Q-learning algorithm for navigation control of autonomous blimp

In this research, an autonomous control system for blimp navigation was developed using reinforcement learning algorithm. The aim of this research is to provide a blimp the capability to approach a goal position autonomously in an environment, where the dynamical models of the blimp and the environment are unknown. Webots™ Robotics Simulator (WRS) was used to simulate and evaluate the control strategy obtained through a one-step Q-learning method. The simulation data generated via WRS were then processed and analysed within MATLAB. The simulation results showed that the control policy acquired from Q-learning is much more effective compared to the traditional control methods.