Pest Control I:

Burmese Pythons

Back to Contents

    Currently infesting southern Florida and spreading, the Burmese Python constitutes an invasive species that has a deleterious impact on many native species, with the possibility of driving at least some of them to extinction. These snakes are dark-colored with a pattern down the back consisting of many brown blotches bordered in black. That skin pattern, because of its attractiveness, contributes to their popularity with both reptile keepers and the leather industry. In the wild, Burmese Pythons grow to a length of 3.7 meters (12 feet) on average, growing larger in rare instances, so these are formidable reptiles.

    The Burmese Python must have a permanent source of water nearby, so it exists naturally in environments in which substantial amounts of standing or running water exist. In the wild its natural habitat extends, east to west, from Vietnam to eastern India and, north to south, from Nepal to Indonesia. In the United States southern Florida provides an ideal environment for these snakes.

    Climatic conditions make any migration of the snakes out of southern Florida unlikely, but in that vast wetland they have become a disaster. Released by people who no longer care for them or by the damage from hurricanes, the snakes have become a menace to other animals, especially in the Everglades. As of 2009 Dec 31 over 1300 of the snakes have been captured in the Sea of Grass, where they eat birds, small alligators, and other, endangered species. If we wish to maintain a restorable natural environment in Florida, we must eliminate the Burmese Pythons, all of them: we must extinguish them completely.

    Achieving that goal requires that someone create a fully autonomous robot predator that will exterminate the Burmese Pythons and nothing else. It must be a self-guided machine that can roam the swamps and grasslands of southern Florida, swimming when necessary, find the snakes, and capture them.

    It will vaguely resemble a centaur. That is, it will have roughly the size of a full-grown man; it will have four legs (to provide a stable support for walking and fighting snakes); and it will have a pair of very long arms with hand-like manipulators at their ends. And jutting up from between the arms a thick stalk will hold up the sensor pod (the head). The sensor pod will be wide, like the head of a hammerhead shark, to give the robot very good depth perception in vision and good directionality in hearing.

    Of all the senses that we want to give the robot, the sense of touch comes easiest, made all the easier by the fact that the robot does not have to feel texture. Pressure sensors on the inside surfaces of the hands and the feet will enable the robot to know when it touches something and with how much force. Angle sensors inside the robotís joints will give the robot a kinesthetic sense, enabling it to know where every part of its body is positioned relative to some fixed point on or in that body. Combining touch and kinesthesia enables the robot, for example, to determine whether itís walking on solid ground or on mud.

    Hearing gives the designers the next hardest sense to devise for a robot, though itís still fairly easy because itís one-dimensional. Sound converted into electrical pulses and brought into the robotís dataspace through microphones (one on each side of the sensor pod to give the robot binaural hearing) will pass into an electronic cochlea. There the electrical vibrations get separated into their component frequencies, whose volumes are monitored over brief intervals of time and passed into a neural net that learns to associate each pattern of sound with suitable words, such as "hoot of an owl" or "snake slithering through grass". The robot can then process those words through the logic of Boolean algebra and thereby associate them with other words, such as words of intent.

    Most animals explore their environment through chemical cues, so our robot needs a sense of smell. Instead of using a nose on its sensor pod, the robot will draw air in through openings in the tips of its fingers and pass it over a miniature Omniphage inside the palm of its hand. As the Omniphage disassembles the organic molecules floating in the air it will identify them and report their relative frequency in the mixture to a neural net that will associate the pattern with words that denote the associated odor.

    Of all the senses that the robot must have, vision will give the designers the most difficulty in developing. The sensor itself is not so difficult. A television camera that uses a charge-coupled device to turn light into electric charge closely mimics the eye with its retina: it produces a pointillist array representing brightness and color. The difficult part consists of associating that pattern with specific words, largely because any scene projected onto a retina, natural or artificial, contains such a large number of objects that itís hard to isolate one object and associate it with a word.

    One obvious way to train a robot to see things mimics the way we teach children to recognize items. The trainers present the item against a completely blank background and condition the robot to associate the item, seen from different angles and under different lighting conditions, with a word that names the item. Then the trainer tests the robot by asking it to find the item in a more cluttered environment, giving the robot a word and seeing how well the robot sees the item amid the visual noise. The designers have achieved their goal when the robot displays the ability to spot a Burmese Python in the natural environment and does not mistake another species of snake for its prey.

    Once the robot has been properly trained and programmed, its builders can copy the connection settings in its computer/brain into other robots as they come off the production line. The robots will retain the ability to learn, but they will begin their existence with the ability to begin carrying out their duties.

    In their primary duty the robots will find and capture Burmese Pythons. Just in case something goes wrong (and it likely will), the designers wonít allow the robots to kill the snakes. When a robot captures a python, it will call a human crew to come and pick it up, though a robot helicopter with a cage may do the actual pickup. We want to focus all of the robotís effort into finding the snakes with such perfection that we donít create a predator/prey cycle; that is, we donít want to create a situation in which the prey adapts to the predation and, thus, continues to exist. We donít want to evolve a snake that can successfully evade the robots. We want to render natural selection inoperative and drive the pythons to extinction in Florida.

    One additional problem comes from our wanting our pest-control robots to wander over wide areas of the landscape in search of their prey. We want to give our robot a power source that will keep it going for a reasonably long time, a matter of days at minimum. Solving that problem involves building a tank into the robotís body. An Omniphage built across the tankís bottom will disassemble organic matter and combine the atoms with atmospheric oxygen to generate power in the manner of a fuel cell. A hatch on the robotís back enables the robot to fill the tank with organic material, such as fallen leaves, dry brush, or dead animals, obtained from its environment. Self-fueled, living off the land, the robot can pursue its prey indefinitely.

    Once the pythons are extinct in Florida the program managers can redirect the robots to pursue other invasive species, such as the walking catfish that also infest Florida. In Australia, the robots will hunt rabbits and cane toads. Given the way in which invasive species travel around the world these days, we can say confidently that these robots will never be out of work.


Back to Contents