Simulation for a volcano monitoring network Rainer Mautz ETH Zurich, Institute of Geodesy and Photogrammetry November 22nd, 2008 Session 9: Natural hazards and risks Motivation Positioning Algorithm Contents 1. Motivation 2. Positioning Algorithm 3. Simulation Setup 4. Simulation Results 5. Conclusion & Outlook Simulation Setup Simulation Results Conclusions & Outlook Motivation Positioning Algorithm Simulation Setup Simulation Results Conclusions & Outlook 1. Motivation Volcanoes experience pre-eruption surface deformation Reason: internal magma pressure cause surface bulge displacements direction: upwards and outwards horizontal: radial pattern up to 10 cm vertical: uplift of 4 - 6 cm / year (typical) area: over 10 km2 goal Mount St. Helens, Washington spatially distributed position based monitoring system for early warning positioning for spatio-temoral referencing of additional sensors e.g. seismicity, geothermal, gravity, geomagnetic data Motivation Positioning Algorithm Simulation Setup Simulation Results Conclusions & Outlook 1. Motivation SAR interferometry: update rate 35 days Geodetic GNSS: expensive, energy consuming Feasibility of a positioning system with deployed location aware sensor nodes tiny nodes low cost battery-powered self positioning ranging capability high density short range – low power Motivation Positioning Algorithm Simulation Setup Simulation Results Conclusions & Outlook 1. Motivation GPS (anchor nodes) tiny nodes inter-node distances Tiny Node GPS Station Motivation Positioning Algorithm Simulation Setup Simulation Results Conclusions & Outlook 2. Positioning Algorithm Principle of Wireless Positioning: Multi-Lateration known node unknown node range measurement Motivation Positioning Algorithm Simulation Setup Simulation Results Conclusions & Outlook 2. Positioning Algorithm Iterative Multi-Lateration: Initial anchors Step 1: becomes anchor Step 2: becomes anchor Step 3: becomes anchor Motivation Positioning Algorithm Simulation Setup Simulation Results Conclusions & Outlook 2. Positioning Algorithm Ambiguity problem when creating the smallest rigid structure (a) (b) 5 3 1 4 2 (c) 5 3 1 3 2 1 4 4 5’ 2 Motivation Positioning Algorithm Simulation Setup Simulation Results Conclusions & Outlook 2. Positioning Algorithm Positioning Strategy input ranges return refined coordinates and standard variations Creation of a robust structure find 5 fully connected nodes failed achieved failed volume test achieved failed ambiguity test achieved assign local coordinates free LS adjustment Coarse Positioning Transformation into a reference system yes anchor nodes available? input anchor nodes no return local coordinates Merging of Clusters (6-Parameter Transformation) Expansion of minimal structure (iterative multilateration) Motivation Positioning Algorithm Simulation Setup Simulation Results Conclusions & Outlook 3. Simulation Setup Object of study: Sakurajima Stratovolcano, summit with three peaks, island 77 km2 1117 m height extremely active: strombolian, plinian densely populated: Kagoshima, 680.000 on island 7.000 monitored by Sakurajima Volcano Observatory (levelling, EDM, GPS) Landsat image, created by NASA Motivation Positioning Algorithm Simulation Setup Simulation Results Conclusions & Outlook 3. Simulation Setup Sakurajima Mountain – Digital Surface Model 10 x 10 m grid Central part of volcano Area 2 km x 2.5 km Data provided by Kokusai Kogyo Co. Ltd Motivation Positioning Algorithm Simulation Setup Simulation Results Conclusions & Outlook 3. Simulation Setup Parameters for Simulation Parameter Default Value Range Number of tiny nodes 400 100 – 1000 Number of GPS nodes (anchors) 10 1–5% 400 m 200 – 500 m 10 4 - 12 1 cm 0–1m Maximum range (radio link) Inter-nodal connectivity Range observation accuracy Node distribution grid / optimised Motivation Positioning Algorithm Simulation Setup Simulation Results Conclusions & Outlook 4. Simulation Results 400 nodes on a 100 m x 125 m grid. 1838 lines of sight with less than 500 m Motivation Positioning Algorithm Simulation Setup Simulation Results 4. Simulation Results Optimised positions. 5024 lines of sight with less than 500 m Conclusions & Outlook Motivation Positioning Algorithm Simulation Setup Simulation Results Conclusions & Outlook 4. Simulation Results Maximum radio range versus number of range measurements Maximum radio range versus number of positioned nodes 30 400 connected nodes ranges per node 25 20 15 200 100 10 5 200 300 250 300 350 400 maximal range [m] 450 500 0 200 250 300 350 400 maximal range [m] 450 500 Motivation Positioning Algorithm Simulation Setup Simulation Results Conclusions & Outlook 4. Simulation Results Number of located nodes in dependency of the number of anchor nodes Number of anchors Anchor fraction Number of located nodes Success rate Number of ranges 3 0.8 % 3 1% 3 5 1.2 % 191 48 % 3556 10 2.5 % 354 88 % 4553 15 3.8 % 371 93 % 4874 20 5.0 % 400 100 % 5024 Motivation Positioning Algorithm Simulation Setup Simulation Results Conclusions & Outlook 4. Simulation Results Correlation between Ranging Error and Positioning Error mean errors / deviation [m] 4 3.5 3 2.5 2 1.5 1 + true deviation ● mean error (as result of adjustment) 0.5 0 0 0.2 0.4 0.6 noise [m] 0.8 1 Motivation Positioning Algorithm Simulation Setup Simulation Results Conclusions & Outlook 4. Simulation Results Mean errors of the X- Y- and Z-components sorted by the mean 3D point errors (P) 0.16 0.14 mean error [m] 0.12 X Y Z P 0.1 0.08 0.06 0.04 0.02 0 0 100 200 node number 300 400 Motivation Positioning Algorithm Simulation Setup Simulation Results 5. Conclusions Feasibility of a wireless sensor network shown Direct line of sight requirement difficult to achieve 10 % GPS equipped nodes required Error of height component two times larger Position error ≈ range measurement error Outlook Precise ranging (cm) between networks to be solved Protocol & power management Conclusions & Outlook Motivation Positioning Algorithm Simulation Setup End Simulation Results Conclusions & Outlook