Route planning using pattern classification and search techniques

A prototype route planner was developed in the context of mission planning for air interdiction. A representative geographical scenario was divided into a rectangular grid, with each intersection having nine characteristics associated with it. These characteristics form a pattern vector which describes the attributes of each intersection. A representative sample of these vectors was rated by pilots as to the desirability of overflying points with those particular characteristics. A minimum-distance pattern classifier was constructed using these data. A route planner was then constructed using an 'algorithm A' search routine. The route planner attempts to find a low-cost path to a target by using a heuristic combining distance and pattern classifier output.<>