Performance Measurement of DEVS Dynamic Structure on Forest Fire Spreading Simulation

Size: px
Start display at page:

Download "Performance Measurement of DEVS Dynamic Structure on Forest Fire Spreading Simulation"

Transcription

1 Performance Measurement of DEVS Dynamic Structure on Forest Fire Spreading Simulation Yi Sun, Xiaolin Hu Department of Computer Science Georgia State University Abstract---Traditional simulation methodology supports only changes in models state variables. Dynamic structure modeling allows to model both changes in state variables and changes in structure and has been applied to different applications. This paper focuses on the performance issue of dynamic structure implemented in the DEVS modeling and simulation environment. Six performance measurement metrics are defined. Measurement results and analysis are provided to evaluate the advantages and limitations of dynamic structure modeling in the context of forest fire spreading simulation. Index Terms DEVS, dynamic structure, forest fire spreading, performance metrics I. INTRODUCTION The motivation of dynamic structure models comes from the requirement of some applications, e.g., sociological and ecological applications. These application systems need dynamic structure capability to meet with structural changes in modeling and simulation. Generally, dynamic structure is useful for modeling living autonomous systems that change their interactions, compositions and behavior patterns to adapt to their environments, or those self-organizing, selfreconfiguring engineered systems, where the structures of the systems adapt to changed requirements. Many research on dynamic structure modeling have been based on the DEVS (Discrete Event Systems Specification) [2] modeling and simulation framework. DEVS has been introduced to support component-based modeling and simulation by emphasizing the theory of hierarchical modular modeling. In DEVS-based environment such as DEVSJAVA [3], a component is a model with clear defined interfaces. DEVS is a component system consisting of individual components and by establishing relationships among them. Such system has high degree of autonomy with well-defined interfaces, so that dynamic structure of components can be achieved during runtime. Barros in [5] presented the formalism on extended dynamic structure DEVS whose basic models are the same as classic DEVS basic models, but the structure of coupled models can change over time. The full and complete semantics for dynamic structure were added in classic DEVS. The dynamic capability supports changes ranging from simple linking relations between models, to more complex changes in model composition such as adding and deleting models during simulation. Dynamic structure DEVS provides several advantages: a) It provides a natural and effective way to model those complex systems which exhibit structure and behavior changes to adapt to different situations. Structure changing and component upgrading is an essential part of these systems. Without the dynamic structure capability, it s very hard, if not impossible, to model and simulate them. b) From the design point of view, dynamic structure provides the additional flexibility to design and analyze a system under development. c) Dynamic structure makes it possible to load only a sub-set of system s components for simulation. This is very useful to simulate very large systems with tremendous number of components, as only the active components need to be loaded dynamically to conduct the simulation. Otherwise, the entire system has to be loaded before the simulation begins. For example, as will be illustrated later, dynamic structure allows us to load only part of a cell space in forest fire spreading simulation. Dynamic structure has been applied to different applications. In [4], Dynamic Structure DEVS (DSDEVS) proved to be able to model a complex adaptive computer architecture. The work of [6] applied the DSDEVS to an example of fire spread simulation. In [7], the authors used dynamics structure to support the development of a dynamic team formation of robots. None of these works have explicitly measure the performance of dynamic structure modeling and simulation. In this paper, we discuss the dynamic structure implementation in DEVSJAVA environment and emphasize on its performance analysis by forest fire spreading experimentation. The structure of remaining paper is organized as follows. First, performance analysis and measurement metrics are presented in section II. Then in section III, the forest fire spreading experiment implementation is given and the experiment results related with performance metrics in section II are discussed in section IV. Finally, the conclusions and future work are presented in section V. II. PERFORMANCE ANALYSIS AND MEASUREMENT METRICS To develop performance measurement metrics for dynamic structure modeling and simulation, we start with the general simulation protocol as implemented by the coordinator in DEVSJAVA. Models.construction(); Models.initialization(); While (stop condition is not met){ simulators.askall( nexttn ) tn = compareandfindtn(); simulators.tellall("computeout,tn)

2 simulators.tellall("sendout") simulators.tellall("applydelt,tn) The simulation process first executes model construction and initialization. We call these two parts as initialization in later sections. Then the simulation goes into a while loop until simulation stops. In each while loop cycle (also referred to as simulation step in the following text), the simulators that are involved in this cycle go through the operations of nexttn(), computerout(), sendout(), and applydelt() following the semantics of DEVS models. Note that this simulation protocol applies to both dynamic structure (DS) and non-dynamic structure (non-ds). The difference between the two cases is that the number of simulators that are involved in the simulation cycle is different. For non-ds, the number of simulators is fixed; for DS, the number of simulators dynamically changes, due to the fact that models are dynamically added and deleted. Based on the above protocol, the total execution time can be counted as the time of initialization plus that of executing the while loop cycles. The initialization time depends on how many models need to be constructed and initialized before simulation starts. The execution time for the while loop depends on how many active models execute functions during the simulation cycles. To give a formal analysis about the execution time of DS model and non-ds model, a general formula is given below: T = T initial + Ti where, T is the total execution time. T initial includes the model construction and initialization time and T i is the execution time of one cycle; m_step is the total simulation cycles. The two parts shown in Formula () are related to DS and non-ds models in different ways. Consider the cellular spacebased forest fire spreading simulation (as described in Section III), where DS implementation uses a cellspacemanager to add/remove forest cells. For this application, initially there is only one component cellspacemanager and no forest cell in the cell space. Thus T initial is cellspacemanager s initial time. T i is the execution time for cycle i, which is the sum of the execution time for all the active cells plus overhead caused by adding/deleting cells at that cycle. So, T initial = t initial-of-cellspacemanager (2) Ti n ( tij + T overhead _ i j = ) In formula (3), tij is cycle execution time of an active cell. n is the number of active cells. T overhead_i consists of initialization of added cells and coupling/decoupling of added and deleted cells. The more number of added and deleted cells is, the larger T overhead_i is. For non-ds model, T initial is all cells initialization time. The cycle execution time includes active cells execution time and non-active cells execution time. So, the formulas are: T initial = N * t initial-of-cell (4) Ti = n j tij N n + j tij () (3) ' (5) In formula (5), tij is cycle execution time of an active cell; tij is cycle execution time of a non-active cell; N is the total number of cells in the cell space. Generally, tij < tij. For example, in the forest fire spreading simulation example described in section III, the value of tij / tij in is around /5 to 2/5. We note that for a particular cycle, the collection of active cells in DS and non-ds implementation is the same. Based on formulas (2), (3), (4) and (5), we get the execution time difference of non-ds and DS: T original - T DS = N * t initial-of-cell - t initial-of-cellspacemanager N n ( tij T overhead _ i j + ' ) Formula (6) shows that a major part of the difference of two models lies in non-ds model s non-active cells cycle execution time and DS model s overhead for adding/deleting cells. When the number of adding/deleting cells is small, the overhead is also small, thus the DS model s execution time performance is better than non-ds model; otherwise the DS model s execution time performance may be worse than that of the non-ds model. Based on the theoretical analysis above, some general observations can be made for the application of cellular spacebased forest fire spreading simulation. In general, we note that compared with the non-ds model, the DS implementation not only decreases execution time but also reduces memory utilization. On one hand, the DS forest fire model decreases execution time. The non-ds model creates all cells at the beginning of simulation, while the DS model only lets active cells exist during simulation. This improves simulation speed at least in two aspects: ) The dynamic structure saves time in the beginning. It does not necessarily create all cells at one time; furthermore, the average number of cells used in DS is much less than non-ds model. This advantage is more when the size of cell space becomes larger, e.g., the 6 * 6 cell space size vs. 2 * 2 cells space size. 2) The dynamic structure saves time during simulation. Since all cells exist in non-ds model, they need to execute internal, external, and output functions; but in DS model, only active (e.g., front fire burning) cells execute these functions, it saves passive cells execution time. The overhead of DS is the time of repeated adding/deleting cells and coupling connections. If non-ds model has a large number of cells to execute internal, external and output functions and DS model s overhead is small, DS is superior to non-ds model. On the other hand, the DS model reduces memory utilization. In DS model, only active (e.g. burning) cells exist in memory during whole simulation, while in non-ds model, all cells occupy memory all the time. So DS model s memory utilization is always less than non-ds model. To further support the above analysis, we develop the following six measurement metrics to compare the performance of DS and non-ds models. Experimental results are collected and presented in Section IV based on these metrics. Initialization time for different cellular space size Execution time for different cellular space size Execution time at different simulation stages Memory usage (6)

3 Execution time for different model behaviors (different # of points) Execution time for onedcoord, DS and non-ds models These metrics are developed based on two important factors for measuring simulation performance. One is execution time, and the other is memory utilization. The first metric is to compare initialization time of different cellular space size. The second metric is comparing execution time between non-ds and DS models on different cellular space size. This metric analyzes how cell space sizes affect DS performance. The third metric is detailed execution time analysis about non-ds model. In this metric, the total execution time, overhead, and added/deleted cells are presented and their relationship is pointed out. The fourth metric compares memory use of non- DS and DS models. The fifth metrics measures the execution time for different model behaviors, i.e., different types of models or the same model with different simulation scenarios. This measurement aims to show that the performance advantage/disadvantage of DS is really application-dependent. While some applications favor the DS, other may turn against the DS from the performance point of view. All these five metrics show DS model brings performance advantage by reducing the number of active cells in minimum. However, overhead is imported. In order to avoid the overhead of dynamically adding/removing cells but in the meantime to make only the active cells participate the simulation cycles, an improved simulation engine is developed in [8]. This simulation engine onedcoord uses a minseltree data structure to sort and find the active cells and then only those active cells go through the simulation cycles. So in the sixth metric, we compare onedcoord with DS and non-ds models. III. FOREST FIRE MODEL AND DYNAMIC STRUCUTRE IMPLEMENTATION In this section, we present our DS simulation for forest fire spreading model. Before that, a brief description of original non-ds model is presented. In original non-ds forest fire model, a forest is modeled as a two-dimensional cell-space composed of individual forest cells coupled together according to their relative physical geometric locations. Each cell is modeled as a twodimensional cell. Specifically, each cell has the following six states: unburned, burning, burned, unburned-wet, burning-wet, and burned-wet. Conditions and rules are defined to govern the state transition of a cell. In the two-dimensional cell space model, each cell has eight neighbor cells. Accordingly, for each cell, eight fixed fire spreading directions are defined. Fire spread in each cell is modeled using Rothermel s [9] stationary model, which is a one-dimension semi-empirical model. To obtain the second dimension, a propagation algorithm that uses maximum rate of spread and wind and slope factors is applied. During simulation, the behavior of a burning cell can be influenced by external inputs from neighboring cells as well as changes in weather conditions. In addition, uncertainty is incorporated in the model by allowing certain critical parameters to be sampled from arbitrary probability distributions during the simulation run. A detailed description of this model and the initial condition of simulation can be found in []. In non-ds model, all forest cells are created at the beginning of the simulation. This adds a lot of computation time for a cell space with thousands of cells. In order to avoid this, DS model creates active cells during simulation. Once a forest cell goes from an active state to a passive state, it is removed from the cell space. This approach only keeps the forest cells along the fire front. To achieve the implementation of the dynamic structure feature, we use DEVSJAVA and take advantage of its capability of dynamical adding/deleting components. In implementation, a DynamicCellSpaceManager model is used be responsible for structure change. It executes external function when there are requests from ForestCell. Part of the external method is shown as below. if (messageonport("inigniter",)) { getmessageonport( inigniter ); create and add fc object; addcouplings; holdin("ignite",.); if (messageonport( "inburning")) { getmessageonport( inburning ); create and add neighbor objects; addcouplings; holdin("added",.); if (messageonport("inburned")) { getmessageonport( inburned ); removemodel(cell.message); holdin("removed",); In ForestCell, two output ports outburning and outburned are defined. These two ports connect to DynamicCellSpaceManager s two input ports, inburning and inburned, separately. When an ForestCell burns, it sends out an adding message via its outburning port to DynamicCellSpaceManager s inburning port. In turn the DynamicCellSpaceManager adds its neighbor cells and adds coupling connections. When a cell is being burned, it sends out a delete message via its outburned port to DynamicCellSpaceManager, which deletes this cell from CellSpace. The process continues until the simulation ends. Since only active (burning) cells are considered in the cell space, this approach has the advantage of using less memory and leading to faster execution. During simulation, a data structure queue s is used to save information of burning cells and to guarantee all cells are added by DynamicCellSpaceManager only once. When a cell is burning and DynamicCellSpaceManager adds its neighbor cells in CellSpace, if the neighbor cell is a burning cell, it is in the queue, so DynamicCellSpaceManager will not add it.

4 Figure. Comparing the non-dynamic structure and dynamic structure implementation Figure shows fire-spreading display of DS and non-ds models. Three simulation stages are given. For the figures at the top row, the red cells are burning; the black cells are burned out; the pink cells are just ignited and transiting to the burning state; all other cells are unburned where different colors represent different fuel models. In the bottom row figures, the white spaces indicate that these cells (either unburned or burned out) are not loaded as part of the model. This comparison shows that the two implementations lead to the same simulation results. Figure also clearly shows the difference between the two implementation approaches. In the dynamic structure implementation, cells are dynamically added when they are about to be ignited by their neighbors, and removed when they are burned out. However, in the nondynamic structure implementation, all cells are loaded from the beginning and kept throughout the simulation. IV. PERFORMANCE MEASUREMENT RESULTS To test and detailed analyze DS performance, the same experiment frame of forest fire model is used in both non-ds and DS models. 6 metrics as described in section 2 are measured in this experimentation section. They are listed as follows. A. Initialization time for different cellular space size Table. Initialization Time (sec.) of Different Cellular Space Size C ellular Non-DS DS M odel Space Size M odel 2*2.7. 4* * Table shows the data of initialization time on 2*2, 4*4 and 6*6 cellular space. The initialization time of non-ds model increases with the increase of cellular space because non-ds model initialize all cells in cellular space at the beginning of simulation. But for DS model, the initialization time is little and does not change with the increase of cellular space size. From this point, DS model shows better performance on initialization than non-ds model. B. Execution time for different cellular space size The execution time is measured on simulation stages for different cellular space, which is depicted in figure 2(a). The speedup that is calculated by non-ds model s execution time divided by DS model s execution time is shown in figure 2(b). In these two figures, simulation stage is referred as the logical simulation time that is denoted by tn (e.g., stage 2 (S2) (tn: 8k) means simulation time period is from to 8). 3 curves for non-ds model and 3 curves for DS model on 2*2, 4*4 and 6*6 cellular space are displayed in figure 2(a). The non- DS 2*2 and DS 2*2 models finish simulation before stage 6 because small cell space runs out of simulation earlier than large cell space. When comparing the non-ds and DS model on the same cellular space, we see that non-ds model s execution time is larger than DS model at the same stage. This discipline is more obvious when cell space is large. Another discipline is that for the same model (e.g., non-ds or DS) and same stage, the execution time of larger cellular space is more than smaller cellular space at the same stage. In figure 2(b), speedup is displayed based on the data in figure 2(a). The speedup is almost in horizontal line. 6*6 cellular space s speedup is the largest. 4*4 cellular space s speedup is larger than 2*2. To analyze the reason behind description, we go back to formula (5) as introduced in earlier section. The non-ds model s cycle execution time consists of active cells and nonactive cells execution time. These two parts are different. The DS model s execution time consists of active cells execution time and overhead as described in formula (3). The difference lies in whether non-active cells execution time in non-ds model is larger than overhead in DS model or not. Therefore, on one hand, when the cellular space becomes larger, the nonactive cell s execution time increases and the proportion that overhead occupies decreases, so the speedup of small cellular space is smaller than large cellular space at same stage. On the other hand, for a certain cellular space size, the model behavior decides that in non-ds model, the proportion of nonactive cells execution time changes little and in DS model, the overhead proportion of execution changes little as well when simulation continues, so the speedup changes little correspondingly. Execution Time (s) Execution Time of non-ds and DS on Different Cell Space S (tn:4k) S2 (tn:8k) S3 (tn:2k) S4 (tn:6k) S5 (tn:2k) S6 (tn:24k) (non-ds)2*2 (non-ds)4*4 (non-ds)6*6 (DS)2*2 (DS)4*4 (DS)6*6 Speedup on Different Cell Space (a) (b) Figure 2. (a) Execution Time of Different Cellular Space (b) Speedup of Different Cellular Space C. Execution time at different simulation stages Detailed data about DS model is shown in figure 3. In this figure, simulation stages are intervals and they are continuous. Added/deleted number of cells, overhead and total execution time are displayed separately. From the figure, all four lines have the same trend, which means the rate of overhead and total execution time are similar with the rate of added/deleted cells. In this metric, we count the simulation time of DS model into two parts: time of active cells execution and time of overhead. The active cells execution time is proportional to the number of active cells. The overhead is due to the number of added/deleted cells. When simulation executes from early to later stage, the overhead increases because the number of added/deleted cells increases as well. Therefore, total Speedup S (tn:4k) S2 (tn:8k) S3 (tn:2k) S4 (tn:6k) S5 (tn:2k) S6 (tn:24k) 2*2 4*4 6*6

5 execution time increase. Note that the convex in stage 2 is caused by the model itself behavior, so the cumulative execution time from stage to stage 2 increases faster than other stage period, which is consistent with (DS)6*6 line in figure 2(a). In all, the execution time is proportional to the number of added/deleted cells. If the number of added/deleted cells becomes very large in some situation (e.g., later simulation stage), the execution time of DS model is possible to be greater than non-ds model. In such situation, the performance of DS model is worse than non-ds model. Execution Time (s) S (tn:-4k) Figure 3. Simulation Time Analysis (6 * 6 Cell Space Size) D. Memory usage Execution Time of 6*6 Cell Space Analysis S2 (tn:4k-8k) S3 (tn:8k-2k) S4 (tn:2k-6k) S5 (tn:6k-2k) S6 (tn:2k-24k) Added# (*) deleted# (*) Toverhead Execution time One of the most advantages of DS model is reducing memory use. In figure 4, the comparison of memory use between non-ds and DS model is presented. The top diagraph shows the result on different simulation stage about 6*6 cellular space. The bottom diagraph shows the result on different cellular space. In non-ds model, the active cells are always equal to cell space cells, so the memory using keeps in top line and does not change during simulation. In DS model, its active cells do not change much during whole simulation process, which is listed in table 2, so the memory using is in the bottom line and is less than non-ds model, as depicted in top diagraph of figure 4. On the other hand, DS model does not change with the increase of cellular space size. But for non-ds model, the memory use increases with the increase of cellular space. So when the cellular space size becomes very large, non-ds model has the deficiency about memory issue. DS model has no such limitations. Table 2. Active Cells of DS Model (6*6) S (tn:4k) S2 (tn:8k) S3 (tn:2k) S4 (tn:6k) S5 (tn:2k) S6 (tn:24k) # Active Cells E. Execution time for different model behaviors (multiple points) In this metric, we implement 2- and 4- in 4*4 cellular space for DS and non-ds models. The sparse and close behaviors are tested separately. In general, the execution time of non-ds model for sparse s is less than close s. The execution time of DS model for sparse s is more than close s. This is because in non-dynamic structure environment, sparse s cause fire spreading faster than close ; while in dynamic structure environment, sparse s lead to more overhead which is caused by added/deleted cells. The execution time comparison and the number of added/deleted cell for DS are listed in table 3 and table 4. The data of DS model in table 3 is consistent with data in table 4. The more number of added/deleted cells, the larger the execution time. Table 3. Execution Time (sec.) for Multiple Ignitions S (tn:4k) S2 (tn:8k) S3 (tn:2k) S4 (tn:6k) S5 (tn:2k) S6 (tn:24k) 2-sparse s(non-ds) close s(non-ds) sparse s(ds) close s(ds) sparse s(non-ds) close s(non-ds) sparse s(ds) close s(ds) Table 4. Added/Deleted Cells for DS Model (4*4) S (tn:4k) S6 (tn:24k) S2 (tn:8k) S3 (tn:2k) S4 (tn:6k) S5 (tn:2k) 2-sparse s(ds) 584/3 97/72 55/ /4 52/296 73/442 2-close s(ds) 523/273 97/689 53/ /4 56/ /442 4-sparse s(ds) 38/68 389/84 576/ /45 78/53 83/58 4-close s(ds) 65/343 7/ /4 426/25 65/37 78/57 Memory (MByte) Memory (MByte) Memory Use (6*6) non-ds 8 6 DS 4 2 S S2 S3 S4 S5 S6 (tn:4k) (tn:8k) (tn:2k) (tn:6k) (tn:2k) (tn:24k) Memory Use non-ds 8 DS *2 4*4 6*6 Cell Space Figure 4. Memory Use between non-ds and DS Models Speedup S (tn:4k) Speedup of Multiple Ignitions (4*4) S2 (tn:8k) S3 (tn:2k) S4 (tn:6k) S5 (tn:2k) S6 (tn:24k) Speedup of sparse 2- Speedup of close 2- Speedup of sparse 4- Speedup of close 4- Figure 5. Speedup of Multiple Ignitions Figure 5 shows the speedup. When multiple s are sparse, the speedup is decreased. If the behavior condition is same (e.g., sparse or close), the speedup of two s is

6 higher than four s. So the speedup is not only related with the added/deleted cells as stated in section C, but also has relationship with the behavior of model. In this case, the behavior means the sparse or close. F. Execution time for onedcoord, DS and non-ds models From the above analysis based on five measurement metrics, we know the simulation time is related with the cell space size, overhead and model structure behavior. The increase of overhead for DS model reduces its performance when simulation goes to later stages. In order to avoid this disadvantage, an improved simulation engine onedcoord is used. onedcoord uses the same idea as DS model to let only active cells execute simulation, but it has no overhead. The execution time comparison among non-ds, DS model and onedcoord is shown in figure 6. From the figure, onedcoord has much less simulation time compared with non-ds and DS models. Note that the initialization time of onedcoord is more than non-ds model because it initializes special data structure as well. The initialization time is listed in table 5. Even though onedcoord uses most initialization time, its total execution time is least. So onedcoord is the best choice from execution time aspect. Table 5. Initialization Time on 6*6 Cellular Space Initialization Time Execution Time (s) non-ds DS OneDCoord Execution Time of non-ds, DS and onedcoord (6*6) S (tn:4k) S2 (tn:8k) S3 (tn:2k) S4 (tn:6k) S5 (tn:2k) S6 (tn:24k) non-ds Figure 6. Execution Time for onedcoord, non-ds, and DS Models From memory point of view, onedcoord needs initialize all cells at the beginning, so its memory using is similar with non-ds model and even a little bit more because it needs extra memory for its special data structure. Detailed discussion about onedcoord can be referenced in [8]. DS onedcoord scale cellular space models such as the forest fire spreading simulation model studied in this paper, the number of added/deleted cells is typically small when compared to the total number of cells in the cell space. Thus DS is generally preferred for these applications. The performance (both memory and execution time) advantage of DS is especially apparent when the cell space size increases significantly. Since onedcoord has much better execution time performance than DS model and the memory use of DS model is better than onedcoord, one of our future work is to combine DS with onedcoord to get the best performance result on both memory and execution time aspects. REFERENCES [] Uhrmacher, A.M., Dynamic Structures in Modeling and Simulation - A Reflective Approach. ACM Transactions on Modeling and Simulation, Vol.. No.2, p , April 2. [2] Zeigler, B.P., T.G. Kim, and H. Praehofer.: Theory of Modeling and Simulation. 2 ed. 2, New York, NY: Academic Press [3] DEVS-Java Reference Guide, [4] B.P. Zeigler, T.G. Kim and C. Lee, Variable structure modelling methodology: An adaptive computer architecture example, Trans. Sot. Comput. Simulation 7 (4) (99) [5] Barros. F.J. Modeling Formalisms for Dynamic Structure Systems. ACM Transactions on Modeling and Computer Simulation, Vol. 7, No. 4, 5-55, 997 [6] Fernando J. Barros, Maria T. Mendes: Forest fire modelling and simulation in the DELTA environment. Simul. Pr. Theory 5(3): (997) [7] X. Hu, B. P. Zeigler, and S. Mittal, Variable Structure in DEVS Component-Based Modeling and Simulation, SIMULATION: Transactions of The Society for Modeling and Simulation International, Vol. 8, No. 2, pp. 9-2, 25 [8] X. Hu, and B. P. Zeigler: A High Performance Simulation Engine for Large-Scale Cellular DEVS Models, High Performance Computing Symposium (HPC'4), Advanced Simulation Technologies Conference, April 24 [9] Rothermel, R., A mathematical model for predicting fire spread in wildland fuels. Research Paper INT-5. Ogden, UT: U.S. Department of Agriculture, Forest Service, Intermountain Forest and Range Experiment Station, 972 [] Lewis Ntaimo, Bithika Khargharia A Cellular Devs Dynamic Fire Spread Simulation Model With An Optimized Cell Space, Project Report, Electrical and Computer Engineering Department, University of Arizona, May, 23 VI. CONCLUSIONS AND FUTURE WORK Dynamic structure capability provides a natural and effective way to model and simulate complex systems that exhibit structure, behavior, and interface changes to adapt to different situations. DEVS-based dynamic structure formalism supports hierarchical and modular modeling building, so it can be used to represent complex systems. In this paper, the performance measurement and analysis of both dynamic structure model and non-dynamic structure model are presented based on the forest fire spreading simulation. The analysis results show that from the memory point of view, DS model is always superior to non-ds model. But from the execution time point of view, DS may become slower than the non-ds if the number of added/deleted cells is large, which causes significant overhead. However, we note that for large

Towards Validation of DEVS-FIRE Wildfire Simulation Model

Towards Validation of DEVS-FIRE Wildfire Simulation Model Towards Validation of DEVS-FIRE Wildfire Simulation Model Feng Gu Dept. of Computer Science Georgia State University Atlanta, GA, 333 fgu1@student.gsu.edu Xiaolin Hu Dept. of Computer Science Georgia State

More information

λ(s)if e = τ(s) Λ(s,e) = φ if e < τ(s)

λ(s)if e = τ(s) Λ(s,e) = φ if e < τ(s) Proceedings of the 1998 Winter Simulation Conference D.J. Medeiros, E.F. Watson, J.S. Carson and M.S. Manivannan, eds. ABSTRACT SIMULATORS FOR THE DSDE FORMALISM Fernando J. Barros Departamento de Engenharia

More information

DYNAMIC STRUCTURE CELLULAR AUTOMATA IN A FIRE SPREADING APPLICATION

DYNAMIC STRUCTURE CELLULAR AUTOMATA IN A FIRE SPREADING APPLICATION DYNAMIC STRUCTURE CELLULAR AUTOMATA IN A FIRE SPREADING APPLICATION Alexandre Muzy, Eric Innocenti, Antoine Aïello, Jean-François Santucci, Paul-Antoine Santoni University of Corsica SPE UMR CNRS 6134

More information

Linking Information and Energy Activity-based Energy-Aware Information Processing

Linking Information and Energy Activity-based Energy-Aware Information Processing Linking Information and Energy Activity-based Energy-Aware Information Processing Xiaolin Hu Department of Computer Science Georgia State University, Atlanta, GA, 30303 xhu@cs.gsu.edu Bernard P. Zeigler

More information

STOCHASTIC CELLULAR AUTOMATA MODEL FOR WILDLAND FIRE SPREAD DYNAMICS

STOCHASTIC CELLULAR AUTOMATA MODEL FOR WILDLAND FIRE SPREAD DYNAMICS STOCHASTIC CELLULAR AUTOMATA MODEL FOR WILDLAND FIRE SPREAD DYNAMICS Rodolfo Maduro Almeida 1 Elbert Einstein Nehrer Macau 2 1 Programa de Pós-graduação em Computação Aplicada INPE São José dos Campos

More information

Improved cellular models with parallel Cell-DEVS

Improved cellular models with parallel Cell-DEVS Improved cellular models with parallel Cell-DEVS Gabriel A. Wainer Departamento de Computación Facultad de Ciencias Exactas y Naturales Universidad de Buenos Aires Pabellón I - Ciudad Universitaria Buenos

More information

Scheduling of Frame-based Embedded Systems with Rechargeable Batteries

Scheduling of Frame-based Embedded Systems with Rechargeable Batteries Scheduling of Frame-based Embedded Systems with Rechargeable Batteries André Allavena Computer Science Department Cornell University Ithaca, NY 14853 andre@cs.cornell.edu Daniel Mossé Department of Computer

More information

The Discrete EVent System specification (DEVS) formalism

The Discrete EVent System specification (DEVS) formalism The Discrete EVent System specification (DEVS) formalism Hans Vangheluwe The DEVS formalism was conceived by Zeigler [Zei84a, Zei84b] to provide a rigourous common basis for discrete-event modelling and

More information

CS 700: Quantitative Methods & Experimental Design in Computer Science

CS 700: Quantitative Methods & Experimental Design in Computer Science CS 700: Quantitative Methods & Experimental Design in Computer Science Sanjeev Setia Dept of Computer Science George Mason University Logistics Grade: 35% project, 25% Homework assignments 20% midterm,

More information

Applying a Dynamic Data Driven Genetic Algorithm to Improve Forest Fire Spread Prediction

Applying a Dynamic Data Driven Genetic Algorithm to Improve Forest Fire Spread Prediction Applying a Dynamic Data Driven Genetic Algorithm to Improve Forest Fire Spread Prediction Mónica Denham, Ana Cortés, Tomàs Margalef, and Emilio Luque Departament d Arquitectura de Computadors i Sistemes

More information

Exploiting Sensor Spatial Correlation for Dynamic Data Driven Simulation of Wildfire

Exploiting Sensor Spatial Correlation for Dynamic Data Driven Simulation of Wildfire Exploiting Sensor Spatial Correlation for Dynamic Data Driven Simulation of Wildfire Haidong Xue and Xiaolin Hu Computer Science Department Georgia State University Atlanta USA Abstract Dynamic data driven

More information

Implementing Parallel Cell-DEVS

Implementing Parallel Cell-DEVS Implementing Parallel Cell-DEVS Alejandro Troccoli Gabriel Wainer Departamento de Computación. Department of Systems and Computing Pabellón I. Ciudad Universitaria Engineering, Carleton University. 1125

More information

Engineering Self-Organization and Emergence: issues and directions

Engineering Self-Organization and Emergence: issues and directions 5/0/ Engineering Self-Organization and Emergence: issues and directions Franco Zambonelli franco.zambonelli@unimore.it Agents and Pervasive Computing Group Università di Modena e Reggio Emilia SOAS 005

More information

An object-oriented design process. Weather system description. Layered architecture. Process stages. System context and models of use

An object-oriented design process. Weather system description. Layered architecture. Process stages. System context and models of use An object-oriented design process Process stages Structured design processes involve developing a number of different system models. They require a lot of effort for development and maintenance of these

More information

SPACE-BASED DATA MANAGEMENT

SPACE-BASED DATA MANAGEMENT SPACE-BASED DATA MANAGEMENT FOR HIGH PERFORMANCE DISTRIBUTED SIMULATION By Jong Sik Lee Copyright? Jong Sik Lee 2001 A Dissertation Submitted to the Faculty of the DEPARTMENT OF ELECTRICAL AND COMPUTER

More information

Real-Time Scheduling and Resource Management

Real-Time Scheduling and Resource Management ARTIST2 Summer School 2008 in Europe Autrans (near Grenoble), France September 8-12, 2008 Real-Time Scheduling and Resource Management Lecturer: Giorgio Buttazzo Full Professor Scuola Superiore Sant Anna

More information

Analytical Modeling of Parallel Systems

Analytical Modeling of Parallel Systems Analytical Modeling of Parallel Systems Chieh-Sen (Jason) Huang Department of Applied Mathematics National Sun Yat-sen University Thank Ananth Grama, Anshul Gupta, George Karypis, and Vipin Kumar for providing

More information

A Truthful Mechanism for Fair Load Balancing in Distributed Systems Λ

A Truthful Mechanism for Fair Load Balancing in Distributed Systems Λ A Truthful Mechanism for Fair Load Balancing in Distributed Systems Λ Daniel Grosu and Anthony T. Chronopoulos Department of Computer Science, University of Texas at San Antonio, 6900 N. Loop 1604 West,

More information

HYCOM and Navy ESPC Future High Performance Computing Needs. Alan J. Wallcraft. COAPS Short Seminar November 6, 2017

HYCOM and Navy ESPC Future High Performance Computing Needs. Alan J. Wallcraft. COAPS Short Seminar November 6, 2017 HYCOM and Navy ESPC Future High Performance Computing Needs Alan J. Wallcraft COAPS Short Seminar November 6, 2017 Forecasting Architectural Trends 3 NAVY OPERATIONAL GLOBAL OCEAN PREDICTION Trend is higher

More information

Description of the fire scheme in WRF

Description of the fire scheme in WRF Description of the fire scheme in WRF March 8, 2010 1 Introduction The wildland fire model in WRF is an implementation of the semi-empirical fire propagation model developed in Coen (2005) and Clark et

More information

Stochastic cellular automata model for wildland fire spread dynamics

Stochastic cellular automata model for wildland fire spread dynamics Journal of Physics: Conference Series Stochastic cellular automata model for wildland fire spread dynamics To cite this article: Rodolfo Maduro Almeida and Elbert E N Macau 2011 J. Phys.: Conf. Ser. 285

More information

Synthesis of Saturating Counters Using Traditional and Non-traditional Basic Counters

Synthesis of Saturating Counters Using Traditional and Non-traditional Basic Counters Synthesis of Saturating Counters Using Traditional and Non-traditional Basic Counters Zhaojun Wo and Israel Koren Department of Electrical and Computer Engineering University of Massachusetts, Amherst,

More information

A Method for Reducing Ill-Conditioning of Polynomial Root Finding Using a Change of Basis

A Method for Reducing Ill-Conditioning of Polynomial Root Finding Using a Change of Basis Portland State University PDXScholar University Honors Theses University Honors College 2014 A Method for Reducing Ill-Conditioning of Polynomial Root Finding Using a Change of Basis Edison Tsai Portland

More information

Wildfires. Chun-Lung Lim and Charles Erwin

Wildfires. Chun-Lung Lim and Charles Erwin Wildfires Chun-Lung Lim and Charles Erwin THE WEATHER RESEARCH AND FORECAST MODEL (WRF) What makes WRF next-generation mesoscale forecast model? Advance the understanding and the prediction of mesoscale

More information

Fire frequency in the Western Cape

Fire frequency in the Western Cape Fire frequency in the Western Cape First year progress report Diane Southey 3 May 27 This report is a summary of the work I have done in the first year of my masters. Each section is briefly discussed

More information

Simulation tools to predict wildfire behavior and to help safety decision making

Simulation tools to predict wildfire behavior and to help safety decision making Simulation tools to predict wildfire behavior and to help safety decision making Marc Castellnou and Bruce Schubert Marc Castellnou, Fire Chief, Catalonia Fire Service Pau Costa Foundation, President Fire

More information

Bandwidth Utilization/Fidelity Tradeoffs in Predictive Filtering 1

Bandwidth Utilization/Fidelity Tradeoffs in Predictive Filtering 1 Bandwidth Utilization/Fidelity Tradeoffs in Predictive Filtering 1 Bernard P. Zeigler George Ball Hyup Cho J.S. Lee Hessam Sarjoughian AI and Simulation Group Department of Electrical and Computer Engineering

More information

Hierarchical Anomaly Detection in Load Testing with StormRunner Load

Hierarchical Anomaly Detection in Load Testing with StormRunner Load White Paper Application Development, Test & Delivery Hierarchical Anomaly Detection in Load Testing with StormRunner Load A fresh approach to cloud-based website load testing is proving more effective

More information

Transposition Mechanism for Sparse Matrices on Vector Processors

Transposition Mechanism for Sparse Matrices on Vector Processors Transposition Mechanism for Sparse Matrices on Vector Processors Pyrrhos Stathis Stamatis Vassiliadis Sorin Cotofana Electrical Engineering Department, Delft University of Technology, Delft, The Netherlands

More information

Practical QSAR and Library Design: Advanced tools for research teams

Practical QSAR and Library Design: Advanced tools for research teams DS QSAR and Library Design Webinar Practical QSAR and Library Design: Advanced tools for research teams Reservationless-Plus Dial-In Number (US): (866) 519-8942 Reservationless-Plus International Dial-In

More information

Human-Caused Fires: An Exploratory Pattern Analysis

Human-Caused Fires: An Exploratory Pattern Analysis Human-Caused Fires: An Exploratory Pattern Analysis J E N A F E R R A R E S E N I C H O L A S J O N E S D I G I T A L C O M P I L A T I O N A N D D A T A B A S E D E S I G N P O R T L A N D S T A T E U

More information

The Weighted Byzantine Agreement Problem

The Weighted Byzantine Agreement Problem The Weighted Byzantine Agreement Problem Vijay K. Garg and John Bridgman Department of Electrical and Computer Engineering The University of Texas at Austin Austin, TX 78712-1084, USA garg@ece.utexas.edu,

More information

Calculating the Spread of Wildfires

Calculating the Spread of Wildfires Calculating the Spread of Wildfires New Mexico Supercomputing Challenge Final Report April 5, 2017 Team 85 Los Alamos Middle School Seventh Grade Physics Programmed in Python Team Members: Jonathan Triplett:

More information

Lecture 18 April 26, 2012

Lecture 18 April 26, 2012 6.851: Advanced Data Structures Spring 2012 Prof. Erik Demaine Lecture 18 April 26, 2012 1 Overview In the last lecture we introduced the concept of implicit, succinct, and compact data structures, and

More information

EE 144/244: Fundamental Algorithms for System Modeling, Analysis, and Optimization Fall 2014

EE 144/244: Fundamental Algorithms for System Modeling, Analysis, and Optimization Fall 2014 EE 144/244: Fundamental Algorithms for System Modeling, Analysis, and Optimization Fall 2014 Discrete Event Simulation Stavros Tripakis University of California, Berkeley Stavros Tripakis (UC Berkeley)

More information

COMP9334: Capacity Planning of Computer Systems and Networks

COMP9334: Capacity Planning of Computer Systems and Networks COMP9334: Capacity Planning of Computer Systems and Networks Week 2: Operational analysis Lecturer: Prof. Sanjay Jha NETWORKS RESEARCH GROUP, CSE, UNSW Operational analysis Operational: Collect performance

More information

Interacting Vehicles: Rules of the Game

Interacting Vehicles: Rules of the Game Chapter 7 Interacting Vehicles: Rules of the Game In previous chapters, we introduced an intelligent control method for autonomous navigation and path planning. The decision system mainly uses local information,

More information

CSE373: Data Structures and Algorithms Lecture 2: Math Review; Algorithm Analysis. Hunter Zahn Summer 2016

CSE373: Data Structures and Algorithms Lecture 2: Math Review; Algorithm Analysis. Hunter Zahn Summer 2016 CSE373: Data Structures and Algorithms Lecture 2: Math Review; Algorithm Analysis Hunter Zahn Summer 2016 Today Finish discussing stacks and queues Review math essential to algorithm analysis Proof by

More information

SYSTEM EVALUATION AND DESCRIPTION USING ABSTRACT RELATION TYPES (ART)

SYSTEM EVALUATION AND DESCRIPTION USING ABSTRACT RELATION TYPES (ART) SYSTEM EVALUATION AND DESCRIPTION USING ABSTRACT RELATION TYPES (ART) Joseph J. Simpson System Concepts 6400 32 nd Ave. NW #9 Seattle, WA 9807 206-78-7089 jjs-sbw@eskimo.com Dr. Cihan H. Dagli UMR-Rolla

More information

Specification of Dynamic Structure Discrete Event Multiagent Systems

Specification of Dynamic Structure Discrete Event Multiagent Systems Specification of Dynamic Structure Discrete Event Multiagent Systems Raphaël Duboz 1, David Versmisse 2, Gauthier Quesnel 2, Alexandre Muzy 3 and Éric Ramat 2 1 Institut de Recherche pour le Développement

More information

Trajectory planning and feedforward design for electromechanical motion systems version 2

Trajectory planning and feedforward design for electromechanical motion systems version 2 2 Trajectory planning and feedforward design for electromechanical motion systems version 2 Report nr. DCT 2003-8 Paul Lambrechts Email: P.F.Lambrechts@tue.nl April, 2003 Abstract This report considers

More information

OBTAINING SQUARES FROM THE PRODUCTS OF NON-SQUARE INTEGERS

OBTAINING SQUARES FROM THE PRODUCTS OF NON-SQUARE INTEGERS OBTAINING SQUARES FROM THE PRODUCTS OF NON-SQUARE INTEGERS The difference between two neighboring squares n 2 and (n+1) 2 is equal to 2n+1 for any integer n=1,2,3,. Thus the numbers generated by n 2 -A

More information

Hashing. Hashing DESIGN & ANALYSIS OF ALGORITHM

Hashing. Hashing DESIGN & ANALYSIS OF ALGORITHM Hashing Hashing Start with an array that holds the hash table. Use a hash function to take a key and map it to some index in the array. If the desired record is in the location given by the index, then

More information

Notation. Bounds on Speedup. Parallel Processing. CS575 Parallel Processing

Notation. Bounds on Speedup. Parallel Processing. CS575 Parallel Processing Parallel Processing CS575 Parallel Processing Lecture five: Efficiency Wim Bohm, Colorado State University Some material from Speedup vs Efficiency in Parallel Systems - Eager, Zahorjan and Lazowska IEEE

More information

CSE 380 Computer Operating Systems

CSE 380 Computer Operating Systems CSE 380 Computer Operating Systems Instructor: Insup Lee & Dianna Xu University of Pennsylvania, Fall 2003 Lecture Note 3: CPU Scheduling 1 CPU SCHEDULING q How can OS schedule the allocation of CPU cycles

More information

NEC PerforCache. Influence on M-Series Disk Array Behavior and Performance. Version 1.0

NEC PerforCache. Influence on M-Series Disk Array Behavior and Performance. Version 1.0 NEC PerforCache Influence on M-Series Disk Array Behavior and Performance. Version 1.0 Preface This document describes L2 (Level 2) Cache Technology which is a feature of NEC M-Series Disk Array implemented

More information

PrepTest. October 1996 Logic Games Setups. Chapter Twenty: October 1996 Logic Games Setups

PrepTest. October 1996 Logic Games Setups. Chapter Twenty: October 1996 Logic Games Setups PrepTest October 1996 Logic Games Setups Chapter Twenty: October 1996 Logic Games Setups 305 PrepTest. October 1996 Game #1: 1. A 2. A 3. D 4. A 5. B This is an Advanced Linear: Unbalanced: Underfunded

More information

A 2-Approximation Algorithm for Scheduling Parallel and Time-Sensitive Applications to Maximize Total Accrued Utility Value

A 2-Approximation Algorithm for Scheduling Parallel and Time-Sensitive Applications to Maximize Total Accrued Utility Value A -Approximation Algorithm for Scheduling Parallel and Time-Sensitive Applications to Maximize Total Accrued Utility Value Shuhui Li, Miao Song, Peng-Jun Wan, Shangping Ren Department of Engineering Mechanics,

More information

WRF-Fire. Janice Coen. WRF Users' Tutorial! Mesoscale & Microscale Meteorological Division / NCAR

WRF-Fire. Janice Coen. WRF Users' Tutorial! Mesoscale & Microscale Meteorological Division / NCAR WRF-Fire Janice Coen 1 Introduction! WRF-Fire is a physics package within WRF ARW that allows users to model the growth of a wildland fire! Fire responds to environmental conditions: terrain slope, fuel

More information

An Smodels System with Limited Lookahead Computation

An Smodels System with Limited Lookahead Computation An Smodels System with Limited Lookahead Computation Gayathri Namasivayam and Miros law Truszczyński Department of Computer Science, University of Kentucky, Lexington, KY 40506-0046, USA Abstract. We describe

More information

DISTRIBUTED COMPUTER SYSTEMS

DISTRIBUTED COMPUTER SYSTEMS DISTRIBUTED COMPUTER SYSTEMS SYNCHRONIZATION Dr. Jack Lange Computer Science Department University of Pittsburgh Fall 2015 Topics Clock Synchronization Physical Clocks Clock Synchronization Algorithms

More information

Parallel programming using MPI. Analysis and optimization. Bhupender Thakur, Jim Lupo, Le Yan, Alex Pacheco

Parallel programming using MPI. Analysis and optimization. Bhupender Thakur, Jim Lupo, Le Yan, Alex Pacheco Parallel programming using MPI Analysis and optimization Bhupender Thakur, Jim Lupo, Le Yan, Alex Pacheco Outline l Parallel programming: Basic definitions l Choosing right algorithms: Optimal serial and

More information

CS505: Distributed Systems

CS505: Distributed Systems Cristina Nita-Rotaru CS505: Distributed Systems. Required reading for this topic } Michael J. Fischer, Nancy A. Lynch, and Michael S. Paterson for "Impossibility of Distributed with One Faulty Process,

More information

The Fast Optimal Voltage Partitioning Algorithm For Peak Power Density Minimization

The Fast Optimal Voltage Partitioning Algorithm For Peak Power Density Minimization The Fast Optimal Voltage Partitioning Algorithm For Peak Power Density Minimization Jia Wang, Shiyan Hu Department of Electrical and Computer Engineering Michigan Technological University Houghton, Michigan

More information

Morphing ensemble Kalman filter

Morphing ensemble Kalman filter Morphing ensemble Kalman filter and applications Center for Computational Mathematics Department of Mathematical and Statistical Sciences University of Colorado Denver Supported by NSF grants CNS-0623983

More information

Season Finale: Which one is better?

Season Finale: Which one is better? CS4310.01 Introduction to Operating System Spring 2016 Dr. Zhizhang Shen Season Finale: Which one is better? 1 Background In this lab, we will study, and compare, two processor scheduling policies via

More information

SPATIAL DATA MINING. Ms. S. Malathi, Lecturer in Computer Applications, KGiSL - IIM

SPATIAL DATA MINING. Ms. S. Malathi, Lecturer in Computer Applications, KGiSL - IIM SPATIAL DATA MINING Ms. S. Malathi, Lecturer in Computer Applications, KGiSL - IIM INTRODUCTION The main difference between data mining in relational DBS and in spatial DBS is that attributes of the neighbors

More information

WIRELESS cellular networks derive their name. Call Admission Control in Mobile Cellular Systems Using Fuzzy Associative Memory

WIRELESS cellular networks derive their name. Call Admission Control in Mobile Cellular Systems Using Fuzzy Associative Memory ICCCN 23 Call Admission Control in Mobile Cellular Systems Using Fuzzy Associative Memory Sivaramakrishna Mopati, and Dilip Sarkar Abstract In a mobile cellular system quality of service to mobile terminals

More information

Static Program Analysis using Abstract Interpretation

Static Program Analysis using Abstract Interpretation Static Program Analysis using Abstract Interpretation Introduction Static Program Analysis Static program analysis consists of automatically discovering properties of a program that hold for all possible

More information

Labelling technique for the fast Star Identification

Labelling technique for the fast Star Identification Labelling technique for the fast Star Identification By Sangkyun Kim, 1) Mengu Cho, 1) 1) Kyushu Institute of Technology, Japan (Received June 21st, 2017) Star sensor provides with the most accurate attitude

More information

Introducing GIS analysis

Introducing GIS analysis 1 Introducing GIS analysis GIS analysis lets you see patterns and relationships in your geographic data. The results of your analysis will give you insight into a place, help you focus your actions, or

More information

INCREMENTAL SAMPLING METHODOLOGY

INCREMENTAL SAMPLING METHODOLOGY INCREMENTAL SAMPLING METHODOLOGY 1. INTRODUCTION Incremental sampling methodology (ISM) is a structured composite sampling and processing protocol having specific elements designed to reduce data variability

More information

Example: 2x y + 3z = 1 5y 6z = 0 x + 4z = 7. Definition: Elementary Row Operations. Example: Type I swap rows 1 and 3

Example: 2x y + 3z = 1 5y 6z = 0 x + 4z = 7. Definition: Elementary Row Operations. Example: Type I swap rows 1 and 3 Linear Algebra Row Reduced Echelon Form Techniques for solving systems of linear equations lie at the heart of linear algebra. In high school we learn to solve systems with or variables using elimination

More information

Theory of Computer Science

Theory of Computer Science Theory of Computer Science E1. Complexity Theory: Motivation and Introduction Malte Helmert University of Basel May 18, 2016 Overview: Course contents of this course: logic How can knowledge be represented?

More information

TEPZZ 95785_A_T EP A1 (19) (11) EP A1 (12) EUROPEAN PATENT APPLICATION

TEPZZ 95785_A_T EP A1 (19) (11) EP A1 (12) EUROPEAN PATENT APPLICATION (19) TEPZZ 978_A_T (11) EP 2 97 81 A1 (12) EUROPEAN PATENT APPLICATION (43) Date of publication: 23.12. Bulletin /2 (21) Application number: 14172928. (1) Int Cl.: F28F 3/04 (06.01) F28F 3/08 (06.01) F28F

More information

Parallelization of the QC-lib Quantum Computer Simulator Library

Parallelization of the QC-lib Quantum Computer Simulator Library Parallelization of the QC-lib Quantum Computer Simulator Library Ian Glendinning and Bernhard Ömer VCPC European Centre for Parallel Computing at Vienna Liechtensteinstraße 22, A-19 Vienna, Austria http://www.vcpc.univie.ac.at/qc/

More information

Analytical Modeling of Parallel Programs. S. Oliveira

Analytical Modeling of Parallel Programs. S. Oliveira Analytical Modeling of Parallel Programs S. Oliveira Fall 2005 1 Scalability of Parallel Systems Efficiency of a parallel program E = S/P = T s /PT p Using the parallel overhead expression E = 1/(1 + T

More information

A Cellular Automata Approach to Population Modeling

A Cellular Automata Approach to Population Modeling A Cellular Automata Approach to Population Modeling Alexa M. Silverman February 24, 2009 Abstract 1 Introduction 1.1 Cellular automata This project provides an agent-based model of the effects of temperature

More information

A New Dominant Point-Based Parallel Algorithm for Multiple Longest Common Subsequence Problem

A New Dominant Point-Based Parallel Algorithm for Multiple Longest Common Subsequence Problem A New Dominant Point-Based Parallel Algorithm for Multiple Longest Common Subsequence Problem Dmitry Korkin This work introduces a new parallel algorithm for computing a multiple longest common subsequence

More information

THE MOVING MAN: DISTANCE, DISPLACEMENT, SPEED & VELOCITY

THE MOVING MAN: DISTANCE, DISPLACEMENT, SPEED & VELOCITY THE MOVING MAN: DISTANCE, DISPLACEMENT, SPEED & VELOCITY Background Remember graphs are not just an evil thing your teacher makes you create, they are a means of communication. Graphs are a way of communicating

More information

Theory of Computer Science. Theory of Computer Science. E1.1 Motivation. E1.2 How to Measure Runtime? E1.3 Decision Problems. E1.

Theory of Computer Science. Theory of Computer Science. E1.1 Motivation. E1.2 How to Measure Runtime? E1.3 Decision Problems. E1. Theory of Computer Science May 18, 2016 E1. Complexity Theory: Motivation and Introduction Theory of Computer Science E1. Complexity Theory: Motivation and Introduction Malte Helmert University of Basel

More information

Announcements. CSE332: Data Abstractions Lecture 2: Math Review; Algorithm Analysis. Today. Mathematical induction. Dan Grossman Spring 2010

Announcements. CSE332: Data Abstractions Lecture 2: Math Review; Algorithm Analysis. Today. Mathematical induction. Dan Grossman Spring 2010 Announcements CSE332: Data Abstractions Lecture 2: Math Review; Algorithm Analysis Dan Grossman Spring 2010 Project 1 posted Section materials on using Eclipse will be very useful if you have never used

More information

Intel Stratix 10 Thermal Modeling and Management

Intel Stratix 10 Thermal Modeling and Management Intel Stratix 10 Thermal Modeling and Management Updated for Intel Quartus Prime Design Suite: 17.1 Subscribe Send Feedback Latest document on the web: PDF HTML Contents Contents 1...3 1.1 List of Abbreviations...

More information

Initial Report. North Central Texas Initial Attack. Hildreth Road Dozer Accident. February 9, 2006

Initial Report. North Central Texas Initial Attack. Hildreth Road Dozer Accident. February 9, 2006 6.152 February 17, 2006 Initial Report North Central Texas Initial Attack Hildreth Road Dozer Accident February 9, 2006 This report is intended to inform interested parties, for a safety and training tool,

More information

Distributed Data Fusion with Kalman Filters. Simon Julier Computer Science Department University College London

Distributed Data Fusion with Kalman Filters. Simon Julier Computer Science Department University College London Distributed Data Fusion with Kalman Filters Simon Julier Computer Science Department University College London S.Julier@cs.ucl.ac.uk Structure of Talk Motivation Kalman Filters Double Counting Optimal

More information

Research Article Simplified Robotics Joint-Space Trajectory Generation with a via Point Using a Single Polynomial

Research Article Simplified Robotics Joint-Space Trajectory Generation with a via Point Using a Single Polynomial Robotics Volume, Article ID 75958, 6 pages http://dx.doi.org/.55//75958 Research Article Simplified Robotics Joint-Space Trajectory Generation with a via Point Using a Single Polynomial Robert L. Williams

More information

Demand Estimation Sub-Committee MOD330 Phase 1 and 2. 7 th November 2012

Demand Estimation Sub-Committee MOD330 Phase 1 and 2. 7 th November 2012 Demand Estimation Sub-Committee MOD330 Phase 1 and 2 7 th November 2012 DESC: Phase 1 Update (1 of 2) 2 Teleconference took place on 18 th October to obtain DESC confirmation of scope of weather data requirements

More information

The AIR Bushfire Model for Australia

The AIR Bushfire Model for Australia The AIR Bushfire Model for Australia In February 2009, amid tripledigit temperatures and drought conditions, fires broke out just north of Melbourne, Australia. Propelled by high winds, as many as 400

More information

2. Accelerated Computations

2. Accelerated Computations 2. Accelerated Computations 2.1. Bent Function Enumeration by a Circular Pipeline Implemented on an FPGA Stuart W. Schneider Jon T. Butler 2.1.1. Background A naive approach to encoding a plaintext message

More information

Intelligent GIS: Automatic generation of qualitative spatial information

Intelligent GIS: Automatic generation of qualitative spatial information Intelligent GIS: Automatic generation of qualitative spatial information Jimmy A. Lee 1 and Jane Brennan 1 1 University of Technology, Sydney, FIT, P.O. Box 123, Broadway NSW 2007, Australia janeb@it.uts.edu.au

More information

Proceedings, International Snow Science Workshop, Innsbruck, Austria, 2018

Proceedings, International Snow Science Workshop, Innsbruck, Austria, 2018 RELEASE OF AVALANCHES ON PERSISTENT WEAK LAYERS IN RELATION TO LOADING EVENTS IN COLORADO, USA Jason Konigsberg 1, Spencer Logan 1, and Ethan Greene 1 1 Colorado Avalanche Information Center, Boulder,

More information

CHATTERING AVOIDANCE IN HYBRID SIMULATION MODELS: A MODULAR APPROACH BASED ON THE HYFLOW FORMALISM

CHATTERING AVOIDANCE IN HYBRID SIMULATION MODELS: A MODULAR APPROACH BASED ON THE HYFLOW FORMALISM CHATTERING AVOIDANCE IN HYBRID SIMULATION MODELS: A MODULAR APPROACH BASED ON THE HYFLOW FORMALISM Fernando J. Department of Informatics Engineering University of Coimbra 3030 Coimbra, Portugal barros@dei.uc.pt

More information

A Cellular Automata Approach to Population Modeling

A Cellular Automata Approach to Population Modeling A Cellular Automata Approach to Population Modeling Alexa M. Silverman March 31, 2009 Abstract 1 Introduction 1.1 Cellular automata This project provides an agent-based model of the effects of temperature

More information

State of the art Image Compression Techniques

State of the art Image Compression Techniques Chapter 4 State of the art Image Compression Techniques In this thesis we focus mainly on the adaption of state of the art wavelet based image compression techniques to programmable hardware. Thus, an

More information

2. REGIONAL DISPERSION

2. REGIONAL DISPERSION Real-time Transport and Dispersion from Illinois Nuclear Power Plants Thomas E. Bellinger, CCM Illinois Emergency Management Agency Springfield, Illinois 1. INTRODUCTION Meteorological data routinely used

More information

A GUI FOR EVOLVE ZAMS

A GUI FOR EVOLVE ZAMS A GUI FOR EVOLVE ZAMS D. R. Schlegel Computer Science Department Here the early work on a new user interface for the Evolve ZAMS stellar evolution code is presented. The initial goal of this project is

More information

Projectile Motion Slide 1/16. Projectile Motion. Fall Semester. Parallel Computing

Projectile Motion Slide 1/16. Projectile Motion. Fall Semester. Parallel Computing Projectile Motion Slide 1/16 Projectile Motion Fall Semester Projectile Motion Slide 2/16 Topic Outline Historical Perspective ABC and ENIAC Ballistics tables Projectile Motion Air resistance Euler s method

More information

SPC Fire Weather Forecast Criteria

SPC Fire Weather Forecast Criteria SPC Fire Weather Forecast Criteria Critical for temperature, wind, and relative humidity: - Sustained winds 20 mph or greater (15 mph Florida) - Minimum relative humidity at or below regional thresholds

More information

Combining Shared Coin Algorithms

Combining Shared Coin Algorithms Combining Shared Coin Algorithms James Aspnes Hagit Attiya Keren Censor Abstract This paper shows that shared coin algorithms can be combined to optimize several complexity measures, even in the presence

More information

Sensitivity of System Reliability to Usage Profile Changes

Sensitivity of System Reliability to Usage Profile Changes Sensitivity of System Reliability to Usage Profile Changes Kim Weyns Department of Communication Systems, Lund University PO Box 118 SE-211 00 LUND, Sweden kimweyns@telecomlthse Per Runeson Department

More information

SP-CNN: A Scalable and Programmable CNN-based Accelerator. Dilan Manatunga Dr. Hyesoon Kim Dr. Saibal Mukhopadhyay

SP-CNN: A Scalable and Programmable CNN-based Accelerator. Dilan Manatunga Dr. Hyesoon Kim Dr. Saibal Mukhopadhyay SP-CNN: A Scalable and Programmable CNN-based Accelerator Dilan Manatunga Dr. Hyesoon Kim Dr. Saibal Mukhopadhyay Motivation Power is a first-order design constraint, especially for embedded devices. Certain

More information

Information System Decomposition Quality

Information System Decomposition Quality Information System Decomposition Quality Dr. Nejmeddine Tagoug Computer Science Department Al Imam University, SA najmtagoug@yahoo.com ABSTRACT: Object-oriented design is becoming very popular in today

More information

Power System Analysis Prof. A. K. Sinha Department of Electrical Engineering Indian Institute of Technology, Kharagpur. Lecture - 21 Power Flow VI

Power System Analysis Prof. A. K. Sinha Department of Electrical Engineering Indian Institute of Technology, Kharagpur. Lecture - 21 Power Flow VI Power System Analysis Prof. A. K. Sinha Department of Electrical Engineering Indian Institute of Technology, Kharagpur Lecture - 21 Power Flow VI (Refer Slide Time: 00:57) Welcome to lesson 21. In this

More information

Entropy and Self-Organization in Multi- Agent Systems. CSCE 990 Seminar Zhanping Xu 03/13/2013

Entropy and Self-Organization in Multi- Agent Systems. CSCE 990 Seminar Zhanping Xu 03/13/2013 Entropy and Self-Organization in Multi- Agent Systems CSCE 990 Seminar Zhanping Xu 03/13/2013 Citation of the Article H. V. D. Parunak and S. Brueckner, Entropy and Self Organization in Multi-Agent Systems,

More information

Module 5: CPU Scheduling

Module 5: CPU Scheduling Module 5: CPU Scheduling Basic Concepts Scheduling Criteria Scheduling Algorithms Multiple-Processor Scheduling Real-Time Scheduling Algorithm Evaluation 5.1 Basic Concepts Maximum CPU utilization obtained

More information

Q-V droop control using fuzzy logic and reciprocal characteristic

Q-V droop control using fuzzy logic and reciprocal characteristic International Journal of Smart Grid and Clean Energy Q-V droop control using fuzzy logic and reciprocal characteristic Lu Wang a*, Yanting Hu a, Zhe Chen b a School of Engineering and Applied Physics,

More information

Revision History Date Version Description Author 03/06/ Initial Version Sergio Miguel Martin. Table of Contents

Revision History Date Version Description Author 03/06/ Initial Version Sergio Miguel Martin. Table of Contents Icarus Physics Engine Manual 1 - Basic Engine Operation Version 1.0 Revision History Date Version Description Author 03/06/2010 1.0 Initial Version Sergio Miguel Martin Table of Contents 1. Introduction

More information

Lecture 14 October 22

Lecture 14 October 22 EE 2: Coding for Digital Communication & Beyond Fall 203 Lecture 4 October 22 Lecturer: Prof. Anant Sahai Scribe: Jingyan Wang This lecture covers: LT Code Ideal Soliton Distribution 4. Introduction So

More information

Compressed Sensing Using Reed- Solomon and Q-Ary LDPC Codes

Compressed Sensing Using Reed- Solomon and Q-Ary LDPC Codes Compressed Sensing Using Reed- Solomon and Q-Ary LDPC Codes Item Type text; Proceedings Authors Jagiello, Kristin M. Publisher International Foundation for Telemetering Journal International Telemetering

More information

DESIGNING CNN GENES. Received January 23, 2003; Revised April 2, 2003

DESIGNING CNN GENES. Received January 23, 2003; Revised April 2, 2003 Tutorials and Reviews International Journal of Bifurcation and Chaos, Vol. 13, No. 10 (2003 2739 2824 c World Scientific Publishing Company DESIGNING CNN GENES MAKOTO ITOH Department of Information and

More information