Hands-On ROS for Robotics Programming

Testing the trained model

Taking into account that we are using the Deep Q-Network (DQN) algorithm, what we have to save from the training process is the structure of the neural network and the weights of the edges. This is what the gopigo3_dqn_stage_2 node performs every 10 episodes. Hence, you can find the saved models inside the ./gopigo3_dqn/save_model folder, and the weights of the network are stored in the h5 file type. Every file contains in its name the scenario (stage_1 or stage_2) and the episode number. Follow these steps to evaluate the trained model:

1. Select the file with the highest episode number, that is, stage_2_1020.h5.

Every h5 file contains the weights of the DQN network as it appears at the end of the episode referenced ...