| dc.description.abstract | Autonomous travel poses challenges in machine learning navigation. Different approaches have been considered, such as reinforcement learning, dynamic programming methodologies, and other artificial intelligence assisted solutions. Dynamic programming methodologies exist to assist an agent, or robot, to interact with the environment based on predetermined environments. Real-world terrain and real-time complexity make learning and interacting with an environment particularly difficult. Reinforcement learning is a methodology that provides the agent with a learning system that is more environment and terrain independent. To this end, reinforcement learning is an effective way to solve the autonomous vehicle problem by proposing a model-free, or nearly model-free approach. The reinforcement learning algorithm used for this autonomous vehicle is Q-learning, which is an unsupervised learning algorithm where the agent explores from state to state until his goal is reached. With this algorithm, learning is achieved by sequential states of winning and losing, subject to obtaining the goal or failing to obtain the goal. The objective of this approach is to use particular goals to obtain navigation independent of the environment. | en |
