Probabilistic robotics, also called statistical robotics, is a field of robotics that involves the control and behavior of robots in environments subject to unforeseeable events. Because reality always involves uncertainty, probabilistic robotics may help robots to more effectively contend with real-world scenarios.
Originally, probabilistic robotics involved the ability of a robot to locate itself using maps of known work environments. A blueprint of the surroundings, along with tools such as proximity sensing and machine vision, allowed a robot to navigate and perform tasks with a minimum of errors or mishaps.
More recently, probabilistic robotics has become concerned with the development of robots that can work effectively in environments that they have not previously encountered. Therefore, a robot must develop a sense of the most likely result of a given movement or action, based on defined statistical functions, and then strive for the optimum outcome.
See a demonstration of grid localization in probabilistic robotics: