Hands-On ROS for Robotics Programming

Summary

This chapter has been a quick and practical introduction to how you can apply reinforcement learning so that a robot can perform useful tasks such as transporting materials to a target location. You should be aware that this kind of machine learning technique is at the very beginning of its maturity, and there are as yet few practical solutions working in the real world. The reason is that the process of training is very expensive in terms of time and cost, since you have to perform thousands of episodes to get a well-trained model, and later replay the process with the physical robot to address behavioral differences between the real world and the simulated environment.

Be aware that the training process in Gazebo is not a substitute ...