabla }_{{\theta }_m^{\mu }}J\approx \frac{1}{Z}\sum _{z=1}^Z{\nabla }_{a}Q(s_z,a|{\theta }_m^Q){|}_{a=\mu \left(s_z\right)}{\nabla }_{{\theta }_m^{\mu }}\mu \left(s_z\right|{\theta }_m^{\mu });$$ 14:         Update target networks: $${\theta }_m^{{\mu }^{\prime }}\leftarrow \zeta {\theta }_m^{\mu }+(1-\zeta ){\theta }_m^{{\mu }^{\prime }}$$ $${\theta }_m^{Q^{\prime }}\leftarrow \zeta {\theta }_m^Q+(1-\zeta ){\theta }_m^{Q^{\prime }}$$ 15:     end for 16:   end for 17:end for