Robot Obstacle Avoidance Controller Based on Deep Reinforcement Learning

Figure 8

Rewards curve for a model trained at 37 Hz for 5,600 rounds and validated 100 times at 90 Hz and 37 Hz.