An issue central to the navigation problem is memory. Traditional systems build symbolic maps of the world for navigational reference. Reactive methods, in contrast, eliminate or minimize the use of memory. These reactive techniques have been remarkably successful at solving a wide range of navigational problems. Some problems, however, still present a challenge to reactive strategies, (box canyons for example). The addition of a local spatial memory allows a robot to avoid areas that have already been visited. “Avoiding the past” offers a solution to the box canyon and other navigational problems. An avoid-past strategy has been implemented using a spatial memory within a schema-based motor control model. Experiments have produced promising results in simulation and on mobile robots.
Tucker R. Balch, Ronald C. Arkin