Wireless Sensor Networks: nodes localization issue Jean Philippe MONTILLET Centre for Wireless Communications University of Oulu, Finland jeanfi@ee.oulu.fi
Summary Introduction Positioning techniques: How to localize tiny sensors nodes in wireless sensor networks Nodes localization issue (error in received signal \ error in space) Beacon nodes and how they can reduce the error on positioning the nodes Results Conclusion 16/01/2019 Event Name
Introduction (1) In a Wireless sensor nodes thousands of sensors need to know their position Many applications need position info: in-home forest-fire detection atmospheric (temperature, pressure, … ) military (target detection, …) police 16/01/2019 Event Name
Introduction (2) Sensors used in WINS (Wireless Integrated Network Sensors) are low-cost, low bit rate, and low-power consumption. Allows mass production no maintenance of sensors 16/01/2019 Event Name
Positioning techniques (1) Why does GPS system not use? Expensive solution to have accuracy <1 m Not work everywhere (dense vegetation) Constraint on the energy … 16/01/2019 Event Name
Positioning techniques (1) Instead of using GPS, various techniques can be used Algorithms are based on: Cooperative ranging: ( TOA, TDOA, RSSI, AOA) Or/and DV-Hop count method Present work based on direct method and quasi-Newtown algorithm 16/01/2019 Event Name
Positioning Techniques (2) Direct Method 16/01/2019 Event Name
Positioning Techniques (3) Equations in 3D: Solution: (i.e. Article for the details of the coefficients) 16/01/2019 Event Name
Positioning Techniques (4) Non linear method (DFP) Where, The solution p is found iteratively: ,with : And Bk is the inverse Hessian 16/01/2019 Event Name
Positioning Techniques (5) Nonlinear Optimization Gauss-Newton Method on Rosenbrock's Function 16/01/2019 Event Name
Nodes localization issue (1) Range estimation degraded due to several factors: Multipath Inaccuracy of the clock Channel fading 16/01/2019 Event Name
Nodes localization issue (2) Error in the received signal Ex:RSSI Xσ : medium scale fading (zero-mean gaussian distributed) 16/01/2019 Event Name
Nodes localization issue (3) Error in space Model the fact that after running algorithm, a residual error is still left. Difficult to model One way is to take a gradient descent error: Or on the coordinates: 16/01/2019 Event Name
Beacon nodes (1) Beacon may help to reduce the error on the node localization. Boundaries: Cost of beacon nodes precludes a dense beacon placement. Very high density self-interferences will appear 16/01/2019 Event Name
Beacon nodes (2) Arithmetic method (u= unknown, b= beacon) 16/01/2019 Event Name
Beacon nodes (3) The more the beacon nodes are around the unknown node which wants to know its position, the less the error is on its coordinate. Idea is through this simple example, To see how the error on the coordinates of the unknown nodes behaves. 16/01/2019 Event Name
Results (1) 1 set of simulations Monitored area size: 70mX70mX10m 16/01/2019 Event Name
Results (2) 16/01/2019 Event Name
Results (3) The graphs with higher number of anchor nodes are not always on the top of those ones with low number of anchor nodes. Maybe due to spatial localization of anchor nodes Second set of simulations: area monitored limited (30mX30mX10m). 16/01/2019 Event Name
Results (4) 16/01/2019 Event Name
Results (5) 16/01/2019 Event Name
Results (6) Same results than in the first set of simulations Maybe due to the fact we chose an error randomly distributed on the coordinates Or beacon nodes are to be put only where the error is maximal (key point) 16/01/2019 Event Name
Results (7) Last set of simulations: to check if the error is gradient descent (as we model in the error in space) Different topology, but only if the nodes are distributed in a ring, spatial localization can be seen. Simulations parameters: 4 anchors nodes (at the centre)\ 100 unknown 16/01/2019 Event Name
Results (8) 16/01/2019 Event Name
Conclusion Two types of error: error in received signal (or time) and error in space To figure out how beacon nodes can help to reduce the error on the position. The gradient descent method to model the error in space shows good results Future work: take into account self-interferences between beacon nodes (MAC layer) 16/01/2019 Event Name
Thanks for coming! 16/01/2019 Event Name