Internal models

As we want to create  more and more complex  machine with  software programs as controllers, we search  that  it is often needful  to create  additional types  of abstraction on top of the one given by a generic  programming language. This is particularly true when  the function of the exact  calculation needs relays on considerably on details that  is only received during the running of the machine.

It is theoretically possible  to create  an enormous digital  computer program or circuit that has a look-up table, in which  we calculated the path  for all pairs  of locations and  save them  away. Then when  we want  to search a path,  we could  simply  take in the table at the place  corresponding to the initial  and  goal position  and  get the computed path. We may create and design a general-purpose algorithm that will result any shortest-path problem in a graph. That algorithm and  implementation will be needful  for a wide  variety of possible  answers.

Another example of using  internal methods is when  the computer has some significant lack of information about  the state of the environment. In such conditions, it may be required to explicitly model  the set of given  possible states  of the external world and  their  related probabilities, and  to modify this  model  over  time  as new  evidence is collected.  We will give an example of this approach near the end of the course