Algorithm 1 Offline Training for Domain Theory

Require: $\mathcal{S}$, $\mathcal{A}$: State and action spaces in the continuous domain Require: $map$: A map used for navigation tasks in the continuous domain  1:Initializing DRL (deep reinforcement learning) parameters $\gamma$  2:Initializing the reward function $r(s,a)$  3:Initializing training iterations T  4:for t = 1 to T do  5:   Randomly generating goal g in the $map$  6:   $\forall a\forall s\ne g,r(s,a)\leftarrow 0$  7:   $\forall a,s=g,r(g,a)\leftarrow C$  8:   $Q_{{\pi }_g}\leftarrow DRL(S,A,\gamma )$  9:end for 10:return ${\mathbb{T}}_{\pi }\left(\mathcal{G}\right)$