PhD thesis on Computer Science from the Technical University of Catalonia (AI program)
Realized at the Artificial Intelligence Research Institute (IIIA-CSIC)
Also published as Monografies de l'IIIA (Vol. 18)

Author: Dídac Busquets
Supervisors: Ramon López de Màntaras and Carles Sierra
Defense date: July 28, 2003

Online document: HTML
Complete document: PDF (1.2 MB)

• Autonomous Robot Navigation page
• ARGOS Project
• Autonomous Agents 2000 demo

PDF (32 KB)

PDF (44 KB)

This chapter gives an overview of this PhD thesis, its motivations, objectives and its main contributions. It also gives a list of the publications originated from the research carried out during the completion of the thesis.

PDF (67 KB)

This chapter is devoted to relevant related work and state-of-the-art on the field of autonomous robots for unknown unstructured environments. The relevant work has been divided in two parts, one for each main thread of research of the thesis: control architectures, and mapping and navigation methods. The relevant work concerning control architectures gives an overview of the different approaches on autonomous robots control, focusing on the behavior-based approach. Regarding the mapping methods, we review and compare the two main approaches for map building, the metric one and the topological one. A comparison between two dif- ferent localisation approaches (landmark-based localisation and model matching) is also given.

PDF (113 KB)

In this chapter we firstly describe Prescott’s model for storing spatial relationships among the landmarks of the environment. After that, we describe how we have extended this model for dealing with imprecise information about the location of the landmarks. We also present the algorithm for building a topological map of the environment and how it is used to compute diverting targets, needed by the robot when it is blocked.

PDF (172 KB)

In this chapter a general coordination architecture based on a bidding mechanism is presented. We also present the particular instantiation of the general architecture that we have used to solve the navigation problem. A detailed description of the multiagent Navigation system is also given in this chapter.

PDF (349 KB)

In this chapter the results of the simulated experiments are presented. These experiments include the testing of our architecture and the application of Machine Learning techniques in order to improve the performance of the system. In particular, we present the application of Reinforcement Learning, which we have used to make the system learn how to appropriately use the camera, and an application of Genetic Algorithms, used to tune some of the parameters of the agents of the Navigation system.

PDF (1,23 MB)

This chapter is devoted to present the results of the experiments on real environments with a real robot. Firstly, the wheeled robot platform and a simple vision system used for the real environments experiments are described. Then, we describe the different scenarios in which the experiments have been carried out. Finally, the results of the experimentation in such scenarios are given and discussed.

PDF (39 KB)

In this chapter, we summarize the main contributions of the thesis, and point out some open problems and future research perspectives that we plan to tackle in the near future.

PDF (26 KB)