Dynamic Fine-Grained Localization in Ad-Hoc Networks of Sensors A. Savvides, C. C. Han, M. B. Srivastava Networked and Embedded Systems Lab University of California, Los Angeles {asavvide, simonhan, mbs}@ee.ucla.edu Localization in Sensor Networks • Context awareness in applications • Network coverage analysis • Report origins of events – Temperature at a specific part of the room – Locate/track objects, people, robots – Assist with routing • Why not GPS? – Costly, power hungry, requires line-of-sight, large form factor, accuracy 2 Problem Statement • Estimate node locations in an adhoc network of nodes Iterative Multilateration – Uniformly deployed nodes on a flat plane • Ad-Hoc Localization System(AHLoS) – Every node contributes to process – Small fraction of nodes (beacons) Collaborative Multilateration are initially aware of their locations – Distributed • Robust to surrounding environment changes and node failures • Energy Efficient • Scalable – Inter-node ranging uses(RSSI, ultrasound) 3 Ranging • Localization relies on the ability of nodes to measure distances Measurement 1 Measurement 2 Multilateration or other Position Estimate Measurement n • Physical layer effects may bias ranging => empirical study – RF Received Signal Strength Indicator (RSSI) – RF + Ultrasound Time-of-Arrival(ToA) 4 Target Platforms Rockwell WINS Node (RSSI) • 200MHz StrongARM • DECT Radio from Connexant Medusa Experimental Node (ToA) • Atmel AVR 8535 MCU • RFM Radio • 40KHz Ultrasound 5 Platform Characterization Ultrasound ToA Max range 3m, accuracy 2cm RSSI in football field Max range 20m, accuracy 7m 6 Localization Algorithms • Atomic Multilateration (base case) – Solution similar to GPS – Formulated as a least squares problem – Requires 3 beacons (if more than 3 beacons are available, the ultrasound propagation speed is also estimated) – May not work if beacons are badly aligned Beacon Unknown 7 Atomic Multilateration 2 Minimize over all 1 f ( xi , x0 , s) sti 0 ( xi x0 ) 2 ( yi y0 ) 2 This can be linearized to the form where x12 y12 xk2 yk2 2 2 2 2 x y x y 2 2 k k y 2 2 2 2 xk 1 yk 1 xk yk MMSE Solution: y Xb 0 i 1,2k 1 4 3 2( xk x1 ) 2( yk y1 ) t k20 t102 2 2 2 ( x x ) 2 ( y y ) t t k 2 k 2 k0 20 b X 2 2 2 ( x x ) 2 ( y y ) t t k k 1 k k 1 k0 ( k 1) 0 b ( X T X ) 1 X T y x0 y 0 s 2 8 Iterative Multilateration • Each node that calculates its location it becomes a beacon that can help other nodes to calculate their locations • Allows Distributed Operation • Problem: – Error accumulation – Reasonable results can be achieved for small networks since ultrasonic distance measurement is accurate • Error accumulation can be limited using weights 9 Iterative Multilateration • Each node that calculates its location it becomes a beacon that can help other nodes to calculate their locations • Allows Distributed Operation • Problem: – Error accumulation – Reasonable results can be achieved for small networks since ultrasonic distance measurement is accurate • Error accumulation can be limited using weights 10 Iterative Multilateration • Each node that calculates its location it becomes a beacon that can help other nodes to calculate their locations • Allows Distributed Operation • Problem: – Error accumulation – Reasonable results can be achieved for small networks since ultrasonic distance measurement is accurate • Error accumulation can be limited using weights 11 Iterative Multilateration Accuracy Ranging + Beacon Error 49 47 44 42 40 38 36 34 31 28 26 24 21 20 18 16 13 0.18 0.16 0.14 0.12 0.1 0.08 0.06 0.04 0.02 0 12 Error Distance (m) Ranging Error Node Id 50 Nodes, 20x20 room, range=3m, 10% beacons 20mm white gaussian ranging error 12 Collaborative Multilateration • Considers location information over multiple hops • More than one unknown node positions are estimated simultaneously • Set of nodes considered MUST have a unique solution 13 Collaborative Multilateration Results 14 Node and Beacon Placement • Nodes are assumed to have a uniform distribution • The success of the iterative multilateration process depends on node connectivity and beacon availability Node range = 10m 15 Node vs. Initial Beacon Densities % Resolved Nodes Total Nodes % Initial Beacons Uniformly distributed deployment in a field 100x100. Node range = 10 Results include only iterative multilateration 16 Experimental Setup • Initially Simulated in ns-2 on top of DSDV • Testbed Implementation • Ultrasound transmitted simultaneously with RF • Distributed Computation 17 Centralized or Distributed? • Where should the computation for location estimation take place? – At a central node? – Inside the network? • How does this decision facilitate – Scalability – Robustness – Energy efficiency 18 Centralized vs. Distributed Localization Distributed Pros • • • More robust to node failure Less traffic => less power Better handling of local environment variations – – • • Speed of ultrasound Radio path loss Rapid updates upon topology changes No time synch. required Centralized Cons • A route to a central point • Time synchronization • High latencies for location updates • Central node requires preplanning • More traffic => higher power consumption Centralized Pros • Can solve more accurately 19 Energy Characterization • Ultrasound penalty is the same for both cases so we did not characterize it • Measured AVR MCU and RFM radio AVR Mode Current Power Active 2.9mA 8.7mW Sleep 1.9mA 5.9mW Power Down 1μA 3μW • Total Power - 20mW 20 Localization Energy Cost Distributed Centralized 8 Energy per node (J) 7 6 5 4 3 2 1 0 100 200 300 400 500 600 700 Netw ork Size Node range 10m, 20% beacons Central node at the center of the network 21 Related Work • Centralized – RADAR [Bahl et. al] – Active BAT [Harter et. al] • Proximity – Cricket System [Priyantha et. al] • Ad-Hoc Distributed Proximity – GPS Less Localization [Bulusu et. al] • Ad-Hoc Centralized – Convex Optimization Methods [Doherty et. al] 22 Conclusions and Future Work • Initial results are encouraging (20 cm accuracy) • A distributed implementation is desirable • This is only the beginning! – Medusa II Node under development • 20 meter ultrasound range • More computation power – New 3D test bed • Collaborative Multilateration is promising and should be further explored • Many new applications are emerging! 23 AHLoS Website http://nesl.ee.ucla.edu/projects/ahlos 24