Simulation-Based Evaluation of Adaptive Safety Zones for Human-Robot Interaction Driven by Deep Reinforcement Learning and LSTM Konferenzpaper