| Author: | Dai, Xiaohui |
| Title: | Studies on intelligent adaptive control of autonomous systems with applications to longitudinal vehicle following |
| Degree: | M.Phil. |
| Year: | 2003 |
| Subject: | Hong Kong Polytechnic University -- Dissertations Intelligent control systems Autonomous robots Robots -- Control systems |
| Department: | Department of Electronic and Information Engineering |
| Pages: | x, 120 leaves : ill. ; 30 cm |
| Language: | English |
| Abstract: | In the last two decades, the design of intelligent vehicles that either assist or replace the driver has attracted a lot of attention from both academic researchers and industrial entrepreneurs. This thesis addresses the important problem of autonomous vehicle control within the academic framework and provides new algorithms for solving longitudinal vehicle following based on adaptive and fuzzy control methodologies. Throughout this thesis, the author reports the design of intelligent controllers with more flexibility but less a priori knowledge about system models. In the first stage, the author concentrates on the development of a novel adaptive fuzzy controller. The vehicle longitudinal control system falls into a class of partially known nonlinear systems. A fuzzy system is employed to approximate the ideal controller. A relationship between approximation error and parameters of the fuzzy controller is established first. Then, the adaptive laws of the fuzzy controller are obtained based on Lyapunov synthesis approach. All the parameters of fuzzy controller are adjustable. This is the major difference between my work and the others. However, a weakness of the proposed adaptive fuzzy controller is that it requires some information about the system and it only aims at a specific nonlinear system. To this end, I investigate Q-learning, a model free reinforcement learning (RL) method, and its applicability as a controller design approach for real systems in a knowledge-poor environment. The focus is on two issues: (i) the structure of the Q estimator network and fuzzy controller, and (ii) the development of learning algorithms for both of them. A Takagi-Sugeno type fuzzy inference system and a multiple-layer feed-forward neural network are employed as action producer and Q estimator respectively. The learning algorithms for the Q estimator network and the fuzzy controller are developed based on the temporal difference methods as well as the gradient descent algorithm. The efficiency of applying RL directly may not always be appropriate. Therefore, the author proposes a controller based on dual heuristic programming (DHP) to enhance the controller performance. The structure and adaptation algorithms of the controller for vehicle following problems are presented. The proposed controller has two advantages compared with other controllers based on adaptive critic designs: (i) the system model is not required directly or indirectly, and (ii) it can take advantage of the TS type fuzzy controller to incorporate a priori knowledge. The simulation results of the controller based on RL and those of the controller based on DHP are compared and the advantages of the technique are also explored. The application of these intelligent adaptive controllers to autonomous vehicle control systems has been described. Conclusions are drawn based on studies performed via theoretical analysis and computer simulations. |
| Rights: | All rights reserved |
| Access: | open access |
Files in This Item:
| File | Description | Size | Format | |
|---|---|---|---|---|
| b17404009.pdf | For All Users | 3.59 MB | Adobe PDF | View/Open |
Copyright Undertaking
As a bona fide Library user, I declare that:
- I will abide by the rules and legal ordinances governing copyright regarding the use of the Database.
- I will use the Database for the purpose of my research or private study only and not for circulation or further reproduction or any other purpose.
- I agree to indemnify and hold the University harmless from and against any loss, damage, cost, liability or expenses arising from copyright infringement or unauthorized usage.
By downloading any item(s) listed above, you acknowledge that you have read and understood the copyright undertaking as stated above, and agree to be bound by all of its terms.
Please use this identifier to cite or link to this item:
https://theses.lib.polyu.edu.hk/handle/200/1077