Mobile Robot Planning with Incomplete Information in Dynamic Environments
Brian M. Yamauchi · 1996
If mobile robots are to be useful in real-world envirorunents containing human beings, these robots will need to be able to deal with dynamic changes. This paper describes the problems of incomplete infonnation raised by dynamic environments and outlines questions intended to open a dialogue between planning researchers studying the role of incomplete infonnation and roboticists building mobile robot systems for dynamic, real-world environments. In addition, this paper provides an overview of one approach to this problem which makes use of representations and algoritruns designed specifically for robot exploration and navigation in dynamic environments. Dynamic envirorunents introduce the problem of incomplete information in a number of different ways. First, the robot may have incomplete infonnation about the future state of the environment, since it may be impossible to predict exactly what changes will occur and when they will occur. Second, the robot may have incomplete infonnation about the present state of the envirorunent, since changes may be have occurred in a different part of the envirorunent since the time of the robot's last visit. As a result, any attempt to complete the robot's infonnation about the world can only be successful for a limited period of time. In large envirorunents. the robot's world model may always be incomplete, regardless of attempts to gain new infonnation. For example: consider a mobile robot navigating through an office environment. Open doors allow travel and closed doors obstruct travel, so a complete model would require the knowledge of whether each door is open or closed. However, in a real office building, people will constantly be opening and closing doors as they move from one location to another, so the robot will never be able to know the complete state of the world.