I was just thinking about a possible project - designed to help automate some problems involved in wilderness search and rescue - and ended up with several places that a rule based subsystem would have been a good thing.
First - there's a lot of paperwork involved and its not easy to keep straight. A rule based system could help decide which forms were required this time and help generate them - it could also prompt for information that is needed for one set of forms but not another.
Second - there's a lot of experience built up - both around the country and the world and locally that can help to guide a search effort. Much of this is easily adaptable to a rule based system : "Alzheimers patients tend to ....", "If a hunter is lost on Bald Peak or that area, they most frequently come down Potter's Creek"
Third - The first thing you get to do when you run a search is do a lost person interview - and the information needed can be extensive. A rule based system could help prompt the interviewer (usually a Search and Rescue person, but sometimes a law enforcement officer who happens to be nearby) to get all the relevant information and ensure that nothing gets missed.
And there were a couple other minor areas as well.
Most of these areas are relatively small, involving dozens to a hundred or so rules. But in some ways its just those areas that are the most fruitful - rule based systems tend to be robust in those situations and small numbers of rules phrased as rules rather than procedural code are much easier to read and keep track of.
|