A Multiagent Approach to Qualitative Navigation in Robotics

Abstract

Navigation in unknown unstructured environments is still a difficult open problem in the field of robotics. In this PhD thesis we present a novel approach for robot navigation based on the combination of landmark-based navigation, fuzzy distances and angles representation and multiagent coordination based on a bidding mechanism. The objective has been to have a robust navigation system with orientation sense for unstructured environments using visual information. To achieve such objective we have focused our efforts on two main threads: navigation and mapping methods, and control architectures for autonomous robots. Regarding the navigation and mapping task, we have extended the work presented by Prescott, so that it can be used with fuzzy information about the locations of landmarks in the environment. Together with this extension, we have also developed methods to compute diverting targets, needed by the robot when it gets blocked. Regarding the control architecture, we have proposed a general architecture that uses a bidding mechanism to coordinate a group of systems that control the robot. This mechanism can be used at different levels of the control architecture. In our case, we have used it to coordinate the three systems of the robot (Navigation, Pilot and Vision systems) and also to coordinate the agents that compose the Navigation system itself. Using this bidding mechanism the action actually being executed by the robot is the most valued one at each point in time, so, given that the agents bid rationally, the dynamics of the biddings would lead the robot to execute the necessary actions in order to reach a given target. The advantage of using such mechanism is that there is no need to create a hierarchy, such in the subsumption architecture, but it is dynamically changing depending on the specific situation of the robot and the characteristics of the environment. We have obtained successful results, both on simulation and on real experimentation, showing that the mapping system is capable of building a map of an unknown environment and use this information to move the robot from a starting point to a given target. The experimentation also showed that the bidding mechanism we designed for controlling the robot produces the overall behavior of executing the proper action at each moment in order to reach the target.