edge Exploration of dangerous environments by reinforcement learning agents combining the reward-maximization task with safety concerns