001 package aima.probability.reasoning;
002
003 import java.util.Arrays;
004 import java.util.List;
005
006 import aima.probability.RandomVariable;
007
008 /**
009 * @author Ravi Mohan
010 *
011 */
012 public class HMMFactory {
013
014 public static HiddenMarkovModel createRobotHMM() {
015
016 // Example adopted from Sebastian Thrun's "Probabilistic Robotics"
017 // Chapter 2
018 // A robot faces a door. The state of the door (open or closed)
019 // constitutes the "hidden state"
020 // The robot's sensor detects a "closed" or "open" state.
021 // (Perception may be out of synch with reality because sensors are
022 // probabilistic).
023 // The robot can either "do nothing" or "push" the door. These are the
024 // possible actions.
025
026 List<String> states = Arrays.asList(new String[] {
027 HmmConstants.DOOR_OPEN, HmmConstants.DOOR_CLOSED });
028 List<String> actions = Arrays.asList(new String[] {
029 HmmConstants.DO_NOTHING, HmmConstants.PUSH_DOOR });
030 List<String> perceptions = Arrays.asList(new String[] {
031 HmmConstants.SEE_DOOR_OPEN, HmmConstants.SEE_DOOR_CLOSED });
032
033 RandomVariable prior = new RandomVariable(states);
034 TransitionModel tm = new TransitionModel(states, actions);
035 // tm.setTransitionProbability(start_state, action, end_state,
036 // probability)
037 // given a start state and an action the probability of the end state is
038 // probability
039 tm.setTransitionProbability(HmmConstants.DOOR_OPEN,
040 HmmConstants.DO_NOTHING, HmmConstants.DOOR_OPEN, 1.0);
041 tm.setTransitionProbability(HmmConstants.DOOR_OPEN,
042 HmmConstants.DO_NOTHING, HmmConstants.DOOR_OPEN, 1.0);
043 tm.setTransitionProbability(HmmConstants.DOOR_OPEN,
044 HmmConstants.DO_NOTHING, HmmConstants.DOOR_CLOSED, 0.0);
045 tm.setTransitionProbability(HmmConstants.DOOR_CLOSED,
046 HmmConstants.DO_NOTHING, HmmConstants.DOOR_CLOSED, 1.0);
047 tm.setTransitionProbability(HmmConstants.DOOR_CLOSED,
048 HmmConstants.DO_NOTHING, HmmConstants.DOOR_OPEN, 0.0);
049
050 tm.setTransitionProbability(HmmConstants.DOOR_OPEN,
051 HmmConstants.PUSH_DOOR, HmmConstants.DOOR_OPEN, 1.0);
052 tm.setTransitionProbability(HmmConstants.DOOR_OPEN,
053 HmmConstants.PUSH_DOOR, HmmConstants.DOOR_CLOSED, 0.0);
054 tm.setTransitionProbability(HmmConstants.DOOR_CLOSED,
055 HmmConstants.PUSH_DOOR, HmmConstants.DOOR_CLOSED, 0.2);
056 tm.setTransitionProbability(HmmConstants.DOOR_CLOSED,
057 HmmConstants.PUSH_DOOR, HmmConstants.DOOR_OPEN, 0.8);
058
059 SensorModel sm = new SensorModel(states, perceptions);
060 // sm.setSensingProbaility(state,perception,p); given a state the
061 // probability of a perception is p
062 sm.setSensingProbability(HmmConstants.DOOR_OPEN,
063 HmmConstants.SEE_DOOR_CLOSED, 0.4);
064 sm.setSensingProbability(HmmConstants.DOOR_OPEN,
065 HmmConstants.SEE_DOOR_OPEN, 0.6);
066 sm.setSensingProbability(HmmConstants.DOOR_CLOSED,
067 HmmConstants.SEE_DOOR_OPEN, 0.2);
068 sm.setSensingProbability(HmmConstants.DOOR_CLOSED,
069 HmmConstants.SEE_DOOR_CLOSED, 0.8);
070
071 HiddenMarkovModel hmm = new HiddenMarkovModel(prior, tm, sm);
072 return hmm;
073 }
074
075 public static HiddenMarkovModel createRainmanHMM() {
076 List<String> states = Arrays.asList(new String[] {
077 HmmConstants.RAINING, HmmConstants.NOT_RAINING });
078 // no actions because the observer has no way of changing the hidden
079 // state and i spassive
080 List<String> perceptions = Arrays.asList(new String[] {
081 HmmConstants.SEE_UMBRELLA, HmmConstants.SEE_NO_UMBRELLA });
082
083 RandomVariable prior = new RandomVariable(states);
084
085 TransitionModel tm = new TransitionModel(states);
086 // tm.setTransitionModelValue(start_state, action, end_state,
087 // probability);
088 // given a start state and an action the probability of the end state is
089 // probability
090 tm.setTransitionProbability(HmmConstants.RAINING, HmmConstants.RAINING,
091 0.7);
092 tm.setTransitionProbability(HmmConstants.RAINING,
093 HmmConstants.NOT_RAINING, 0.3);
094 tm.setTransitionProbability(HmmConstants.NOT_RAINING,
095 HmmConstants.RAINING, 0.3);
096 tm.setTransitionProbability(HmmConstants.NOT_RAINING,
097 HmmConstants.NOT_RAINING, 0.7);
098
099 SensorModel sm = new SensorModel(states, perceptions);
100 // sm.setSensingProbaility(state,perception,p); given a state the
101 // probability of a perception is p
102 sm.setSensingProbability(HmmConstants.RAINING,
103 HmmConstants.SEE_UMBRELLA, 0.9);
104 sm.setSensingProbability(HmmConstants.RAINING,
105 HmmConstants.SEE_NO_UMBRELLA, 0.1);
106 sm.setSensingProbability(HmmConstants.NOT_RAINING,
107 HmmConstants.SEE_UMBRELLA, 0.2);
108 sm.setSensingProbability(HmmConstants.NOT_RAINING,
109 HmmConstants.SEE_NO_UMBRELLA, 0.8);
110
111 HiddenMarkovModel hmm = new HiddenMarkovModel(prior, tm, sm);
112
113 // hmm.setSensorModelValue(state,perception,p); given a state the
114 // probability of a perception is p
115
116 return hmm;
117
118 }
119
120 }