Distributed Physicallybased Art and Live Animation on the GRID Presented at Prof. Joe Kearney’s lecture Jun Ni, Ph.D. M.E Research Services, ITS Interactive Kites Flying Shalini Venkataraman, Dept. of CS EVL, University of Chicago NCSA, University of Illinois Outline Introduction to Grid Computing State of the art of high performance computing tele-immersive VR application Motivation and background Physically based model Implementation with VR and no-grid simulation Grid computing based simulation Introduction to Grid Computing Geologically distributed “virtual supercomputer” in virtual organization NSF Middleware NSF and DOE supported globus project, TeriGrid (CalTech, NPACI, ANL, NCSA) (ongoing $54 millions) NSF ITR projects Grids everywhere (next generation of computing) Combination of grid computing together with teleimmersive VR application State of the art of high performance computing tele-immersive VR application on internet Globally network-based Physical model based scientific animation Tele-immersive VR application Art design Motivation and background French sculpter and light artist, Jackie Matisse creates teflon or crepe kites, with artistic tails as long as 15 feet, that can soar through the air, ripple through water, or undulate with the air currents in a room. Randomly influenced by natural forces, the kitetails move, and metamorphose in faint air currents and dramatically changing natural light Motivation and background The VR piece was inspired by the threescreen collaborative video Sea Tails created in 1983 by Matisse with filmmaker Molly Davies. The film follows ten kitetails on their dancing flight through the air and into the water. Physically Based Model To ensure stability, the simulation has to be performed in very small time steps making them very computationally intensive. Implicit approaches to mass-spring systems in the context of VR environments using a grid computing system with its geographically dispersed processors linked by high-speed networks Physically Based Model Each kite is modeled as a cloth object treated as a cluster of masses and springs Using fundamental laws of dynamics to calculate various forces acting on these masses and springs in order to account for the movement of each kite Physically Based Model Mesh Model is introduced to each grid point P(i,j) and each point has its mass and linked to neighboring points Position x(i,j) obeys dynamic laws dx(i,j)/dt = F(i,j)/m(i,j) Newton’s second law Discretize dynamic law x (i,j) t+dt = x(i,j) t + t v(I,j) t+dt Physically Based Model Internal and external forces acting on each point of grids Internal forces: structural, shearing and bending forces Fin(i,j) = k (Lt – Lo)[ P(i,j)-P(k,l) ] Elasticity Physically Based Model Internal and external forces acting on each point of grids External gravitational force Fg(i,j) = m(i,j) g Gravitational acceleration Physically Based Model Internal and external forces acting on each point of grids Wind forces Fw(i,j) = m n(i,j) [ w – v(i,j) ] n (i,j) Air or fluid viscosity Physically Based Model Internal and external forces acting on each point of grids viscous forces Fw(i,j) = - m n(i,j) v(i,j) Damping coefficient Implementation with VR CAVE VR environment EVL’s CAVE CAVE Library Implementation with VR Standalone kite Texture mapped onto the kitetail mesh User can use wand to grab on the kite head and move or change its imagery Wind direction is controlled by wand orientation (constant wind speed) Head controlled by wand in CAVE system Implementation with VR Standalone kite Other properties such as stiffness, length, width and visual attributes like texture maps can be specified at the rum-time by user Each kite dimension is 2 ft by 30 ft in virtual space modled by 250 masspoints Head controlled by wand in CAVE system Implementation with VR Standalone kite (no grid) Simulation rate for on kite takes 125 iteration per second. Each iteration takes 8 ms. In 3-kite simulation, each kite has 41 ms/s. Small time step makes more stable but more computer intensive SGI ONYX Inifite Reality with 8 198 MHz MPIS R10000 processors and 2G memory. Head controlled by wand in CAVE system Grid computing based simulation Distributed simulation Small time steps Grid enhanced High-speed network based Architecture of gird enhanced application to kite simulation Grid Computing Based Simulation Distributed simulation Configure several simulation nodes globally distributed QUANTA middleware (collection of network programming tools for optimizing data sharing over high-speed networks) Grid Computing Based Simulation Distributed simulation kiteServer (database server for wind direction as a 3float array; any user interaction results will be received and broadcast to other nodes) kiteSim (simulation server for computing each kite’s position and directly transmitted through UDP socket to dispply client running in CAVE system) kiteDisplay (client) Implementation (displays the kitetails and userinteraction. The kite positions will read from kiteServer and display texture mapped with images Grid Computing Based Simulation Results Distributed simulation rate (1000iterantions/s) is significant higher than standalone simulation (125 iterations/s) Simulation rate is dependent of the number of kites due to network bandwidth. With increasing number of kites, simulation rate approaches to constant. Grid Computing Based Simulation Discussion Network latency Interactions among kites Fluid models Communication between kites Virtual space for flying aircrafts (Jun Ni’s proposal) using physically based mathematical models in CFD fro fluid flow along each craft and deformable body model for each object of craft Interactive sound tracks What about your suggestions? Reference http://www.evl.uic.edu/research/template_res_ project.php3?indi=231