Fusion & AI > Tezler

In this study, path planning applications were developed for autonomous vehicles using more than one reinforcement learning method. Reinforcement learning methods are among the most successful methods in autonomous control systems. Reinforcement learning can provide optimum actions for agent. Firstly, Q-learning method is developed for path planning in static environments. Q-learning works in discrete state and action spaces. Second section of study Dueling Double Deep Q-learning method applied to agent in dynamic environment. The extended kalman filter is used in environment observation. In the last part, all methods, which applied to the problem, compared each other’s. [pdf]