Transcript Slide 1
An Empirical Characterization of Radio Signal Strength Variability in 3-D IEEE 802.15.4 Networks Using Monopole Antennas Dimitrios Lymberopoulos, Quentin Lindsey and Andreas Savvides Embedded Networks and Applications Laboratory (ENALAB) http://www.eng.yale.edu/enalab Yale University Can RSSI provide reliable distance estimation? Quantify variability in typical office environments 3-D deployments Low power radios What other type of information can RSSI provide? More than 150.000 measurements were acquired 40 wireless sensor nodes were used Acquired data along with ground truth data are available at: http://www.eng.yale.edu/enalab/rssidata EWSN 2006 February 15th Dimitrios Lymberopoulos Background MAP based approaches (RADAR, Bahl et. al) Create a database of RSSI fingerprints: < [RSSI], Position > Find the fingerprint with the minimum distance to the recorded RSSI array 15ft error using 802.11 wireless radios RSSI distance prediction (Ecolocation, Yedavalli et. al) Use ordering or triangulation to refine the initial estimates 10ft error in a small indoor experiment with CC1000 wireless radios Probabilistic approaches (Madigan et. al) Every node computes a belief about its location A probabilistic signal propagation model is assumed 20ft error using 802.11 wireless radios EWSN 2006 February 15th Dimitrios Lymberopoulos Infrastructure XYZ sensor node designed at Yale (http://www.eng.yale.edu/enalab/XYZ) CC2420 wireless radio from Chipcon 2.4 GHz IEEE 802.5.14/Zigbee-ready RF transceiver DSSS modem with 9 dB spreading gain Effective data rate: 250 Kbps 8 discrete power levels: 0, -1, -3, -5 , -7, -10, -15 and -25 dBm Power consumption: 29mW – 52mW Monopole antenna with length equal to 1.1inch. EWSN 2006 February 15th Dimitrios Lymberopoulos Received Signal Strength Indicator (RSSI) The power P at the input RF pins can be obtained directly from RSSI: P = RSSI + RSSIOFFSET [dBm] RSSI is an 8-bit value computed by the radio over 8 symbols (128μs) RSSIOFFSET is determined experimentally based on the front-end gain. It is equal to -45dbm for the CC2420 radio Sources of RSSI Variability Intrinsic Radio transmitter and receiver calibration Extrinsic Antenna orientation Multipath, Fading, Shadowing EWSN 2006 February 15th Dimitrios Lymberopoulos Path Loss Prediction Model Log-normal shadowing signal propagation model: RSSI(d) = PT – PL(d0) – 10ηlog10(d/d0) + Xσ -20 RSSI(d) is the RSSI value recorded at distance d PL(d0) is the path loss for a reference distance d0 η is the path loss exponent Xσ is a gaussian random variable with zero mean and σ2 variance -25 RSSI (dbm) PT is the transmission power Averaged RSSI values log-fit -30 -35 -40 -45 0 5 10 15 20 25 Distance(feet) Model verification using data from a basketball court EWSN 2006 February 15th Dimitrios Lymberopoulos Radio Calibration For each location and orientation 20 packets were sent @ -15dBm 090180270- 090180270090180270- 090180270- Receiver 1.31ft Transmitter 090180270EWSN 2006 090180270- 090180270- February 15th 090180270- Dimitrios Lymberopoulos Radio Calibration For each location and orientation 20 packets were sent @ -15dBm 090180270- 090180270090180270- 090180270- Receiver EWSN 2006 Transmitter 1.31ft 090180270- 090180270- 090180270- February 15th 090180270- Dimitrios Lymberopoulos Radio Calibration For each location and orientation 20 packets were sent @ -15dBm 090180270- 090180270090180270- 090180270- Receiver 1.31ft 090180270EWSN 2006 Transmitter 090180270- 090180270- February 15th 090180270- Dimitrios Lymberopoulos Radio Calibration For each location and orientation 20 packets were sent @ -15dBm 090180270- 090180270- 090180270EWSN 2006 090180270- Receiver Transmitter 090180270- 090180270- 1.31ft 090180270- February 15th 090180270- Dimitrios Lymberopoulos Radio Calibration Experiment in an empty room TX calibration: 9 different transmitters RX calibration: 6 different receivers TX Standard Deviation: 2.24dBm 30 30 30 25 25 25 25 20 20 20 15 20 15 10 10 5 5 1 2 3 4 5 6 7 Transmitter ID 8 0 9 RSSI(dbm) 30 RSSI(dbm) 35 0 15 10 1 2 3 4 5 6 7 Transmitter ID 8 0 9 15 10 5 5 180 Degrees 1 2 4 3 Receiv er ID 0 5 30 30 25 25 25 25 20 20 20 15 10 10 5 5 0 1 EWSN 2006 2 3 4 5 6 7 Transmitter ID 8 9 0 RSSI(dbm) 30 RSSI(dbm) 30 15 15 10 2 3 4 5 6 7 Transmitter ID 8 9 February 15th 0 4 3 Receiv er ID 5 15 10 5 5 1 2 270 Degrees 35 20 1 180 Degrees 270 Degrees 35 RSSI(dbm) RSSI(dbm) 90 Degrees 0 Degrees 90 Degrees 35 RSSI(dbm) RSSI(dbm) 0 Degrees RX Standard Deviation: 1.86dBm 1 2 4 3 Receiv er ID 5 0 1 2 4 3 Receiv er ID 5 Dimitrios Lymberopoulos Antenna Characterization Experiment took place in a basketball court Minimize multipath effect At each measurement point 20 packets @ -15dBm were received Side View Top View 8ft 2ft 6.5ft 3.5ft 2ft 1.25ft 2ft : measurement point EWSN 2006 February 15th Dimitrios Lymberopoulos Antenna Characterization -5 Optimal antenna length-1.1inch Large communication range Suboptimal antenna with 2.9inch length -15 -20 RSSI (dbm) Random RSSI values due to multipath Optimal Antenna Suboptimal Antenna -10 -25 -30 -35 -40 -45 -50 0 2 4 6 8 10 12 14 16 Distance (ft) EWSN 2006 February 15th Dimitrios Lymberopoulos Antenna Characterization -36 -40 -42 -20 0 45 90 135 180 225 270 315 -32 -34 -36 RSSI (dbm) -38 RSSI (dbm) -30 0 45 90 135 180 225 270 315 0 45 90 135 180 225 270 315 -25 RSSI (dbm) -34 -38 -40 -30 -35 -42 -44 -44 -40 -46 -46 -48 5 10 15 20 Distance(feet) 1.25ft 25 30 -48 0 5 10 15 Distance(feet) 3.5ft 20 25 30 -45 0 5 10 15 20 25 Distance(feet) 6.5ft Similar distances (<1ft difference) can produce very different RSSI values (even up to 11dBm) Very different distances ( even >18ft) can produce the same RSSI values EWSN 2006 February 15th Dimitrios Lymberopoulos Antenna Characterization -20 -25 6.5ft 3.5ft 1.5ft -30 RSSI (dbm) RSSI (dbm) -25 -30 -35 -40 -45 6.5ft 3.5ft 1.5ft -35 -40 -45 0 5 10 15 20 25 -50 Distance(feet) 0 5 10 15 20 25 Distance(feet) Best antenna orientation Worst antenna orientation Antenna orientation effect For a given height of the receiver very different RSSI values are recorded for different antenna orientations EWSN 2006 February 15th Dimitrios Lymberopoulos Radiation Pattern Side View Top View Communication range Symmetric Region Communication range EWSN 2006 February 15th Antenna orientation independent regions Dimitrios Lymberopoulos Antenna Effects in Indoor Environments The basketball court experiment was performed inside our lab We focused on the best antenna orientation -20 6.17ft 5.65ft -25 4.6ft RSSI (dbm) 1.25ft -30 -35 -40 -45 -50 0 2 4 6 8 10 12 14 16 Distance (ft) EWSN 2006 February 15th Dimitrios Lymberopoulos Large Scale Indoors Experiment 40 nodes were placed on the testbed (15ft (W) x 20ft(L) x 10ft(H)) installed in ENALAB Each node transmitted 10 packets at each one of the 8 power levels. The recorded RSSI values were transmitted to a base station for logging. Placement and Connectivity Connectivity at Power Level 7 Connectivity at Power Level 4 13 13 15 15 16 16 11 Z coordinate 2 18 2.5 9 14 17 10 26 24 4 3 7 22 1 41 0.5 5 1 39 40 38 28 3 2 7 22 1 41 33 27 31 6 3 Y coordinate X coordinate February 15th 6 5 37 4 3 2 1 2 0 27 35 2 1 28 33 36 3 5 30 34 38 4 1 39 40 4 37 2 5 42 29 1 0 5 36 3 0 4 20 0.5 35 EWSN 2006 24 1.5 8 6 25 32 30 34 4 Y coordinate 18 31 32 0 5 42 29 10 26 2 2 9 14 17 23 6 25 20 21 8 Z coordinate 21 23 1.5 11 19 19 2.5 12 12 1 0 0 X coordinate Dimitrios Lymberopoulos Large Scale Indoors Experiment Maximum (0dBm) Medium (-5dBm) Low (-15dbm) RSSI does not change linearly with the log of the distance Multipath 3-D antenna orientation EWSN 2006 February 15th Dimitrios Lymberopoulos Link Asymmetry Asymmetric link between nodes A and B RSSI(A) ≠ RSSI(B) One way links Asymmetric links 36 55 Percentage of assymetric links Percentage of One-Way Links One way Links 34 32 30 28 26 24 22 50 45 40 35 >=2 >=3 30 >=4 >=5 25 >=6 20 1 2 3 4 5 6 7 8 Power Level (1= Maximum) EWSN 2006 20 1 2 3 4 5 6 7 8 Power Level (1 = Maximum) February 15th Dimitrios Lymberopoulos What else can we do? Detection of: Human presence Human motion More than 30% of the links are affected by human presence or motion EWSN 2006 February 15th Dimitrios Lymberopoulos Conclusions Radio calibration has minimal effect on localization 3-D space is very different that 2-D space Antenna orientation effects are dominant in 3-D deployments 3-D deployments are a more realistic for evaluating RSSI localization methods RSSI distance prediction in 3-D deployments is almost impossible Ordering of the RSSI values is not helpful Even if antenna orientation is known! Probabilistic approaches A probabilistic model of RSSI exists for the symmetric region of the antenna Generalizing this model to 3-D deployments is extremely difficult if not impossible. EWSN 2006 February 15th Dimitrios Lymberopoulos Useful Lessons Learned Dynamic Loadable Binary Modules Dynamic Loadable Binary Modules Matlab interface to the network to: Wire up multiple services to create user specific services Static SOS Kernel Module Hardware Abstraction Communication Log data from the network Memory Manager Push data to the network Outdoor space Corridor Lab Offices Other XYZ Becton Center IT support To Davies Auditorium AKW #000 ENALAB MTC LAB Professor’s Kuc Lab Machinery Room Loading Dock Ed Jackson MTC LAB EWSN 2006 February 15th Dimitrios Lymberopoulos THANK YOU!