Artificial Intelligence for Autonomous Navigation
A robot is a device that is able to function in ways based on the form of a human. Limbs such as arms and legs are roughly modeled after the living body. The electronics are analogous to a living heart, the hardware is equivalent to the living body, and the mind is most similar to the human brain. In order for any body to be able to interact with its surroundings, real-time sensing/reaction is required. Artificial Intelligence (AI) allows humans to teach robots to have a likeness to themselves.
AI is used in order for a non-living robot to be able to understand what good science is. It allows for subjective viewpoints to produce better results with given surroundings. AI is used whenever a robot is expected to encounter unknown, unsafe, or unexpected situations.
AI Techniques are broken up into six categories: sensing, planning, fuzzy processing, neural networks, genetic algorithms, and teleoperation. The combining of the first five allow for increasingly more lifeline intelligence based on how much interaction is possible.
The ability to sense the circumstances of the surroundings allows an AI-enabled robot to make an aim for the best possible results. The Mars Exploration Rovers, currently operating on Mars, are able to detect surface hazards, and steer around them in order to reach a goal. Without this ability, either a robot would commonly become stuck in a hazard, or a controller on Earth would have to steer the robot with a several-minute delay. Teleoperation is a technique that allows someone on Earth to mostly let a robot do as it wants, but steps in when the subtleties of the situation, and knowledge known from other places, gives another option better possibilities. The more information that a robot collects, the better the ability to make decisions.
Fuzzy Processing allows an AI-enabled robot to interact with its surroundings in a more human practice; rather than considering a condition to be true or false, Fuzzy Logic allows a robot to understand the subtleties of "almost," "nearly," and "partially." Rather than every response being true or false, a response can be a possibility along a gradient. Fuzzy Logic can give a robot the freedom to identify the better of two imperfect options.
Neural Networks enable a robot to identify by association. This can play a basic role in teaching a robot to become an object that can "know" things. Supervised Neural Networks teach robots much like children learn in the classroom; Unsupervised Neural Networks allow the robot to analyze an effect that takes place, a perform accordingly in the future. Conversely, Genetic algorithms model the learning process as an evolution as seen in nature.
Humans wish to model robots based on things they do themselves. This gives robots human qualities and human knowledge to what was an inanimate object. From driving on Mars to tasks more down to Earth, such an accomplishment has numerous possibilities, limited only by our ability to give life.
