ARK: autonomous mobile robot for an industrial environment

The ARK (Autonomous Robot for a Known Environment) Project is a precompetitive research project involving Ontario Hydro, the University of Toronto, York University, Atomic Energy of Canada Ltd., and the National Research Council of Canada. The project started in September 1991 and completed in August 1995. The technical objective of the project was to develop a sensor-based mobile robot that can autonomously navigate in a known industrial environment.

There are many types of industrial operations and environments for which mobile robots can be used to reduce human exposure hazards, or increase productivity. Examples include inspection for spills, leaks, or other unusual events in large industrial facilities, materials handling in computer integrated manufacturing environments, and the carrying out of inspections, the cleaning up of spills, or the carrying out of repairs in the radioactive areas of nuclear plants - leading to increased safety by reducing the potential radioactive dose to workers.

The industrial environment is significantly different from office environments in which most other mobile robots operate. The ARK project will produce a self-contained mobile robot with sensor-based navigation capabilities specifically designed for operation in a real industrial setting. The ARK robot is currently being tested in the large engineering laboratory at AECL CANDU in Mississauga, Ontario. This open area covers approximately 50,000 sq. feet of space and accommodates one hundred and fifty employees. Within the Laboratory, there are test rigs of various sizes, mockups of reactor components, a machine shop, a fabrication facility, a metrology lab and assembly area. There are no major barriers between these facilities and therefore at any one time there may be up to fifty people working on the lab floor, three fork lift trucks and floor cleaning machines in operation. Such an environment presents many difficulties that include: the lack of vertical flat walls; large open spaces (the main isle is 400' long) as well as small cramped spaces; high ceilings (50'); large windows near the ceiling resulting in time dependent and weather dependent lighting conditions, a large variation in light intensity, also highlights and glare; many temporary and semi-permanent structures; many (some very large) metallic structures; people and forklifts moving about; oil and water spills on the floor; floor drains (which could be uncovered); hoses and piping on the floor; chains hanging down from above, protruding structures, and other transient obstacles to the safe motion of the robot.

Large distances, often encountered in the industrial environment, require sensors that can operate at such ranges. The number of visual features (lines, corners and regions) is very high and techniques for focusing attention on specific, task dependent, features are required. Most mobile robotic projects assume the existence of a flat ground plane over which the robot is to navigate. In the industrial environment this ground plane is generally flat, but regions of the floor are marked with drainage ditches, pipes and other unexpected low lying obstacles to movement. The ARK robot requires sensors that can reliably detect such obstacles.

The ARK robot's onboard sensor system consists of sonars and one or more ARK robotic heads and a floor anomaly detector (FAD). The head consists of a colour camera and a spot laser range finder mounted on a pan-tilt unit. The pan, tilt, camera zoom, camera focus and laser distance reading of the ARK robotic head are computer controlled.

The ARK robot must navigate through its environment autonomously and cannot rely on modifications to its environment such as the addition of radio beacons, magnetic strips beneath the floors, or the use of visual symbols added to the existing environment. In order to navigate within this environment the ARK robot uses naturally occurring objects as landmarks. The robot relies on vision as its main sensor for global navigation, using a map of permanent structures in the environment (walls, pillars) to plan its path. While following the planned path, the robot locates known landmarks in its environment. Positions and salient descriptions of the landmarks are known in advance and are stored in the map. The robot uses the measured position of the detected landmarks to update its position with respect to the map.

The ARK project built a total of three robots. ARK1 was the original research vehicle and was built around a CYBERMOTION Navmaster platform. ARK1 is housed at the University of Toronto. ARK-lite was the second research vehicle and is housed at York University. Finally ARK-2 is the industrial prototype vehicle and is housed at AECL.