5 Nidle_worker = num_idle_worker() 6 while Nidle_worker > 0 do /* Create new individual */ 7 new_individual = population.sample(πμ, Σ) 8 Create new individual by sampling from N (πμ,Σ) 9 actor_worker = get_idle_worker() 10 actor_worker.set_actor_network(new_individual) 11 Calculate prl according to Eq. (3) 12 if rand() < prl and total_step >= rl_start_step then /* Train the actor with shared critic */