Local Perception Systems

Abstract

Before we can proceed with an account of the needs of autonomous – or more appropriately semiautonomous – vehicles in terms of real-time perception and local awareness, it would be useful to describe the potential environments around which intelligent off-road vehicles are going to rove. Agricultural production sites are very diverse, which means that the potential surroundings of a vehicle that performs mechanized tasks are equally as varied. They range from vast regions in the North American Midwest, Australia, or Central Europe, the intensive farming and horticulture of Japan and Mediterranean countries, the highly mechanized specialty crops of South America, the west coast of the United States, and Florida, to the abundant paddy fields of Southeast Asia. All of these settings might appear to demand very special needs, but from a robotic vehicle standpoint, open environments (excluding greenhouses) can be classified into three main categories. Of course, each application will require specific tuning, but, generally speaking, the core systems will need to cope with the following three situations:

• rows of equally-spaced plants (Scene I);

• rows of equally-spaced trees (Scene II);

• rows of plants or trees growing within limits imposed by man-made structures (Scene III).