"... In PAGE 8: ... { 20-SIM ( quot;Twente Sim quot;) is a modeling and simulation program developed at the UniversityofTwente [4]. Table1 shows that both time and state events are typically handled by these packages, though the implementation mayvary [23]. Also, they facilitate adding and removing equations that do not change causality.... ..."

"... In PAGE 14: ... 4.1 GPGP Simulation: Issues Our model of an abstract task environment, used in these experiments, has ten parameters; Table1 lists them and the values used in the experiments described in the next two sections.5 5Our earlier work focussed on the analysis of distributed sensor network task environments [6, 7].... In PAGE 15: ...1 .5 .9 .5 Table1 : Environmental Parameters used to generate the random episodes The primary sources of overhead associated with the coordination mechanisms include action executions (communication and information gathering), calls to the local scheduler, and any algorithmic overhead associated with the mechanism itself. Table 2 summarizes the total amount of overhead from each source for each coordination mechanism setting and the coordination substrate.... In PAGE 15: ... L represents the length of processing (time before termination), and d is a general density measure of coordination relationships. We believe that all of these amounts can be derived from the environmental parameters in Table1 , they can also be measured experimentally. Interactions between the presence of coordination mechanisms and these quantities include: the number of methods or tasks in E, which depends on the non- local view mechanism; the number of coordination relationships jCRj or the subsets RCR (redundant coordination relationships), HPCR (hard predecessor coordination relationships), SPCR (soft predecessor coordination relationships), which depends on the number of tasks and methods as well; and the number of commitments jCj, which depends on each of the three mechanisms that makes commitments.... In PAGE 16: ... Parallel refers to the centralized reference scheduler that uses a heuristic parallelization of an optimal single agent schedule using a complete global view. The experiments were based on the same environmental parameters as the facilitation experiments ( Table1 ). There are several important things to note about this class of environments: The size of the episodes was kept artificially small so that the centralized reference scheduler could find an optimal schedule in a reasonable amount of run time.... In PAGE 18: ... Here we give an example with respect to the soft coordi- nation mechanism (Mechanism 5), which will make commitments to facilitation relationships. We ran 234 randomly generated episodes (generated with the environmental parameters shown in Table1 ) with four agents both with and without the soft coordination mechanism. Because the variance between these randomly generated episodes is so great, we took advantage of the paired response nature of the data to run a non-parametric Wilcoxon matched-pairs signed- ranks test [3].... ..."

"... In PAGE 14: ... 4.1 GPGP Simulation: Issues Our model of an abstract task environment, used in these experiments, has ten parameters; Table1 lists them and the values used in the experiments described in the next two sections.5 5Our earlier work focussed on the analysis of distributed sensor network task environments [6, 7].... In PAGE 15: ...1 .5 .9 .5 Table1 : Environmental Parameters used to generate the random episodes The primary sources of overhead associated with the coordination mechanisms include action executions (communication and information gathering), calls to the local scheduler, and any algorithmic overhead associated with the mechanism itself. Table 2 summarizes the total amount of overhead from each source for each coordination mechanism setting and the coordination substrate.... In PAGE 15: ... L represents the length of processing (time before termination), and d is a general density measure of coordination relationships. We believe that all of these amounts can be derived from the environmental parameters in Table1 , they can also be measured experimentally. Interactions between the presence of coordination mechanisms and these quantities include: the number of methods or tasks in E, which depends on the non- local view mechanism; the number of coordination relationships jCRj or the subsets RCR (redundant coordination relationships), HPCR (hard predecessor coordination relationships), SPCR (soft predecessor coordination relationships), which depends on the number of tasks and methods as well; and the number of commitments jCj, which depends on each of the three mechanisms that makes commitments.... In PAGE 16: ... Parallel refers to the centralized reference scheduler that uses a heuristic parallelization of an optimal single agent schedule using a complete global view. The experiments were based on the same environmental parameters as the facilitation experiments ( Table1 ). There are several important things to note about this class of environments: The size of the episodes was kept artificially small so that the centralized reference scheduler could find an optimal schedule in a reasonable amount of run time.... In PAGE 18: ... Here we give an example with respect to the soft coordi- nation mechanism (Mechanism 5), which will make commitments to facilitation relationships. We ran 234 randomly generated episodes (generated with the environmental parameters shown in Table1 ) with four agents both with and without the soft coordination mechanism. Because the variance between these randomly generated episodes is so great, we took advantage of the paired response nature of the data to run a non-parametric Wilcoxon matched-pairs signed- ranks test [3].... ..."

"... In PAGE 14: ... 4.1 GPGP Simulation: Issues Our model of an abstract task environment, used in these experiments, has ten parameters; Table1 lists them and the values used in the experiments described in the next two sections.5 5 Our earlier work focussed on the analysis of distributed sensor network task environments [6, 7].... In PAGE 15: ...1 .5 .9 .5 Table1 : Environmental Parameters used to generate the random episodes The primary sources of overhead associated with the coordination mechanisms include action executions (communication and information gathering), calls to the local scheduler, and any algorithmic overhead associated with the mechanism itself. Table 2 summarizes the total amount of overhead from each source for each coordination mechanism setting and the coordination substrate.... In PAGE 15: ... L represents the length of processing (time before termination), and d is a general density measure of coordination relationships. We believe that all of these amounts can be derived from the environmental parameters in Table1 , they can also be measured experimentally. Interactions between the presence of coordination mechanisms and these quantities include: the number of methods or tasks in E, which depends on the non- local view mechanism; the number of coordination relationships jCRj or the subsets RCR (redundant coordination relationships), HPCR (hard predecessor coordination relationships), SPCR (soft predecessor coordination relationships), which depends on the number of tasks and methods as well; and the number of commitments jCj, which depends on each of the three mechanisms that makes commitments.... In PAGE 16: ... Parallel refers to the centralized reference scheduler that uses a heuristic parallelization of an optimal single agent schedule using a complete global view. The experiments were based on the same environmental parameters as the facilitation experiments ( Table1 ). There are several important things to note about this class of environments: #0F The size of the episodes was kept artificially small so that the centralized reference scheduler could find an optimal schedule in a reasonable amount of run time.... In PAGE 18: ... Here we give an example with respect to the soft coordi- nation mechanism (Mechanism 5), which will make commitments to facilitation relationships. We ran 234 randomly generated episodes (generated with the environmental parameters shown in Table1 ) with four agents both with and without the soft coordination mechanism. Because the variance between these randomly generated episodes is so great, we took advantage of the paired response nature of the data to run a non-parametric Wilcoxon matched-pairs signed- ranks test [3].... ..."

"... In PAGE 10: ... We also measured the average and median time used to create query images using the different combinations. Table2 shows the overall tool use for query types, while Figure 4 shows the distribution of the various combinations used to create the query images, e.g.... ..."

"... In PAGE 4: ...send messages that provide feedback about user actions. A proposed set of messages is shown in Table1 . To facilitate reuse, any message that can be sent by a tool should be understood by a tutor, and any message that can be sent by a tutor should be under- stood by a tool.... ..."

"... In PAGE 56: ... Table 12 Outcomes of tool demonstrations in design, consulting and regulating organisations Potential uses Identified qualities Providing preliminary analysis to HAZOP Providing education to inexperienced designers Providing safety training for staff who work on operating sites Supporting the design of work systems Supporting reviews with checklists Supporting engineering cases to improve designs in negotiations with project managers Favourable Augments experience base of the organisation with failures that by chance it has not encountered Raises awareness of operator assumptions and operability issues Users would naturally test some of the assumptions without prompting Provides information about operator behaviour that operators would be reluctant to report to designers Unfavourable Leaves some concepts open to interpretation Employs cases that will be perceived as being irrelevant due to their origins in other industries Omits data on the frequency with which the assumptions contribute to accidents Fails to provide positive instructions on what to do Neglects cost of implied action that designer has to carry out Needs to encouragee users to find their own cases to represent particular concepts Fails to reflect the fact that equipment design is constrained by legislation anyway Needs an expert facilitator to use effectively Assumptions list lacks specificity Table13 shows the observations that were made of a meeting in which five staff used the tool to examine the operation of loading a hazardous material. This served as a case study.... In PAGE 57: ...The group was debriefed but the observations made in the debriefing are not reproduced as they mostly repeated the observations recorded in Table13 . There were, however, some additional observations.... ..."