Navigation in unknown environment by building instantaneous spatial structures

摘要

A strategy typically employed for mobile robot navigation in an unknown environment is to follow a nominal straight-line path to the goal point. During travelling on the nominal path, the robot uses distance information, e.g. derived from sonar sensors, and geometric information to determine the spatial relations between the robot and the environment. Navigation in an unknown environment is still a challenging issue especially in the presence of cluttered objects or obstructions.There are two possible ways to path planning in an unknown environment: the first is to map the environment and navigate based on the map; the second is to assign a nominal path, which the robot follows whilst at the same time it senses obstacles and reacts to achieve a collision free trajectory. In both cases the robot circumnavigates obstructions and generates a new path from the initial location to the goal point. Often the strategies used for navigation employ simple path planning techniques aided by specific methods to recognize objects and construct a structure for the environment. In Chronis’ PhD thesis is this area, a ring of low level sonar sensors is used to get spatial relations between a mobile robot and its environment. The eventual goal is to use spatial relations for navigation of the mobile robot in an unstructured, unknown environment. However, Chronis’ work does not construct any model of perceived structures in the environment and does not involve any tolerance to sensor failure. The approach described in this thesis improves this earlier work in precisely these two areas.The proposed approach uses low level sensors, such as sonar sensors, to achieve navigation in an unknown and cluttered environment. It integrates sonar sensors and geometric information to construct structures of the environment and consequently establish a system that navigates effectively and quickly through cluttered objects and obstructions. It is shown that this strategy achieves efficiency and effectiveness in mobile robot navigation. The approach is also shown to be robust and tolerant to sensor failures. The strategy is not dependent on the number or type of sensors on the robot and does not assume a particular type of robot; it can work with any sensory method that can provide an object representation in two dimensions.