In this presentation, a new robot navigation system that operates in corridor environments, using a user-provided sketch map, is proposed. This system make it unnecessary for the robot to explore the whole environment and successfully create the map before it becomes capable of navigation, as it is in most case to date. The sketch map of the environment provided by the user is a two-dimension map similar to floor plans usually found in buildings. It has a major different to commonly-used auto-generated map in its lack of details and accurate metric and geometric information. This presents an important problem, especially when robot can observe its environment only partially.
The navigation system extracts structural information of the corridors from the user-given map, and use it for planning and localization during navigation. This information is represented as a topological map augmented with node properties.A localization method based on Multiple Hypothesis Tracking (MHT) is implemented to address the ambuguity caused by discrepancy of the map and the observed environment.
In this presentation, first, the sketch floor map and its characteristics are described, and a method to extract topological map and other information that are used in navigation process is described. After a brief overview of how a robot models the environment using input from range sensor, a localization technique based on Multiple Hypothesis Tracking (MHT) is presented. Finally, results of navigation experiment, conducted in both simulated and real environments, are discussed and a conclusion and future works are discussed.