The Research And Application Of Hierarchical Reinforcement Learning And Affordance Model

Since artificial intelligence has been proposed, to make the robot smart as to think andlearn like human beings is considered a very difficult issue. Scientists summed up theautonomous learning methods for robot sprout the robot autonomous mental developmentideas after years explore. Research and development of robots and humans are inextricablylinked to their own. In1979, the concept of Psychology Affordance had been proposed. Itdiscussed and studied what human see in the environment to elaborated how we percept theworld around us. Through this giving an inspiration to how robot extract information from theenvironment to form Affordance.Robot path planning problem is a core issue of artificial intelligence. The conventionalpath planning algorithm can solve this problem under physical obstacle environment. But thisalso make the robot fall into a mindset that have to avoid obstacles. Changing a think, robotmat not escape the obstacle all the time, but could perform other strategies. In the paper, Ianalysis the essential attribute of the environment, study the Affordance of the obstacle,combined with hierarchical reinforcement learning. Then apply them in the obstacleenvironment of robot path planning problem.The first chapter Introduction section, starting from the robot profile, leads to the conceptof robot autonomous mental development and potential actions.Then analyzes thesignificance of the Affordance of the robot autonomous mental development.The second chapter introduces what is Affordance of robot, and then analyzes theresearch status of Affordance model. Based on the advantages and disadvantages ofAffordance model, I propose the energy of Affordance model.The third chapter introduces hierarchical reinforcement learning theory and thedecomposition of the value function and status of abstract theories. Explain the importance ofthe state of abstraction, and then propose a hierarchical model of the Affordancereinforcement learning algorithm.Chapter four set up the Affordance model to simple and complex obstacles environment.Then complete the experiment tasks combined with hierarchical reinforcement learning. Last use the two Affordance model to a scene application.