A Machine Learning Approach Using MATLAB on Robotics Control and Navigation for Autonomous Systems

Smart autonomous robotics must be able to navigate in complicated surroundings. Developing complex navigational platforms to transport autonomous robotics from one location to another has taken years of engineering and study. Notwithstanding their general effectiveness, a new area of study is focused on creating machine learning (ML) methods to deal with the identical issue. Nevertheless, there hasn't been any direct evaluation of the developing and traditional approaches to this issue up to this point. Since the optimum management issue has gained popularity as a study topic, reinforcement learning (RL) methods are preferred for supervised learning in the ML field due to the impossibility of selecting every possible solution for continuous-state issues and the lack of a specified trainer. This work proposes an approach to tackle autonomous robotic navigation using the 2 most widely used RL methods, Sarsa <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$(\boldsymbol{\lambda})$</tex> and Q <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$(\boldsymbol{\lambda})$</tex>. To boost efficiency, the suggested framework, built using MATLAB, utilizes condition and action collections that are constructed uniquely. Both in simulation and real-world settings, the platform has an elevated performance percentage in navigating autonomous robotics to a specified destination by overcoming barriers. Furthermore, it is feasible to compare the results of the Sarsa and Q techniques and evaluate how the starting variables employed by the RL techniques, such as <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$\boldsymbol{\lambda}$</tex>, affect learning.