## State Estimation Using Particle Filters in Wildfire Spread Simulation

Citations: | 2 - 2 self |

### BibTeX

@MISC{Gu_stateestimation,

author = {Feng Gu and Xuefeng Yan and Xiaolin Hu},

title = {State Estimation Using Particle Filters in Wildfire Spread Simulation},

year = {}

}

### OpenURL

### Abstract

Keywords: particle filters, DEVS-FIRE, state estimation, dynamic systems A fundamental issue in data assimilation of wildfire simulation is to estimate the dynamically changing states, e.g., the fire front position of wildfire, based on observation data of fire sensors. This is a challenging task because of the dynamic and non-linear behavior of fire spread. In this paper, we apply particles filters, also called sequential Monte Carlo methods, to data assimilation in wildfire simulation for estimating the dynamically evolving fire front of a spreading fire. The framework of applying particle filters to the DEVS-FIRE simulation model is presented. Preliminary experiment results show that the particle filtering algorithm was able to track the dynamically changing fire front based on fire sensor data, and thus to provide more accurate predictions of wildfire spread. 1.

### Citations

1060 |
A F M. Novel approach to nonlinear-non-gaussian Bayesian state estimation. IEEProceeding-F
- Gordon, Salmond, et al.
- 1993
(Show Context)
Citation Context ...licate some samples with larger weights as ~[ i] [ j] ( ) ~ [ j] P x = x = q( x ) , i=1, 2, …, N, so as to overcome the limitation that the entire process is dependent on initially generated samples [=-=Gordon et al. 1993-=-]. To summarize, particle filters with sampling and importance resampling have the following steps: 1) sampling – generates a number of particles; 2) compute and normalize their corresponding importan... |

890 | Sequential Monte Carlo Methods in Practice - Doucet, Freitas, et al. - 2001 |

657 | On sequential Monte Carlo sampling methods for Bayesian filtering
- Doucet, Godsill, et al.
- 2000
(Show Context)
Citation Context ... w t are two independent random variables to generate the noises of both of these variables; θ is a set of all other parameters in the system. Applying hidden Markov property and hidden Markov model [=-=Doucet et al. 2000-=-], we can obtain the following formulas, ⎧ ⎪ st + 1 ~ pθ ( st + 1 st ) = pv ( st + 1 − f ( st , θ, t)), t ⎨ ⎪⎩ mt ~ pθ ( mt st ) = pw ( mt − g( st , θ, t)), t which are the basis of state estimation o... |

454 | Sequential monte carlo methods for dynamic systems - Liu, Chen - 1998 |

382 | Understanding the Metropolis-Hastings algorithm - Chib, Greenberg - 1995 |

33 |
FARSITE: Fire area simulator – Model development and evaluation
- Finney
- 1998
(Show Context)
Citation Context ... understand the process of wildfire spread and to estimate fire intensity, direction, location, rate of speed so as to contain the wildfire. Several wildfire models were developed, including FARSITE [=-=Finney 1998-=-], BehavePlus [Andrews et al. 2005], and DEVS-FIRE [Natimo et al. 2008], and Hfire [Morais 2001]. Using these models, the states of wildfire spread at various time steps are computed based on the info... |

23 |
C.: BehavePlus fire modeling system, version 4.0: User’s guide revised
- Andrews, Bevins, et al.
- 2008
(Show Context)
Citation Context ... wildfire spread and to estimate fire intensity, direction, location, rate of speed so as to contain the wildfire. Several wildfire models were developed, including FARSITE [Finney 1998], BehavePlus [=-=Andrews et al. 2005-=-], and DEVS-FIRE [Natimo et al. 2008], and Hfire [Morais 2001]. Using these models, the states of wildfire spread at various time steps are computed based on the information, of spatial fuel data, wea... |

21 |
Particle filters for mobile robot localization. Sequential Monte Carlo Methods in Practice
- Fox, Thrun, et al.
- 2001
(Show Context)
Citation Context ...e found in [Chen 2004]. [Gustafsson et al. 2002] developed a framework for the problems of positioning, navigation, and tracking using particle filters, and presented a couple of general algorithms. [=-=Fox et al. 2001-=-] used particle filters to solve robot localization, one of the important problems of mobile robots. Several other variations of particle filters techniques were implemented in robot localization in s... |

18 |
DEVS-FIRE: Towards an integrated simulation environment for surface wildfire spread and containment, Simulation: Transactions of The Society for Modeling and
- Ntaimo, HU, et al.
- 2008
(Show Context)
Citation Context ... intensity, direction, location, rate of speed so as to contain the wildfire. Several wildfire models were developed, including FARSITE [Finney 1998], BehavePlus [Andrews et al. 2005], and DEVS-FIRE [=-=Natimo et al. 2008-=-], and Hfire [Morais 2001]. Using these models, the states of wildfire spread at various time steps are computed based on the information, of spatial fuel data, weather data and landscape data. Althou... |

16 | Upper Saddle - Hall - 1995 |

15 |
Stochastic processes and filtering theory. Mathematics in science and engineering
- Jazwinski
- 1970
(Show Context)
Citation Context ...More details about these particle filters applications can be found in these papers. 3. OVERVIEW OF PARTICLE FILTERS We use the following formulas to represent a typical non-linear state space model [=-=Jazwinski 1970-=-] ⎧st + 1 = f ( st , θ, t) + vt , ⎨ ⎩mt = g( st , θ, t) + wt , where s t and m t stands for the state variable and the measurement variable of the model respectively; f and g are functions to specify ... |

15 |
Comparing spatially explicit models of fire spread through chaparral fuels: A new model based upon the Rothermel fire spread equation
- Morais
- 2001
(Show Context)
Citation Context ... rate of speed so as to contain the wildfire. Several wildfire models were developed, including FARSITE [Finney 1998], BehavePlus [Andrews et al. 2005], and DEVS-FIRE [Natimo et al. 2008], and Hfire [=-=Morais 2001-=-]. Using these models, the states of wildfire spread at various time steps are computed based on the information, of spatial fuel data, weather data and landscape data. Although these models were used... |

11 | A wildland fire model with data assimilation - Mandel, Bennethum, et al. - 2008 |

8 |
Monte Carlo Filters for Non-linear State Estimation,” Automatica
- Bølviken, Acklam, et al.
- 2001
(Show Context)
Citation Context ...algorithms, containing perfect sampling, importance sampling, acceptancerejection sampling, and Metropolis-Hastings independence sampling, are proposed. More details can be found in [Liu et al. 1998; =-=Bølviken et al. 2001-=-; Chib et al. 1995]. Among all these sampling algorithms, SIR (Sampling Importance Resampling) construct the approximation of the target N [ i] [ i] density function tˆ N ( x) = ∑ q~ ( x ) δ ( x − x )... |

6 | Efficient Particle Filtering Using RANSAC with Application to 3D - Lu, Hager - 2006 |

5 |
Nonlinear Kalman filtering algorithms for on-line calibration of dynamic traffic assignment models
- Antoniou, Ben-Akiva, et al.
- 2007
(Show Context)
Citation Context ... estimate the probability density functions. Generally speaking, these algorithms fall into two categories. The first is to linearize the model, for example, Kalman filters in [Kailath et al. 2000]. [=-=Antoniou et al. 2007-=-] considered three extensions of Klaman filter ,including extended Kalman filter, limited extended Kalman filter, and unscented Kalman filter, to find the solutions to non-linear state space models. ... |

5 | A wildland fire dynamic data-driven application system - Coen, Beezley, et al. - 2007 |

4 |
Particle filter based mosaicking for forest fire tracking
- Bradley
(Show Context)
Citation Context ...ps from burned areas of a spreading wildfire. This paper differs from the previous work by estimating the state of the wildfire spread, e.g., the fire front, from randomly deployed fire sensor data. [=-=Bradley 2007-=-] proposed an approach to estimate and track forest fires based on sequential Monte Carlo methods in video processing. In this implementation, images information from miniature air vehicles was used a... |

4 |
Towards applications of particle filters in wildfire spread simulation
- Gu, Hu
- 2008
(Show Context)
Citation Context ...pendence sampling. The advantages and disadvantages of these techniques can be found in [Simandl et al. 2007]. There are several possible applications of particle filters in DEVSFIRE as described in [=-=Gu and Hu 2008-=-]. In this paper, we apply particle filters to the DEVS-FIRE model for estimating the fire front of a spreading fire based on observation data from fire sensors. The rest of this paper is organized as... |

3 | Evaluation of Farsite Simulator in A - Arca - 2007 |

3 |
Dynamic data rectification using particle filters
- Chen, Morris, et al.
- 2008
(Show Context)
Citation Context ...on of particle filters of biology, in which populations of compact long chain polymers were created by the Monte Carlo methods to study the relationships between packing density and chain length. In [=-=Chen et al. 2008-=-], authors set up the probabilistic framework for the dynamic data rectification, which provides the basis of process fault diagnosis. With the formulation of wildfire spread models, particle filters ... |

3 |
Validation of BEHAVE fire behavior predictions in Oak Savannas using five fuel models
- Grabner, Dwyer, et al.
- 1997
(Show Context)
Citation Context ...ation, of spatial fuel data, weather data and landscape data. Although these models were used to predict and analyze the fire propagation in some real wildfire, which can be validated to some extent [=-=Grabner et al. 1997-=-; Arca et al. 2007; Gu and Hu 2007], they still exist drawbacks. In particular, the wildfire simulation models and their parameters unavoidably have errors or inaccuracies and thus generating deviatio... |

2 |
Application of particle filtering to image enhancement. CERTIS
- Azzabou, Paragios, et al.
- 2005
(Show Context)
Citation Context ... terms of signal strength, by which the Mobile Station’s position and speed can be correctly estimated. Another important application domain of sequential Monte Carlo methods is image processing. In [=-=Azzabou et al. 2005-=-], particle filtering approach was used to improve the image quality. Also, the 3D face pose tracking problem was modeled by particle filters in [Lu et al. 2005] with a small number of particles and w... |

2 |
USDA Testimony to Congress
- Rey
- 2004
(Show Context)
Citation Context ...s of natural forest resources, endangered species, lives of human beings, and properties. It is estimated that more than 11, 000 communities close to federal land are subject to threat from wildfire [=-=Rey 2004-=-]. In the 2007 wildfire season, over 85, 500 fires across the whole US, burned land up to more than 9.3 million acres. It cost 1.8 billion dollars in effort to fight wildfires and a potential 2.5 bill... |

2 |
Estimation of Nonlinear Dynamic Systems–Theory and Applications
- Schön
- 2006
(Show Context)
Citation Context ..., use a series of statistical methods to represent dynamic systems by recursively estimating the probability density function, based on which we compute and estimate their states [Doucet et al. 2001; =-=Schön 2006-=-]. In the practical applications of particle filters, the probability density function is a function of a set of samples, each of which corresponds to a particle and its weight. The ways to generate s... |

1 |
A Modified Particle Filter for Simultaneous Localization and Landmark Tracking
- Hu, Downs, et al.
- 2004
(Show Context)
Citation Context ...ot localization, one of the important problems of mobile robots. Several other variations of particle filters techniques were implemented in robot localization in specific environment. These include [=-=Hu et al. 2004-=-], in which the modified particle filters are the solutions to the simultaneous robot localization in an indoor environment, and [Yang and Atiken 2005] proposed and evaluated uniform clustered particl... |

1 | Mobility Tracking in Cellular Networks Using Particle Filtering - Canagarajah, Ristic - 2007 |

1 | Sampling Densities of Particle Filter: A - Simandl, Straka - 2007 |

1 | Uniform Clustered Particle Filtering for Robot Localization - Yang, Aitken |