A Simple World Model for an Autonomous Vehicle

A critical component of the guiding intelligence of an autonomous vehicle is the system's computational model of its physical environment. This world model must be able to represent the important attributes of the vehicle's environment, the vehicle's placement in that environment, and the system status (or health) of the vehicle. For a submersible autonomous vehicle, important environment attributes include the bottom topography, the surf ace topography (if under an ice layer), and objects of note (including itself) . System status is somewhat more vehicle specific, but includes those aspects of the vehicle which are critical to sustained operation: sensors, effectors, energy budget, and various "vital signs". This paper presents a prototypical world model which ad dresses these issues. The model was implemented using an object oriented programming language. This approach was found to be quite advantageous in that the decomposition of the physical environment into smaller units resulted in a set of easily coded entities. in this model, each environment attribute is represented by a computational object which records data and behaviors concerning that attribute. For example, bottom and surface topographies are represented by a collection of programming objects which are arranged in a quadtree data structure and maintained by functions which are bound to those objects.