State space search system

• US 5,301,256 A
• Filed: 09/17/1992
• Issued: 04/05/1994
• Est. Priority Date: 09/20/1991
• Status: Expired due to Term

• Current Assignee: Mitsubishi Denki Kabushiki Kaisha (Mitsubishi Electric Corporation)
• Original Assignee: Mitsubishi Denki Kabushiki Kaisha (Mitsubishi Electric Corporation)
• Inventors: Nakamura, Shizuka
• Primary Examiner(s): Fleming, Michael R.
• Assistant Examiner(s): Hafiz, Tariq R.
• Application Number:
  US07/946,105
• Time in Patent Office: 565 Days
• Field of Search: 395/23, 395/22, 395/13, 395/10
• US Class Current: 706/62
• CPC Class Codes: G06N 3/08 Learning methods

Abstract

A state space search system searches for an optimal goal state by searching in the direction of a steepest descent with respect to an evaluation function. When the search becomes impossible upon reaching a local minimum of the evaluation value, the evaluation value is increased to continue the search in the direction of steepest descent.

Claims

1. A state space search method for searching for a goal state within a state space consisting of a plurality of states, said state space search method comprising the steps of:

selecting a state for searching;

determining current neighboring states of state under search according to a transition rule;

determining an evaluation value of the state under search according to an evaluation function;

comparing evaluation value of the current state under search with a predetermined evaluation value of the goal state;

determining evaluation values of the neighboring states according to the evaluation function when the evaluation value of the current state under search is different from the predetermined value; and

designating a neighboring state as a next state under search when the evaluation value of the current state under search is different from the predetermined value, the neighboring state having a minimum evaluation value amongst the evaluation values of the neighboring states and the current state under search.

2. A state space search system for searching for an optimal goal state within a state space comprising a plurality of states, and state space search method comprising the steps of:

(a) initializing a current state and evaluating said current state by means of an evaluation function;

(b) determining neighboring states of said current state according to a transition rule;

evaluating the neighboring states by means of the evaluation function;

(c) selecting neighboring state having a minimum evaluation value of a next state;

(d) determining whether the next state is a goal state, the search being terminated if the next state is a goal state;

(e) judging whether an evaluation value of the next state is smaller than an evaluation value of the current state;

(f) increasing the evaluation value of the current state to a value greater than the evaluation value of the next state;

(g) updating the current state to the next state; and

(h) repeating steps (b) through (g).

3. A state space search method for searching for a goal state within a state space comprising a plurality of states, said state space search method comprising the steps of:

selecting a current state for search;

determining an evaluation value of the current state according to an evaluation function;

determining neighboring states of the current state according to a transition rule;

comparing the evaluation value of the current state with a predetermined evaluation value of the goal state;

determining evaluation values of the neighboring states according to the evaluation function when the evaluation value of the current state is different from the predetermined evaluation value of the goal state; and

selecting a neighboring state having an extreme evaluation value amongst the evaluation values of the neighboring states and the current state for search when the evaluation value of the current state is different from the predetermined evaluation value of the goal state, wherein an extreme evaluation value is defined as a minimum or a maximum value.

4. A state space search method according to claim 3 further comprising the step of adjusting the evaluation value of the current so that the evaluation value of the current state is less extreme than the most extreme evaluation value of any of the neighboring states prior to the step of selecting a neighboring state.