Dissertation > Excellent graduate degree dissertation topics show

Research on Reinforcement Learning Based on Value Function Approximation and State Space Decomposition

Author: ZuoLei
Tutor: XuZuo
School: National University of Defense Science and Technology
Course: Control Science and Engineering
Keywords: Enhance learning Value function approximation Said policy iteration Space Decomposition Autonomous Obstacle Avoidance
CLC: TP242
Type: Master's thesis
Year: 2011
Downloads: 17
Quote: 0
Read: Download Dissertation

Abstract


Enhanced learning can effectively solve the uncertain sequential decision optimization problem in recent years has developed into a hot research topic in the field of machine learning. How to overcome the \On the other hand, with the expansion of the scope of application, the mobile robot will face a more complex and changeable unknown environment, intelligent navigation of mobile robot control technology put forward higher requirements. How to improve the autonomous navigation capabilities of the mobile robot and the adaptive capacity of the environment, is key to the successful application of the mobile robot in an unknown environment. Reinforcement learning methods based value function approximation with state space decomposition in-depth study, and autonomous obstacle avoidance control applied to mobile robot in an unknown environment. The research achievements include: 1. Propose a representation policy iteration k-means clustering-based learning methods. This paper studies the algorithm based on graph Laplace operator representation policy iteration (RPI), and then use cluster analysis to improve the composition point selection method, proposed RPI based on k-means clustering algorithm, the simulation results show that the The method can effectively improve the generalization performance of RPI algorithm. 2. Research and achieve real-time learning control for the inverted pendulum system. In this paper, the linear value function approximation methods based on policy iteration (RPI) algorithm and its improved algorithm for the model inverted pendulum real-time learning control, better control effect and to enhance the study of the actual Engineering Applications meaningful exploration. 3. Proposed a method based on the space decomposition structured policy iteration (HRPI). First structured to enhance the learning algorithm, and then combined the RPI algorithm with the state space decomposition method, based on state space decomposition structured to enhance learning method HRPI. This method is based on the approximation of function of the state space is decomposed into different sub-space, and then in each sub-space strategy learning. The simulation results show that this method has good generalization performance in solving the time optimal. 4. Proposed control method based on the improvement of RPI's mobile robot obstacle avoidance. This paper first introduces autonomous mobile robot in an unknown environment the MDP modeling methods to avoid the obstacle problem, then scroll window path planning and RPI algorithm combining an RPI-based mobile robot autonomous obstacle avoidance control method, and simulation The experiment tested the effect of the method generalization performance and obstacle avoidance. The experimental results show that the mobile robot obstacle avoidance RPI-based reaction navigation control method can effectively achieve autonomous obstacle avoidance in an unknown environment.

Related Dissertations

  1. The Research of Mobile Machine Vision Navigation and Obstacle Avoidance System,TP242
  2. Reinforcement Learning-based Routing Mechanism for Multi-sink Wireless Sensor Networks,TP212.9
  3. Research on Obstacle-avoiding Fuzzy Control System of Mine Rescue Robot,TP242
  4. Distributed Coordination iron production process optimization methods,TF58
  5. Research on Reinforcement Learning Methods for Navigation and Control of Autonoumous Mobile Robots,TP242.6
  6. The Development of Control System for a Domestic Partner Robot,TP242.6
  7. The Spatial Scale Effects of the Regional Economic Disparities in the Central China Area,F224
  8. Research on Kernel-based Reinforcement Learning in Continuous State and Action Spaces,TP181
  9. Research and Implementation of Reinforcement Learning Based Network Adapttive Realtime Video Transport System,TN919.8
  10. Neuron-online Optimization Algorithm of Conveyor -serviced Production Station,TH237.1
  11. Research on Behavior-based Obstacle Avoidance Control for Intelligent Exploration Vehicle in Unknown Environments,TP242
  12. Visual attention and cognitive behavioral model and its application,TP391.41
  13. Research on Fast Packet Classification Algorithm,TP393.01
  14. Research on Motion Control of Mobile Robots Based on Reinforcement Learning,TP242.6
  15. Computer Animation Technology,TP391.41
  16. Design and Implementation of Tour-guide Robot,TP242.6
  17. Surface Mesh Automatic Generation System Based on Delaunay Subdivision,TP391.41
  18. Fractional non- Gaussian noise frequency estimation method and program implementation,TN911.7
  19. The Research of Autonomous Obstacle Avoidance for Air Duct Cleaning Robot Based on Fuzzy Control,TP242
  20. Research and Realization on Collision Detection in Virtual Environment,TP391.9

CLC: > Industrial Technology > Automation technology,computer technology > Automation technology and equipment > Robotics > Robot
© 2012 www.DissertationTopic.Net  Mobile