Navigation of autonomous mobile robots in unknown and unpredictable environments is a challenging domain to test and/or demonstrate knowledge representation and reasoning techniques because it involves a number of unique characteristics:
The input to the control system, particularly when provided by sonar range finders and odometric wheel encoders, is inaccurate, sparse, uncertain, and/or unreliable.
No complete mathematical representation exists of the process termed “navigation,” although, as demonstrated by humans, a set of skills for accomplishing this process exists that can typically be represented in a linguistic manner as IF-THEN rules (e.g., if the goal is to the left, then turn left; if an obstacle is detected to the right, then bear left).
The approximations involved in the numerical representation of the system and its environment (e.g., geometric representations, map discretization in grid) are significant.
A navigation environment is in general dynamic and unpredictable, typically leading to large uncertainties in its representation.