Academia.eduAcademia.edu

Figure 4 – uploaded by Syed Danial

See full PDFdownloadDownload figure
Figure 2.2. The general methodology for modeling landmarks-based route learning.

Figure 2 2. The general methodology for modeling landmarks-based route learning.

Related Figures (62)

On the realization of the Recognition-Primed Decision model for artificial agents
On the realization of the Recognition-Primed Decision model for artificial agents
or knowledge for the agent. An agent might be developed that uses certain strategies
or knowledge for the agent. An agent might be developed that uses certain strategies
Figure 1.2. Stages of a BDI agent from perceiving to situation recognition to decision- making. Note that the probability that the situation is really what the alarm means is based on proposed agent model. ‘igure 1.2. Stages of a BDI agent from perceiving to situation recognition to decision- emergency by exploiting dedicated escape routes. The capability in (2), is taken here
Figure 1.2. Stages of a BDI agent from perceiving to situation recognition to decision- making. Note that the probability that the situation is really what the alarm means is based on proposed agent model. ‘igure 1.2. Stages of a BDI agent from perceiving to situation recognition to decision- emergency by exploiting dedicated escape routes. The capability in (2), is taken here
Figure 2.1. The sample path of a typical CTMC. Source: (Adopted from Ajmone Marsan (1990).) depicted in Figure 2.1, where each horizontal line represents the duration of time the
Figure 2.1. The sample path of a typical CTMC. Source: (Adopted from Ajmone Marsan (1990).) depicted in Figure 2.1, where each horizontal line represents the duration of time the
Figure 2.4. GSPN model of landmark based route learning. Transitions ts - t17 are the stochastic transitions. All other transitions are immediate transitions. Transitions, tg, tio, ti2, tia, and tie, are executed when the decision is not to store a landmark and its associated navigation command, whereas, the transitions, ts, ti1, ti3, tis, and ti7, are used for saving the information about landmarks and associated navigation commands. The operator ++ is used to construct a multiset of tokens. The operator * is a binary operator in Snoopy. It takes a non-negative integer as a left operand that specifies the number of copies of the element provided as the right operand. The place Az3 has thirteen tokens. Notice that the current distribution of tokens as shown here constitutes the initial marking Mo.
Figure 2.4. GSPN model of landmark based route learning. Transitions ts - t17 are the stochastic transitions. All other transitions are immediate transitions. Transitions, tg, tio, ti2, tia, and tie, are executed when the decision is not to store a landmark and its associated navigation command, whereas, the transitions, ts, ti1, ti3, tis, and ti7, are used for saving the information about landmarks and associated navigation commands. The operator ++ is used to construct a multiset of tokens. The operator * is a binary operator in Snoopy. It takes a non-negative integer as a left operand that specifies the number of copies of the element provided as the right operand. The place Az3 has thirteen tokens. Notice that the current distribution of tokens as shown here constitutes the initial marking Mo.
Table 2.2. The stochastic transition rates of the GSPN model in Figure 2.4. The rates Q; and Q2 are taken at random from the range a, whereas Q3 and Q4 are randomly selected from near the lower-end and the upper-end values in the range a, respectively.
Table 2.2. The stochastic transition rates of the GSPN model in Figure 2.4. The rates Q; and Q2 are taken at random from the range a, whereas Q3 and Q4 are randomly selected from near the lower-end and the upper-end values in the range a, respectively.
Figure 2.5. Training exposure to participants Source: Adopted from (J. Smith, 2015). on offshore petroleum platform. The experiment involved 36 participants that formed learning and competence of human participants during emergency egress situations
Figure 2.5. Training exposure to participants Source: Adopted from (J. Smith, 2015). on offshore petroleum platform. The experiment involved 36 participants that formed learning and competence of human participants during emergency egress situations
Figure 2.6. (a) The route, R1, from worksite (S) in the engine room to the primary muster station (D). Decision points, pi, 1 < i < 13, with required navigation commands are shown. The route is scaled for better readability. (b) Description of landmarks at the decision points.
Figure 2.6. (a) The route, R1, from worksite (S) in the engine room to the primary muster station (D). Decision points, pi, 1 < i < 13, with required navigation commands are shown. The route is scaled for better readability. (b) Description of landmarks at the decision points.
Table 2.3. Difficulty level, y, associated with each decision points in R1.
Table 2.3. Difficulty level, y, associated with each decision points in R1.
the number of runs at 500 for samples S1 and S3, and at 5000 for samples S2, S4, S5,
the number of runs at 500 for samples S1 and S3, and at 5000 for samples S2, S4, S5,
Table 2.4. States in the CTMC of the GSPN model in Figure 2.4, obtained for the sample S4. States containing the places As—Ai. are shown for brevity. The actual markings involved unfolding (of the GSPN) due to the colored subnet N3. For clarity only A22 was replaced by its corresponding uncolored place in the markings below. The rest of the places are shown as they appeared in Figure 2.4. The steady state probability distribution for £ exists and is reported in Table 2.4. The unfolding (of the GSPN) due to the colored subnet N3. For clarity only Az2 was replaced by
Table 2.4. States in the CTMC of the GSPN model in Figure 2.4, obtained for the sample S4. States containing the places As—Ai. are shown for brevity. The actual markings involved unfolding (of the GSPN) due to the colored subnet N3. For clarity only A22 was replaced by its corresponding uncolored place in the markings below. The rest of the places are shown as they appeared in Figure 2.4. The steady state probability distribution for £ exists and is reported in Table 2.4. The unfolding (of the GSPN) due to the colored subnet N3. For clarity only Az2 was replaced by
a navigator’s state of remembering or forgetting a landmark with associated direction
a navigator’s state of remembering or forgetting a landmark with associated direction
Figure 3.4. (a) A schematic of the primary escape route (R1) in AVERT simulator from the worksite to the muster stations, (b) description of the landmarks near the decision-points (DP), pi, there are fourteen decision-points. The distances are not scaled but are presented for better readability. two (S2) was designed to train and test the participants for emergency alarm recognition
Figure 3.4. (a) A schematic of the primary escape route (R1) in AVERT simulator from the worksite to the muster stations, (b) description of the landmarks near the decision-points (DP), pi, there are fourteen decision-points. The distances are not scaled but are presented for better readability. two (S2) was designed to train and test the participants for emergency alarm recognition
Table 3.1. A sample of data from four scenarios LE4, TE2, TE4, and LA1 with 10 participants tagged as CAG1, CSG1,..., MWGI. Each column represents the data for each decision-points pi, ..., pi4. A ‘1’ shows that the participant has made the right decision at the respective decision-point and a ‘0’ means that the participant did not follow the required action. Each entry contains four values, for example, for the participant CAG1, the values for p; are 0011 where the first 0 is for LE4, the second 0 is for TE2, the third value ‘1’ is for TE4. and the last value is recorded from the scenario LA1.
Table 3.1. A sample of data from four scenarios LE4, TE2, TE4, and LA1 with 10 participants tagged as CAG1, CSG1,..., MWGI. Each column represents the data for each decision-points pi, ..., pi4. A ‘1’ shows that the participant has made the right decision at the respective decision-point and a ‘0’ means that the participant did not follow the required action. Each entry contains four values, for example, for the participant CAG1, the values for p; are 0011 where the first 0 is for LE4, the second 0 is for TE2, the third value ‘1’ is for TE4. and the last value is recorded from the scenario LA1.
Figure 3.6. The average number of participants failed to follow the navigation commands corresponding to decision-points pi-ps6. Percentage values are shown on the y-axis. that also goes directly near the mess hall. Since po is considered as the highest
Figure 3.6. The average number of participants failed to follow the navigation commands corresponding to decision-points pi-ps6. Percentage values are shown on the y-axis. that also goes directly near the mess hall. Since po is considered as the highest
Figure 3.7. The average number of participants failed to follow the navigation commands corresponding to decision-points p7-p14. Percentage values are shown on the y-axis.
Figure 3.7. The average number of participants failed to follow the navigation commands corresponding to decision-points p7-p14. Percentage values are shown on the y-axis.
Figure 3.8. A method to apply the GSPNRL model for a sequential route learning task. The model inputs are (a) the landmarks (regarding difficulty levels denoted by y)
Figure 3.8. A method to apply the GSPNRL model for a sequential route learning task. The model inputs are (a) the landmarks (regarding difficulty levels denoted by y)
along R1 near each decision point, and (b) the navigation commands required at each
along R1 near each decision point, and (b) the navigation commands required at each
Table 3.3. Initial difficulty levels assigned to each decision-points in R1.
Table 3.3. Initial difficulty levels assigned to each decision-points in R1.
Figure 3.9. Simulated results for the scenario LE4. Represents an agent that models remembering/forgetting the decision-points and the associated navigation commands. (A) The boxplots shown here are based on the percentage of times a decision-point is remembered during 5000 runs of the simulation of LE4. (B) The boxplots shown here are based on the percentage of times a decision-point is forgotten during 5000 runs of the simulation of LE4. or forget a decision-point. Moreover, in all cases where a decision-point is
Figure 3.9. Simulated results for the scenario LE4. Represents an agent that models remembering/forgetting the decision-points and the associated navigation commands. (A) The boxplots shown here are based on the percentage of times a decision-point is remembered during 5000 runs of the simulation of LE4. (B) The boxplots shown here are based on the percentage of times a decision-point is forgotten during 5000 runs of the simulation of LE4. or forget a decision-point. Moreover, in all cases where a decision-point is
Figure 3.10. Simulated results for the scenario TE4. Represents an agent that models remembering/forgetting the decision-points and the associated navigation commands. (A) The boxplots shown here are based on the percentage of times a decision-point is remembered during 5000 runs of the simulation of TE4. (B) The boxplots shown here are based on the percentage of times a decision-point is forgotten during 5000 runs of the simulation of TE4. scenario, i.e., LE4. After every one or two simulations, the difficulty levels associated
Figure 3.10. Simulated results for the scenario TE4. Represents an agent that models remembering/forgetting the decision-points and the associated navigation commands. (A) The boxplots shown here are based on the percentage of times a decision-point is remembered during 5000 runs of the simulation of TE4. (B) The boxplots shown here are based on the percentage of times a decision-point is forgotten during 5000 runs of the simulation of TE4. scenario, i.e., LE4. After every one or two simulations, the difficulty levels associated
Figure 3.11. Simulated results for the scenario LA4. Represents an agent that models remembering/forgetting the decision-points and the associated navigation commands. (A) The boxplots shown here are based on the percentage of times a decision-point is remembered during 5000 runs of the simulation of LA4. (B) The boxplots shown here are based on the percentage of times a decision-point is forgotten during 5000 runs of the simulation of LA4.
Figure 3.11. Simulated results for the scenario LA4. Represents an agent that models remembering/forgetting the decision-points and the associated navigation commands. (A) The boxplots shown here are based on the percentage of times a decision-point is remembered during 5000 runs of the simulation of LA4. (B) The boxplots shown here are based on the percentage of times a decision-point is forgotten during 5000 runs of the simulation of LA4.
Figure 3.12. Simulated results for the scenario TA4. Represents an agent that models remembering/forgetting the decision-points and the associated navigation commands. (A) The boxplots shown here are based on the percentage of times a decision-point is remembered during 5000 runs of the simulation of TA4. (B) The boxplots shown here are based on the percentage of times a decision-point is forgotten during 5000 runs of the simulation of TA4.
Figure 3.12. Simulated results for the scenario TA4. Represents an agent that models remembering/forgetting the decision-points and the associated navigation commands. (A) The boxplots shown here are based on the percentage of times a decision-point is remembered during 5000 runs of the simulation of TA4. (B) The boxplots shown here are based on the percentage of times a decision-point is forgotten during 5000 runs of the simulation of TA4.
Figure 3.13. The average number of participants failed to follow the navigation commands corresponding to the decision-points p1-ps are represented by black filled bars. Simulation results corresponding to each scenario are shown by unfilled rectangular bars. Percentage values are shown on the y-axis. a minimum. As an alternative to this approach, one may consider changing the rates
Figure 3.13. The average number of participants failed to follow the navigation commands corresponding to the decision-points p1-ps are represented by black filled bars. Simulation results corresponding to each scenario are shown by unfilled rectangular bars. Percentage values are shown on the y-axis. a minimum. As an alternative to this approach, one may consider changing the rates
Figure 3.14. The average number of participants failed to follow the navigation commands corresponding to the decision-points p7-p14 is depicted with black rectangular bars. The unfilled bars show the simulated results. Percentage values are shown on the y-axis.
Figure 3.14. The average number of participants failed to follow the navigation commands corresponding to the decision-points p7-p14 is depicted with black rectangular bars. The unfilled bars show the simulated results. Percentage values are shown on the y-axis.
Figure 3.15. The derivatives of steady-state probabilities (that is, the probability that the output place of the transition tn, the place A,-3, has one token) concerning small changes in the parameter 0 are shown. The parameters 01, 02, 03, 04, and 5 are the rates of the stochastic transition tg, tio, tiz, tia, and tie respectively. The parameter values, that is, the values of the rates, are taken in the ranges defined
Figure 3.15. The derivatives of steady-state probabilities (that is, the probability that the output place of the transition tn, the place A,-3, has one token) concerning small changes in the parameter 0 are shown. The parameters 01, 02, 03, 04, and 5 are the rates of the stochastic transition tg, tio, tiz, tia, and tie respectively. The parameter values, that is, the values of the rates, are taken in the ranges defined
Table 4.1. Variable/predicate names and description predicate holds is estimated by employing an inference mechanism, such as by using
Table 4.1. Variable/predicate names and description predicate holds is estimated by employing an inference mechanism, such as by using
Table 4.2. The FOL rules that are showing the knowledge base for basic emergency preparedness. fire into a larger fire that obstructs the primary escape-route leading to the primary
Table 4.2. The FOL rules that are showing the knowledge base for basic emergency preparedness. fire into a larger fire that obstructs the primary escape-route leading to the primary
empirical data. The objective of Smith’s experiment was to assess VE training effects
empirical data. The objective of Smith’s experiment was to assess VE training effects
Figure 4.3. Floor map for decks A and C in AVERT simulator. A participant starts from Cabin (S) in part (1) and ends either at the mess hall or the lifeboat station in part (2) using external stairwell or main stairwell. The dotted lines show the alternate route, and the solid lines refer to the primary route. followed by another PA. Initially, a participant is situated in their cabin (see the floot
Figure 4.3. Floor map for decks A and C in AVERT simulator. A participant starts from Cabin (S) in part (1) and ends either at the mess hall or the lifeboat station in part (2) using external stairwell or main stairwell. The dotted lines show the alternate route, and the solid lines refer to the primary route. followed by another PA. Initially, a participant is situated in their cabin (see the floot
Figure 4.4. Portion of ground MN obtained using grounding of the predicates in rules 2-5. Querying the proposed MLN based model of agent SA is the same as querying a
Figure 4.4. Portion of ground MN obtained using grounding of the predicates in rules 2-5. Querying the proposed MLN based model of agent SA is the same as querying a
Table 4.4. Weights assigned to rules using datasets D1 and D2. Only 12 out of a total of 59 ground rules obtained by different groundings of the rules in Table 4.2 are shown for brevity. knowledgebase. We use the MC-SAT algorithm using the Alchemy inference engine
Table 4.4. Weights assigned to rules using datasets D1 and D2. Only 12 out of a total of 59 ground rules obtained by different groundings of the rules in Table 4.2 are shown for brevity. knowledgebase. We use the MC-SAT algorithm using the Alchemy inference engine
Table 4.5. Query results. The symbol ‘—’ is the logical not operator. A predicate followed by a symbol ‘—’ has a truth value of false, otherwise true. The column for empirical results contains the results obtained from participants. Corresponding to each empirical result is a probability the model generated for that predicate. For example, R(P1G1, GPA, 10) meaning that the participant P1G1 has recognized the GPA alarm during time interval t0. The probability that this predicate R(P1G1, GPA, 10) is true is 0.91. Notice that the constant EVAC means EVACUATE. in the following predicates.
Table 4.5. Query results. The symbol ‘—’ is the logical not operator. A predicate followed by a symbol ‘—’ has a truth value of false, otherwise true. The column for empirical results contains the results obtained from participants. Corresponding to each empirical result is a probability the model generated for that predicate. For example, R(P1G1, GPA, 10) meaning that the participant P1G1 has recognized the GPA alarm during time interval t0. The probability that this predicate R(P1G1, GPA, 10) is true is 0.91. Notice that the constant EVAC means EVACUATE. in the following predicates.
decision-makers plan, in the event of an emergency, to mitigate the aftereffects or to Figure 5.1. Integrated G. Klein’s RPD model. Source (G. Klein, 1998, p. 27). centered artificial intelligence (AI). RPDM (see Figure 5.1) explains how humar
decision-makers plan, in the event of an emergency, to mitigate the aftereffects or to Figure 5.1. Integrated G. Klein’s RPD model. Source (G. Klein, 1998, p. 27). centered artificial intelligence (AI). RPDM (see Figure 5.1) explains how humar
more information for the recognition of the situation, but the same mechanism, BN,
more information for the recognition of the situation, but the same mechanism, BN,
operates, a module to performs mental simulation as a cause-and- effect mechanism
operates, a module to performs mental simulation as a cause-and- effect mechanism
Figure 5.3. Activities in the process of developing a realization of RPDM. the environment to having a decision for what needs to be done when a situation
Figure 5.3. Activities in the process of developing a realization of RPDM. the environment to having a decision for what needs to be done when a situation
model considers this diverse nature of experiences among experts by generally
model considers this diverse nature of experiences among experts by generally
where w is the weight assigned to the rule. The operator ‘=>’ is for logical implication. probabilities as shown in Table 5.3, provided w > 0. Here Z is the partition function as described in Section 5.2.2. The probability for a world to be true depends on the
where w is the weight assigned to the rule. The operator ‘=>’ is for logical implication. probabilities as shown in Table 5.3, provided w > 0. Here Z is the partition function as described in Section 5.2.2. The probability for a world to be true depends on the
predicate appearing in a formula. An edge between two nodes means that the Figure 5.4. The nodes in Figure 5.4 are obtained for each possible grounding of each
predicate appearing in a formula. An edge between two nodes means that the Figure 5.4. The nodes in Figure 5.4 are obtained for each possible grounding of each
ontology for offshore emergency situations.
ontology for offshore emergency situations.
Table 5.4. Variables (and corresponding predicate names) to be used in the ELDS modul development along with parameter types and description are shown. proceeded in a scenario towards making a required decision. Factors that play important roles in deciding about the kind of emergency (FIRE or EVACUATE),
Table 5.4. Variables (and corresponding predicate names) to be used in the ELDS modul development along with parameter types and description are shown. proceeded in a scenario towards making a required decision. Factors that play important roles in deciding about the kind of emergency (FIRE or EVACUATE),
Table 5.5. A sample of empirical observations for the decision choices made by the five participants. The time interval t 0 starts with the beginning of the scenario until the GPA alarm ends. The interval t1 starts when the PAPA alarm begins sounding, i.e., just after the GPA ends and the situation escalates from FIRE to EVACUATE. It ends at the end of the scenario.
Table 5.5. A sample of empirical observations for the decision choices made by the five participants. The time interval t 0 starts with the beginning of the scenario until the GPA alarm ends. The interval t1 starts when the PAPA alarm begins sounding, i.e., just after the GPA ends and the situation escalates from FIRE to EVACUATE. It ends at the end of the scenario.
able 5.6. The evidence/test data collected as 20% of the empirical observations. Columns with predicate names having a preceding ‘?’ contait nulated results, which are the probabilities these predicates are true given the evidence data. a) M1 is short for messhall, the primary muster station, and M2 is short for the alternate muster station, the lifeboat station. (b) True and false are the empirically observed state: or the predicates in each column. (c) A predicate starting with a ‘?’ is the one that has been queried in the ELDS module. The remaining predicates are used as evidence/test dat 1 the query process. ELDS module that comprises MLN is queried by employing the MC-SAT inference algorithm. (d) The results of querying for R, HITR, and HES are reporte s the probability that the predicates are true given the evidence data. These probabilities are reported as values in parenthesis under the respective columns. A probability valu vith a subscript M1 stands for the result related with the primary muster station, whereas the one having a subscript M2 refers to the probability that the predicate is true involvins 1e alternate muster station. (e) H1 refers to FIRE emergency, H2 refers to EVACUATE emergency. The probability that a FIRE emergency is occurred is referred to by nui, ani 1e probability that an EVACUATE emergency has occurred is referred to by nu2, where0 <n <1.
able 5.6. The evidence/test data collected as 20% of the empirical observations. Columns with predicate names having a preceding ‘?’ contait nulated results, which are the probabilities these predicates are true given the evidence data. a) M1 is short for messhall, the primary muster station, and M2 is short for the alternate muster station, the lifeboat station. (b) True and false are the empirically observed state: or the predicates in each column. (c) A predicate starting with a ‘?’ is the one that has been queried in the ELDS module. The remaining predicates are used as evidence/test dat 1 the query process. ELDS module that comprises MLN is queried by employing the MC-SAT inference algorithm. (d) The results of querying for R, HITR, and HES are reporte s the probability that the predicates are true given the evidence data. These probabilities are reported as values in parenthesis under the respective columns. A probability valu vith a subscript M1 stands for the result related with the primary muster station, whereas the one having a subscript M2 refers to the probability that the predicate is true involvins 1e alternate muster station. (e) H1 refers to FIRE emergency, H2 refers to EVACUATE emergency. The probability that a FIRE emergency is occurred is referred to by nui, ani 1e probability that an EVACUATE emergency has occurred is referred to by nu2, where0 <n <1.
Figure 5.7. A model of mental simulation during deliberation of the plan of moving to the LIFEBOAT station. The agent weighs its chances of being trapped for each case of choosing PER and SER. Figure 5.7. A model of mental simulation during deliberation of the plan of moving to the LIFEBOAT station. The agent weighs its chances of being trapped for each case of choosing
Figure 5.7. A model of mental simulation during deliberation of the plan of moving to the LIFEBOAT station. The agent weighs its chances of being trapped for each case of choosing PER and SER. Figure 5.7. A model of mental simulation during deliberation of the plan of moving to the LIFEBOAT station. The agent weighs its chances of being trapped for each case of choosing
Figure B.1. A portion of the floor map of A-deck, and C-deck (accommodation block). The primary escape route is shown with arrows pointing towards the main stairwell from the cabin. The participant has to go down two levels from the cabin to A-deck, where the muster stations are located. Hazards at different locations are shown to illustrate an emergency. work in the current scenario.
Figure B.1. A portion of the floor map of A-deck, and C-deck (accommodation block). The primary escape route is shown with arrows pointing towards the main stairwell from the cabin. The participant has to go down two levels from the cabin to A-deck, where the muster stations are located. Hazards at different locations are shown to illustrate an emergency. work in the current scenario.
Related topics:
Computer ScienceMemorial University of Newfoundland
580 California St., Suite 400
San Francisco, CA, 94104
© 2025 Academia. All rights reserved