nextuppreviouscontents
Next:Robotics Up:Mobile Robotics Previous:Geon Recognition using a

ORB: Object Recognition for Real-time Autonomous Mobile Robot Navigation

D.T. Bui, M.D. Levine

An object recognition system called ORB (Object Recognition and map
Building using the QUADRIS sensor platform on a mobile robot.) is proposed and implemented for use on a mobile robot. ORB utilizes the QUADRIS sensor platform developed at CIM (Center for Intelligent Machines) at McGill University, which is composed of two BIRIS (Official trademark of the National Research Council of Canada.) laser range finders. ORB performs a series of sensory and perceptual tasks in conjunction with a mobile robot control architecture called SPOTT. ORB's main task is to sense the mobile robot's surroundings and provide laser range data, in the form of line segments, for SPOTT's map database. In an office environment, ORB also identifies and labels the structural objects (i.e., walls, doors) in this map. While navigating through an office space, the mobile robot may be required to search for certain objects in the area. In this scenario, ORB is used to recognize the movable objects (i.e., chairs, tables and desks). ORB is able to perform its tasks in a fast and efficient manner by using simple models to represent the structural and movable objects in the database. ORB's recognition procedures only require sparse sets of range  scans to identify  aforementioned
objects. The structural object models are built from prior knowledge of the office environment. For example, the doorway model would consist of the known doorway widths found on the experimental
office floor. ORB has been tested extensively in the CIM environment, but it can also be applied to any office space provided the structural dimensions are known a priori. ORB's models for the movable
  objects are idealized descriptions with the object's surfaces represented by planes. The physical dimensions of the movable object models are defined by Architectural Standards, as
office furniture is built to conform to these standards.


Annual Report

Fri Nov 26 23:00:32 GMT 1999