# An FPGA processor for modelling wildfire spreading ---  --- ## Wildfires are important disasters - Country people lose their homes - Rare animal species are threatened - The number of sources of oxygen is decreased → the greenhouse effect --- ## Computational fire spread models is important. wind and ground slope affect how a fire will develop through time. differences in fuel beds between parts of the landscape also affect the way fires spread. To predict the spreading of a fire is very difficult. A fire front. **help optimize the response of the fire fighters** --- ## Related research partial differential equations(PDE) - Rothermel's equations - The BEHAVE system - The FARSITE model --- ## PDEs problems - 扱いがむずかしい - 計算速度の問題 --- ## CA approach  --- ## CA approach CA with square cells, that uses weight factors to describe the effect of different kinds of fuel, wind and slope. the accuracy decreases somehow in more complex cases 要素が足りない、精度が低い The model proposed in this paper, can be considered inspired by the ones already described above, but features a number of changes and additions compared to them, aimed to make the model faster and less resource-demanding, while remaining reasonably accurate in the representation of reality --- ## CA problems the time a wildfire needs to spread < the time needed for execution over a wide area --- ## Real-time fire spread model The aim of this work is to formulate a CA-based fire spread model focusing on implementing an accurate, real-time yet light on resources approach to wildfires. The result is an algorithm directly implemented in hardware, able to yield useful results in a short time. --- ## Using FPGA Furthermore, because of the inherent parallelism of CAs, the proposed model is hardware implemented with the help of the Very High Speed Integrated Circuit (VHSIC) Hardware Description Language (VHDL) synthesizable code in order to speed up the application of CAs to the study of wildfire spreading. More specifically, a translation algorithm is used, that checks the CA parameters values previously determined by the user and automatically produces the synthesizable VHDL code that describes the aforementioned CA. It should be mentioned that CAs are one of the computational structures best suited for hardware realization. he CA architecture offers a number of advantages and beneficial features such as simplicity, regularity, ease of mask generation, silicon-area utilization, and locality of interconnections --- ## Cellular automata - space and time are discrete - interactions are local --- ## A CA is characterized by 1. the number of spatial dimensions (n); 2. the width of each side of the array (w). wj is the width of the jth side of the array, where j = 1, 2, 3, . . . , n; 3. the width of the neighbourhood of the cell (d). dj is the width of the neighbourhood along the jth side of the array; 4. the states of the CA cells; 5. the CA rule, which is an arbitrary function F . --- ### CA vs PDE complicated boundary and initial conditions CAs run quickly on digital computers serial computers because they exploit the inherent parallelism of the CA structure --- ## Fire spread model is to synthesize a model maintaining as much of the functionality of the aforementioned model as possible, while concentrating on lowering computation time and resource needs, making it quick to run as real-time model and power-efficient. The goal of this effort was to design a model able to be the core around which a mobile device can be designed, where computational resources are limited and power consumption is critical --- ## Model overview hexagonal lattice. approaching the shape of a circle much better than square cells cell value: spread rate - the fuel in the cell - the wind in the cell - the slope of the ground uphill wind --- ## The wind and the slope $$ \phi_{s} = 5.257 \times \tan a^2 $$ $$ \phi_{w} = 0.276 \times p_{wr} $$ $$ R_{max} = R_0(1+\phi) $$ $$ \vec{\phi} = \vec{\phi_w} + \vec{\phi_s} $$ 楕円の画像 --- ## The corresponding cellular automaton model fire そのセルの火の強さ A is deducted from B fuel 60度の平均がR --- ## Calculation of the next state of the CA wind state of the cell $$ fire \times R(\theta) $$ ランダムに2つ以上離れた場所も燃える enable フラグは、1つ飛びで火がくるイベント --- ## FPGA realization CA consist of a uniform structure composed of many finite state machines matching the inherent design layout of FPGA hardware 計算速度 電力消費などで優位に4 ユーザーはMatlabインターフェースにパラメーターを入力する --- ## Simulation results --- ## Comparison between hardware and software 小数点以下の計算精度の違いにより、software と hardware に差がでた ---  ---  --- ### Fire starting on a hillside  --- ### Fire starting on a hillside  --- ### Fire starting inside a pit  --- ### Fire spreading over an area with obstacles  --- ### Simulation conclusions software: 2.5s hardware: 1μs as the fire front progresses the calculations needed to be carried out increase exponentially, and therefore the serial calculation of the proposed algorithm will eventually take too long to perform. On the other hand, the hardware implementation will take one clock cycle regardless of the size of the matrix and the spread of the fire, and the complexity of the under study conditions, fully exploiting its inherent parallelism. --- ### Conclusions portable system with GPS Note: hello