Project Overview. ⢠AG channel measurement/modeling for UAS. ⢠Dual-band SIMO measurements 2012-2013. 6. Lockheed S-3B Viking. Transportable Tower ...
UNIVERSITY of SOUTH CAROLINA Department of Electrical Engineering
Air-Ground Channel Measurements & Modeling for UAS David W. Matolak Professor
Ruoyu Sun Ph.D. Student
24 April 2013 University of South Carolina
Outline • Introduction
• Project motivation/overview • Measurement campaign
• Measurements to Models • Initial Results
• Conclusion
University of South Carolina Ca aroli ro olilina na
2
Introduction • UAS expected to see INCREASING use – Government, industry, academia studying UAS integration into the NAS
• UAS missions & operation conditions can be markedly different from traditional, e.g., – Low altitude & elevation angle – Ground clutter
• Consequence: air-ground channel will be important to reliable performance University of South Carolina
3
Motivation • UAS control & non-payload communications (CNPC) must be fast, reliable
• Channel effects can be performance limiting – Obstructions → blockage, multipath propagation delay dispersion & frequency selectivity – Mobility causes Doppler time variation
• For evaluating any potential waveforms (air interface), quantitative channel models needed University of South Carolina
4
Motivation (2) • No comprehensive, validated, wideband models exist for time-varying AG channel
• Existing measurements
- Sparse, & for different frequency bands - For widely different environments - Not parameterized as function of - Elevation angle - GS antenna height & beamwidth - GS local environment
- Do not include airframe shadowing If channel not quantified, signal design suboptimal University of South Carolina
5
Project Overview • AG channel measurement/modeling for UAS • Dual-band SIMO measurements 2012-2013
Lockheed S-3B Viking
Transportable Tower System • 7 kW Generator • Pneumatically Extendable Tower (to ~20 m) • Cabinet for Test Equipment University of South Carolina
6
Project Goals 1. Conduct AG channel measurements in two bands, simultaneoulsy – In multiple GS environments – With two Rx antennas in each band
2. From these measurements, develop empirical channel models
University it off South S th Carolina C li
7
Measurement Plan: NASA GRC • Dual-band SIMO AG measurements – – – –
Over water (CLE, CA) Flat: Urban, Suburban, Rural, Desert, Forest Hilly: Urban, Suburban, Rural, Desert, Forest Mountainous (Rockies)
• Spread-spectrum stepped correlator, collect/compute – Time-varying CIRs – Path loss, shadowing, Doppler, correlations
University of South Carolina
8
Test Sites w/Frequency Authorizations
NTD – Point Mugu NAS PMD – Palmdale
BDU – Boulder Municipal TEX – Telluride Regional IOW – Iowa City University of South Carolina
CLE – Cleveland BKL – Burke Lakefront, Cleveland UNI – Ohio University Airport LBE – Westmoreland County (Arnold Palmer)
Channel Sounder Structure, Specs Aircraft
GS Transmitter L-band Tx
C-band Tx
to Transmit Antennas Band Signal Bandwidth (MHz) L 5 C 50
Receiver 1 L-band Rx
Receiver 2
C-band Rx
L-band Rx
from Receive Antennas Frequency Span (MHz) 960-977 5000-5100
• PTx~10 W, “blade” antennas • C-band: HPA G~10 dB, LNAs G~30 dB • Rrep,max~ 3 kHz University of South Carolina
C-band Rx
from Receive Antennas Band L L C C
DS-SS Sequence Length N 511 1023 511 1023
WMAX (Ps) 102.2 204.6 10.2 20.4
10
Channel Sounder (2)
Sounder Tx
Sounder Rx (1 of 2)
University of South Carolina
11
Channel Modeling Options A. Stored CIRs obtained from measurements • Stored samples “played back” in simulation • Depending on accuracy & resolution, samples represent real channel conditions for given setting • Drawback: only limited sets of measured data B. High-frequency approximations (e.g., ray optics) • Assumes all objects in environment have dimensions large w.r.t. O unable to model diffuse scattering • Requires large environment database that includes all object dimensions, locations, & electrical properties (HVP), computation time large 12 University of South Carolina
Channel Modeling Options (2) C. Stochastic models • Typically in TDL form, & traditionally assume widesense stationarity (WSS) • Recent models circumvent WSS limitation & employ non-stationary elements to improve realism • Generally most efficient in implementation D. Parametric stochastic models • Generalizations of stochastic TDL, parameterizing CIR features themselves – e.g., statistical distributions for MPC delays, #MPCs, angles of arrival University of South Carolina
13
Channel Modeling “Process” Geometry Flight Paths (d, T… ) (& attitudes) GS Frequency Environment Features Band Type Desired Model Features
Time Duration
• Measurements • Data processing • Validation
LOS & Ground Ray Computations
Obstruction Attenuation Model(s)
MPC Model(s) University of South Carolina
AG Channel Model: Time-Domain Samples
14
TDL Channel Model: h(W;t) = CIR h(W ; t )
L 1
¦ z (t )D k
k
(t ) exp{ j[ZD,k (t W k (t )) Zc (t )W k (t )]}G [W W k (t )]
k 0
zk(t) = persistence process ({0,1}) to mimic MPC “birth/death” from Tx
E(t)
sk
×
D0 ( t )e jI ( t ) 0
W1
sk-1
D1( t )e jI ( t ) 1
(shadowing) z0(t)
×
z1(t)
W2
WL
…
zL-1(t)
× +
to Rx
University of South Carolina
D L1( t )e jI
rk
¦
L 1
sk-L+1
L1 ( t
)
×
s h (t )
i 0 k i i
15
Initial Results • Laboratory tests
Table 2. C-band back-to-back single-path RMSDS statistics (ns), for data of 11 Feb 2013, Rx 1.
RMS-DS Statistic
-6
5
Multipath Threshold (dB) 25
Analytical Mean 9.7 Median 9.7 Max 9.8 Min 9.5 Standard 0.12 deviation
30
35
40
10.0 9.9 10.3 9.7
8.3 10.2 10.2 10.5 10.0
10.4 10.4 10.7 10.1
0.22
0.16
0.18
x 10
Back-to-Back Test Box 1, L-band 390th PDP I values No threshold applied
4
I values in Linear scale
– Back to back – One or two-path channels
University of South Carolina
3
2
1
0
-1 30
40
50
60
70
80
90
100
110
120
Sample Index (Total domain 0 to 2045)
16
Example Power Delay Profile (PDP) 0
B2B 15Nov L-band Box 1 Single path test 390th PDP Full delay span
Normalized Power (dB)
-10
-20
-30
-40
-50
-60
0
200
400
600
800
1,000
1,200
1,400
1,600
1,800
2,045
Sample Index (Total domain 0 to 2045)
• L-band (Rx1), GRC lab, November 2012 • Power vs. delay (0 to 204.6 Ps) University of South Carolina
17
Initial Results: Lab 2-path Tests
Table 3. C-band back-to-back 2-path RMSDS statistics (ns), for both C-band receivers, data of 11 March 2013 (25 dB threshold).
RMS-DS Statistic Analytical Mean Median Max Measured Min Standard deviation
Rx 1 49.0 48.7 48.7 52.0 42.8
Rx 2 49.0 49.0 49.0 51.5 46.3
1.3
0.7
University of South Carolina
18
Analytical: Two-Ray Model l1 l2
A I Te
v
B
E
hF( s ) (W ; t ) D 0 (t )G (W W 0 ) D g (t )e
j 2S'Rk / O
Flat Earth
*(t )G (W W g )
§ ¨D ¨ 1,k ©
p
1/ 2
· ¸ ¸ ¹
q=T1+T d=kaq k=4/3 a=earth radius C
University of South Carolina
ª " 1,k " 2,k º 2 «1 » ¬« ka sin(\ k ) " 1,k " 2,k ¼»
Curved Earth 19
CLE Flight Test 140
Path Loss (dB)
130
120
110
C-band Free Space C-band Flat Earth C-band Curved Earth L-band Free Space L-band Flat Earth L-band Curved Earth
100
5 Dec 2012 Flight Path 90 6
x 10
0.2
4.215
0.6
0.8
1
1.2
1.4
1.6
1.8
2
Tx-Rx distance (meters) Rx (Airplane) Tx (Ground)
4.21
Z
0.4
4
x 10
Fig. 13. C- and L-band analytical two-ray model path loss & fits, vs. distance.
4.205 4.2 4.195 -4.73 -4.735 6
x 10
6.9 6.8
-4.74
6.7
-4.745
Y
-4.75
6.6 6.5
X
5
x 10
University of South Carolina
20
Analytical Results: CLE Flight 90 80
70
60 50 40 30
60 50 40 30
20
20
10
10
0
0.5
1
1.5
Tx-Rx distance (m)
Vertical Flat Horizontal Flat Vertical Curved Horizontal Curved
80
RMS delay spread (ns)
70
RMS delay spread (ns)
90
Vertical Flat Horizontal Flat Vertical Curved Horizontal Curved
0
2 4
x 10
C-band
0.5
1
1.5
Tx-Rx distance (m)
2 4
x 10
L-band
• 2-ray RMS delay spread VW DDW when CIR W0=0, and energy normalized to 1 University of South Carolina
21
Computed Doppler Shift (LOS path) 285
1500
Doppler shift in L-band Max Doppler shift in L-band
280
1450
Doppler shift (Hz)
275
Doppler shift (Hz)
Doppler shift in C-band Max Doppler shift in C-band
270 265 260
L-band
255 250
1400
1350
C-band
1300
245 240
21.5 km
21.2 km 10
20
30
40
50
60
70
80
90
100
110
1250
21.5 km
21.2 km 10
20
30
40
50
60
70
80
90
100
110
Time (second)
Time (second)
• Computed from measured GPS • “Measured”=“max” since aircraft flying away from GS during this segment of flight University of South Carolina
22
Received Power -48
Total power in data file (dBm) sum(I2+Q2)(dBm)
Total power in data file (dBm)
-49 -50 -51 -52 -53 -54 -55 -56 -57
21.2 km 0
500
21.5 km 1,000
1,500
2,000
2,500
3,000
3,500
4,000
4,500
5000 5326
PDP Index
• Sounder provides two (equivalent) methods of estimating PRx • Near-perfect agreement with link budget! Pr = Pt + Gt + Gr - Lc - Lfs University of South Carolina
23
Next Steps • Software “engineering” data files • Flight tests – CLE, LBE – VBG/EDW – Telluride • Data analysis • Model Development • Integration w/UAS network simulations University of South Carolina
24
Conclusion • Accurate AG channel models important to UAS • Dual-band, SIMO measurement system developed for AG channel measurements • Flight tests Spring/Summer 2013 (+?) in multiple environments • Empirical AG channel model development following data analysis – Integration of channel models into net simulations University of South Carolina
Cleveland Hopkins
25