Download as pdf or txt
Download as pdf or txt
You are on page 1of 11

2018 IEEE 26th International Conference on Network Protocols

Networking Support For Physical-Layer


Cross-Technology Communication
Shuai Wang† , Zhimeng Yin† , Zhijun Li‡ , and Tian He†
† Department of Computer Science and Engineering, University of Minnesota, U.S.
‡ School of Computer Science and Technology, Harbin Institute of Technology, China.
{wang3367, yinxx283, tianhe}@umn.edu, lizhijun.hit@gmail.com

Abstract—Recent research on physical layer cross technology ment on existing infrastructure. Most importantly, PHY-CTC
communication (PHY-CTC) brings a timely answer for escalated provides the communication support for explicit message (e.g.,
wireless coexistence and open spectrum movement. PHY-CTC channel, device and sensing information) exchange among
achieves direct communication among heterogeneous wireless
technologies (e.g.,WiFi, Bluetooth, and ZigBee) in physical layer heterogeneous technologies and thus brings great benefits to
and thus brings communication support for coexistence service upper layer applications including, but not limited to channel
such as spectrum management and IoT device control. To put coordination and IoT device management/control.
PHY-CTC into service, however, there still exists a gap due to To obtain these benefits for upper layer applications, we
its transmission failure and asymmetric link (i.e., one-way PHY- have to address several unique challenges introduced by PHY-
CTC) issues. In this paper, we propose NetCTC – the first net-
working support design for PHY-CTC to establish feedbacks (e.g.,
CTC. First, PHY-CTC is facing asymmetric link problem due
ACKs) and thus meet the upper layer networking requirements to emulation capability. For example, the WiFi commercial
in heterogeneous unicast, multicast and broadcast. The core device is able to emulate ZigBee signals but not vice ver-
design of NetCTC is a real-time interaction mechanism which sa [1]. Second, the signal emulation technique in PHY-CTC
achieves reliable, transmission efficient and parallel interactive inevitably introduces emulation errors which further leads to
communication among heterogeneous devices. We implement and
evaluate NetCTC on commodity devices (Laptops with Atheros
transmission failures. Note that an ACK/ARQ between hetero-
AR2425 WiFi NIC, smart phones with Broadcom BCM4330 WiFi geneous devices is challenging because of the heterogeneity
chip and MicaZ CC2420) and the USRP-N210 platform. Our as well as the asymmetric link problem. The upper layer
extensive evaluation demonstrates that NetCTC achieves reliable applications may not even know which packet is lost.
bidirectional cross technology communication under a full range In this paper, we aim to conquer the above challenges and
of wireless configurations including stationary, mobile and duty-
propose NetCTC – a networking support design for PHY-
cycled settings.
CTC to meet the requirements of upper layer protocols (e.g.,
I. I NTRODUCTION unicast, multicast, and broadcast). In a nutshell, NetCTC is
Wireless technologies have been widely used in applications an interaction mechanism between heterogeneous technologies
such as personal communication, mobile internet surfing, which consists of three components. First, NetCTC introduces
and smart home automation. To accommodate requirements randomness in PHY-CTC packets to eliminate packet se-
from these different applications (e.g., range, throughput, and quence. It encodes PHY-CTC packets with Fountain codes [3],
energy), multiple heterogeneous technologies including WiFi, [4] and continuously transmits the rateless stream of encoded
BlueTooth and ZigBee have been proposed and achieved great packets to receivers, and thus reducing the number of ACKs
success in their respective applications. With the continuous from one ACK for each packet to one ACK for a batch of
growth of wireless devices, however, these technologies now packets. Second, since acknowledgements are necessary in
increasingly become victims of their own success since they cross technology communication scenarios, a one-bit CTC-
are largely ignorant of each other. This leads to operations ACK is designed to notify heterogeneous sender whether
that are non-cooperative, or even disruptive to each other. For the receiver receives enough packets for successful decoding.
example, coexisting wireless technologies compete with each Finally, a dynamic CTC link model is proposed based on
other to preempt channels and they destroy packets of others the feedback of one-bit CTC-ACK. This link model helps
with their own, causing significant spectrum inefficiency. heterogeneous senders decide the optimal stop condition to
To address the coexistence issue among heterogeneous tech- eliminate both unnecessary (re)transmissions and ACKs.In
nologies, researchers propose Physical-layer Cross Technology summary, our technical contributions are as follows:
Communication (PHY-CTC), which builds direct communi- • Leveraging upon physical layer CTC, we propose
cation among heterogeneous technologies via signal emula- NetCTC, the first networking support for heterogeneous
tion, despite their incompatible physical-layer modulation [1], networks. NetCTC addresses the transmission failure and
[2]. The PHY-CTC technique requires neither hardware nor one-way communication issue of PHY-CTC and makes
firmware changes in commodity devices, which eliminates ton- PHY-CTC meet the requirements of upper layer proto-
s of multi-radio gateways and promises zero-cost fast deploy- cols including transmission efficiency and reliability in

978-1-5386-6043-0/18/$31.00 ©2018 IEEE 259


DOI 10.1109/ICNP.2018.00042
unicast, multicast and broadcast, which is an important etc.), despite their incompatible physical-layer modulation.
step towards real-world applications. The core technique of PHY-CTC is software-based signal
• We address a few unique challenges in networking sup- emulation, which utilizes one wireless technology’s signal
port of CTC for upper layer applications, which include (e.g., WiFi signal) to emulate another’s (e.g., ZigBee signal)
(i) packet sequence elimination using Fountain codes, without hardware or firmware changes. The pioneer research
(ii) closed-loop feedback with one-bit CTC-ACK, (iii) WEBee [1] demonstrates that emulated signals from WiFi
transmission control with CTC link modeling, (iv) one- can be successfully detected and decoded by ZigBee. As
to-many CTC covering multiple receivers, and (v) con- shown in Figure 1, WEBee constructs the payload of a WiFi
current feedbacks from multiple receivers. These tech- frame elaborately so that the RF waveform of the payload
niques provide sufficient support to meet the networking resembles that of ZigBee signals. When such a WiFi frame
requirement from upper layers. is received by ZigBee devices, the WiFi header, preamble,
• We implement and evaluate NetCTC on commodity de- and trailer are ignored by ZigBee as noise, while the payload
vices (Atheros AR2425 WiFi NIC and MicaZ CC2420) will successfully pass the ZigBee preamble detection and
and the USRP-N210 platform. Our comprehensive eval- the emulated ZigBee frame will be then demodulated at the
uation reveals that NetCTC achieves a reliable, trans- ZigBee receiver.
mission efficient and parallel cross technology commu-
nication under a full range of wireless configurations, ^ŵĂƌƚ,ŽŵĞ
^ĞƌǀŝĐĞƐ ^ĞƌǀŝĐĞƐ ^ĞƌǀŝĐĞƐ
^ŵĂƌƚ,ŽŵĞ
^ĞƌǀŝĐĞƐ ^ĞƌǀŝĐĞƐ ^ĞƌǀŝĐĞƐ

tŝƚŚ'ĂƚĞǁĂLJƐ WWϭ WWϮ WWϯ tŝƚŚW,zͲd WWϭ WWϮ WWϯ


including stationary, mobile, and duty-cycled settings.
• To examine the networking support for upper layer ser-
vice, we evaluate NetCTC on one-to-one IoT device con- /ŶƚĞƌŶĞƚ
/ŶƚĞƌŶĞƚ
W,z W,z
trol and one-to-many neighbor discovery applications. In d 灅 灅 d
the one-to-one NetCTC application, we embed NetCTC DƵůƚŝͲǀĞŶĚŽƌ'ĂƚĞǁĂLJƐ
灅 灅 tŝ&ŝ W
/ŶƚĞƌŶĞƚ,Ƶď
into Nexus 5 smart phones with Broadcom BCM4330 ŝŐĞĞĞǀŝĐĞƐ W,z W,z
ŝŐĞĞĞǀŝĐĞƐ

d d ůƵĞdŽŽƚŚ ĞǀŝĐĞƐ


WiFi chip to control ZigBee Smart bulb directly. The ůƵĞdŽŽƚŚ ĞǀŝĐĞƐ

experiment results show that NetCTC achieves a reliable (a) Smart home w/ gateway (b) Smart home w/ CTC
control with 62% fewer transmissions. In the one-to-many
Fig. 2. An application of CTC in smart home
application, we use WiFi to help ZigBee devices discover
their neighbors. Compared with traditional approaches,
NetCTC saves 94% of the neighbor discovery time. B. The Benefits of CTC
The paper is organized as follows. Section II presents
Given the trends of escalated wireless coexistence, PHY-
preliminaries. The design overview is presented in Section III,
CTC provides a timely communication paradigm by sup-
followed by basic and advanced designs in Sections IV and V.
porting direct message/data exchange among heterogeneous
Section VI evaluates the performance. We summarize related
wireless technologies. PHY-CTC has several unique features
work in Section VII. Section VIII concludes the paper.
including (i) transparency to today’s networks and will not
 disturb existing traffic, (ii) compatibility to existing infras-
 tructure and can be implemented directly on off-the-shelf
heterogeneous devices, and (iii) capability to achieve the
 

communication requirements (e.g., maximum communication

 
rate) defined in existing standards. With these features, PHY-

 CTC has great potential to improve existing coexistent wire-



 less networks in areas including, but not limited to channel


coordination, IoT device management and cyber physical
  control, by providing explicit message (e.g., channel, device
and sensing information) exchange. For example, in the smart
Fig. 1. PHY-CTC with signal emulation [1] home scenario in Figure 2, PHY-CTC enables to leverage
upon the deployed WiFi AP to connect smart ZigBee and
II. P RELIMINARIES Bluetooth devices without deploying gateways. The connec-
In this section, we first introduce some preliminaries about tivity among heterogeneous devices further enables us to
the PHY-CTC paradigm. Then, we summarize the benefits as remotely manage/control IoT devices, such as smart bulbs,
well as limitations of PHY-CTC. temperature/humidity controller, and audio equipment, through
our phone.
A. What’s CTC
Cross technology communication (CTC) is defined as direct C. The Limitations of PHY-CTC
communication (i.e., message/data exchange) among hetero- PHY-CTC’s significant benefits and great potential oppor-
geneous wireless devices (e.g., WiFi, BlueTooth, ZigBee and tunities come with some limitations including (i) transmission

260
failures due to signal emulation errors and (ii) asymmetric lost packets) information. In addition, due to the difference of
links due to signal emulation infeasibility. wireless standards at sender and receiver (e.g., MAC and chan-
Transmission Failures: in the signal emulation process of nel access precision), the timing of the interaction should be
PHY-CTC, due to the restrictions of technology standards designed carefully to make sure that packets are successfully
(e.g., 802.11g and 802.15.4), not all positions in the payload is transmitted/received. Such an interaction mechanism becomes
allowed to freely change (e.g., due to cyclic prefix in 802.11) more challenging when multiple heterogeneous devices require
and not all signals can be freely generated (e.g., due to QAM networking support at the same time.
quantization in 802.11). Therefore, the emulated signal may
not be perfectly match with the target signal, which further 澷濈澷澡濄澿濈濓澥
leads to transmission failures. 澷濈澷澡濄澿濈濓澦
Asymmetric Links: PHY-CTC adopts software based signal 灅
emulation under hardware restrictions, which may lead to
澷濈澷澡濄澿濈濓 煋
the asymmetric link problem due to emulation infeasibility.
Existing research demonstrates that PHY-CTC from WiFi
and BlueTooth to ZigBee is available on commodity devices 澷濈澷澡澵澷澿澔濆澹濅
without any hardware change but not vice versa [1].
Due to these limitations, there still exists a gap between 濋濝澺濝 濃濢濙澡澶濝濨澔澷濈澷澡澵澷澿 濎濝濛澶濙濙
PHY-CTC and real-world service. For example, in the smart 澸濙濪濝濗濙 澸濙濪濝濗濙
home scenario in Figure 2, without the reverse feedbacks
Fig. 3. The architecture of NetCTC between WiFI and ZigBee devices
from ZigBee/BlueTooth, it is unknown whether PHY-CTC
has successfully finished the control task. In this paper, we
aim to address the transmission failure and asymmetric link B. NetCTC Architecture
issues of PHY-CTC [1] and provide networking support to Figure 3 shows the architecture of NetCTC. Specifically,
achieve efficient bidirectional cross technology communica- NetCTC provides a networking support of two-way interaction
tion. Compared with state of the art, the technical advantages mechanism between WiFi and ZigBee devices. As shown
of NetCTC is summarized in Table I. NetCTC achieves high- in Figure 3, at the sender side (i.e., the WiFi device), we
throughput two-way communication between heterogeneous apply Fountain codes to each cross technology communication
devices without extra hardware costs (e.g., gateways). (CTC) transmission. The introduction of Fountain codes is
used to address the issue that heterogeneous receivers can not
Spectrum efficiently send back the transmission failure information (e.g.,
Expense Throuput Bidirectionality
Efficiency
√ ACKs). With the help of Fountain codes, the number of ACKs
Gateway Medium Medium High
ESense [5] Low Low Low ×
are reduced from one ACK for each packet to one ACK for a

FreeBee [6] Low Medium Low batch of encoded packets. Furthermore, since Fountain codes
WEBee [7] Low High High ×
√ introduce randomness to each packet and eliminate packet
NetCTC Low High High sequence, it is not necessary for receivers to notify the exact
TABLE I lost packet. Instead, receivers only have to notify the sender
C OMPARISON OF N ET CTC AND EXISTING CTC SOLUTIONS
whether they have received enough packets for successful
decoding, which requires only one bit information.
In our NetCTC design, when the sender sends out enough
III. D ESIGN OVERVIEW encoded transmissions, it will send out a “CTC-ACK REQ”
This section presents the design objective and challenges, message which is designed to (i) notify receivers to send
followed by the architecture of NetCTC. back ACKs and (ii) clear the channel for cross technology
transmissions from low-power devices such as ZigBee. When
A. Design Objective and Challenges receivers get the request, it will send back a “One-Bit CTC-
Despite the encouraging capability of cross technology ACK” message to let the sender know whether they have
communication in physical layer, there still exist a gap be- received enough packets for successful decoding. The sender
tween PHY-CTC and upper layer applications. We aim to fill will utilize the one bit information to (i) decide whether it
this gap by providing networking support for PHY-CTC to should starts a new batch of packets or not, and (ii) update
meet the networking requirement from upper layers. We note the timing of sending out “CTC-ACK REQ” messages for
that the heterogeneity leads to some fundamentally different future transmissions.
challenges compared with networking designs in homogeneous Up to now, we have briefly introduced the general process
networks, calling for unique designs in NetCTC. Specifically, of the interaction mechanism in NetCTC as well as the rational
the phenomenon that transmission failure and asymmetric link behind it. In the following, we will introduce design details
issues appear at the same time is prominent in heterogeneous s- including (i) how to apply Fountain codes to CTC packets,
cenarios. This phenomenon leads to the dilemma that receivers (ii) how to achieve One-Bit CTC-ACK, and (iii) how to
can not (efficiently) send back the transmission failure (i.e., decide the timing for the sender to stop deluging encoded

261
packets and send out the “CTC-ACK REQ” message, in WiFi preamble and header until it detects the emulated ZigBee
Section IV-A, IV-B, and IV-C respectively. preamble and considers it as a normal ZigBee packet. Our
design is thus compatible to current ZigBee standard.
IV. BASIC N ET CTC For a transmitted CTC packet, if it passes the cyclic redun-
Our NetCTC design provides networking support for both dancy check (CRC) at the ZigBee receiver, ZigBee buffers this
one-to-one (i.e., unicast) and one-to-many (i.e., multicast and packet for the later decoding. Otherwise, the ZigBee receiver
broadcast) cross technology communication. For sake of clar- throws this packet and marks it as a lost packet. By doing
ity, in the basic design, we focus on introducing one-to-one this, the ZigBee receiver has collected a subset of successfully
NetCTC. We will introduce how to achieve (i) networking transmitted CTC packets, whose number is denoted as N  .
support for one-to-many cross technology communication and Since the coding matrix is known in advance, it is able to
(ii) parallel acknowledgements, on advanced designs in Sec- compute the source packets, if N  > K. As for the probability

tion V-A and V-B respectively. of decoding failure, it is bounded by δ(E) < 2−(N −K) , which
drops significantly with the increase of correctly received
tŝ&ŝ tŝ&ŝ packets; for example, the error probability is less than 0.1%
tŝ&ŝWĂLJůŽĂĚ
WƌĞĂŵďůĞ ,ĞĂĚĞƌ when N  = K + 10. As a result, Fountain code is able
to increase the decoding reliability by sending out a larger
ŝŐĞĞ ŝŐĞĞ & ŽĚĞĚŝŐĞĞ number of coded packets, although some of them can be lost
WƌĞĂŵďůĞ ,ĞĂĚĞƌ ,ĞĂĚĞƌ WĂLJůŽĂĚ due to the unreliability of CTC.

ŵƵůĂƚĞĚŝŐĞĞWĂĐŬĞƚ LJtŝ&ŝ    

Fig. 4. The structure of cross technology communication (CTC) packets


 
    

with Fountain code

A. CTC-PKT with Fountain Code


As introduced in previous sections, PHY-CTC has inherent
error in signal emulation, and is prone to packet error. For       
example, in the state-of-the-art WEBee [1], the first PHY-CTC,
the success decoding rate is about 50%∼60%. To address this Fig. 5. Data recovery via Fountain Code
transmission failure issue, we adopt Fountain code [3], [4] in
Figure 5 depicts an example of utilizing Fountain code in
our design to improve the overall reliability.
PHY-CTC scenarios. In this example, there are five source
Fountain code ensures the successful decoding of the re-
packets, while Fountain encoder generates 12 coded packets.
ceived data, if the overall length of the received data is larger
Although the receiver only receives 7 packets due to the packet
than the input data length. With Fountain code, the sender
loss, it has received more packet than the source packets, thus
(e.g., the WiFi device) generates the N encoded packets as the
able to successfully demodulate.
linear combinations of K source packets. To offer reliability,
N > K, meaning that Fountain code generates more coded
B. One-Bit CTC-ACK
packets than source packets.
The structure of a cross technology communication (CTC) When Fountain codes are applied to PHY-CTC packets, only
packet with Fountain code is shown in Figure 4. From the one ACK is required for a batch of packets which significantly
figure, we can see that the WiFi device transmits a normal reduces the requests of sending back ACKs. In addition, we
WiFi packet including preamble, header, and payload while only need to include one bit information in the ACK to tell
its payload is used to emulate a standard ZigBee packet. the heterogeneous sender whether the receiver has successfully
Specifically, the emulated ZigBee packet has a ZigBee header received enough packets for decoding.
as well as a Fountain code (FC) header which includes Due to the heterogeneous PHY layers of ZigBee and WiFi
encoding vector and batch information. Such a CTC trans- as well as ZigBee devices’ capability on signal emulation,
mission is transparent to legacy networks, in the sense that the WiFi device can not decode the ZigBee packets directly.
CTC packets with Fountain code is recognized by commodity However, it can sample the RSSI, which is a mandatory
ZigBee devices transparently as a legitimate ZigBee frame. process for the CSMA purpose. By sampling the RSSI, the
The packet structure makes our design compatible to exist- WiFi is able to detect the start and the end of the ZigBee
ing standards. ZigBee devices can not differentiate emulated packet for calculating the transmission duration, and thus
packets (sent by WiFi) from regular ZigBee devices. When detect ACKs from ZigBee. In addition, since the ZigBee
a WiFi device transmits regular WiFi packets, these packets utilizes the OQPSK modulation, the amplitude remains almost
will be considered as noise by ZigBee devices since they the same during the transmission. In the CTC-ACK design, we
cannot pass ZigBees preamble detection. When a WiFi device utilize this feature for reliably detecting transmitted ZigBee
transmits a NetCTC packet, the ZigBee device will ignore packets at the WiFi side.

262
6KRUW=LJ%HH3DFNHW /RQJ =LJ%HH3DFNHW
qd be the decoding failure rate introduced by PHY-CTC.
Then the probability that a transmission is considered to
7LPH 7LPH
be failed is given by (qt + (1 − qt )qd ). Here the second
term (1 − qt )qd corresponds to the event that the packet is
 
566, G%P

566, G%P

XV
 XV successfully received with probability (1 − qt ) but failed to
  decode with probability qd .
         
In the cross technology communication with Fountain
(a) ACK with ‘0’ (b) ACK with ‘1’ codes, the probability that the receiver can successfully obtain
a batch of K encoded packets with K transmissions is given
Fig. 6. One-bit ACK with information ‘0’ and ‘1’
by P r{τ = K} = (1−(qt +(1−qt )qd ))K . Now, if K+1 trans-
missions are required to successfully obtain K original packets
Figure 6 depicts the key idea of sending ACK ‘0’ and at receivers, one of the first K coded packets must have
ACK ‘1’. The ZigBee device sends ‘0’ (as demonstrated in failed due to either transmission failure or decoding failure.
the upper part of Figure 6(a)) with one short ZigBee packet No matter which packet has failed, a new coded packet must
to represent that it has received enough packets. In the same be retransmitted successfully in the (K + 1)th transmission.
way, it sends out one long ZigBee packet to represent ACK ‘1’ The probability of this event is P r{τ = K + 1} = (K 1 )(qt +
(as demonstrated in the upper part of Figure 6(b)), meaning (1 − qt )qd )(1 − (qt + (1 − qt )qd ))K−1 (1 − (qt + (1 − qt )qd )). In
that it still needs more packets for successful decoding. In the general, the probability that K + i transmissions are required
CTC-ACK design, we choose packet sizes which are rarely to let a receiver successfully decode all K coded packets is
introduced by normal ZigBee traffic for distinguishing the given by
ACK from background ZigBee communication.
Note that the WiFi AP waits 10ms to detect the one-bit P r{τ = K + i}
CTC-ACK. This is because that the ZigBee device needs to = (K+i−1 )(qt + (1 − qt )qd )i (1 − (qt + (1 − qt )qd ))K .
i
perform CSMA before accessing the channel, which takes
about 10ms at the worst case for one transmission. Since the In our NetCTC design, the maximum number of retransmis-
packet rate of ZigBee is very low, the probability that a regular sions for one packet is limited to R (e.g., R = 7). The expected
ZigBee transmission appears at the same time when a one- number of transmissions to let a receiver successfully obtain
bit CTC-ACK is transmitting within the specified 10ms time a batch of K packets is given by
period is low.
We now introduce how to achieve the key idea in Figure 6 in 
R(K−1)

implementation. The lower parts of Figure 6(a) and 6(b) depict E{τ } = (K + i)P r{τ = K + i}. (1)
the RSSI samples of ZigBee packets received at WiFi devices. i=0

These RSSI samples are used to construct one-bit CTC-ACKs. Up to now, we obtain E{τ } – the expected number of
Specifically, to send CTC-ACK with ‘0’, the ZigBee device transmissions in the interaction mechanism in Figure 3. Note
sends a designed short packet, whose duration is about 300us. that in Eq. (1), the decoding failure rate – qd is caused by
Since the RSSI value of noise is less than -90dbm which are emulation error which is uniquely decided by the PHY-CTC
much lower than that of packets. The WiFi device is thus mechanism. For example, the value of qd is ranged from
able to obtain the CTC-ACK information ‘0’ by detecting the 0.4 to 0.5 in WEBee-type PHY-CTC [1]. Eq. (1) is used to
amplitude of RSSI. In Figure 6(b), a long ZigBee packet is compute the initial value of expected transmissions E{τ }ini
transmitted, leading to a long duration (i.e., 500us) of high as well as the minimal expected number of transmissions
RSSI samples at the WiFi side. The WiFi device recognizes E{τ }min . In our NetCTC design, the values of E{τ }ini and
the CTC-ACK with ‘1’. E{τ }min is given by E{τ }ini = E{τ }{qt =0.5,qd =0.45} and
C. NetCTC Link Model E{τ }min = E{τ }{qt =0,qd =0.4} respectively. When the sender
receives one-bit CTC-ACK from the receiver, it updates its
We now introduce the link model of cross technology
current value of τ with the following rules:
communication when Fountain codes and one-bit CTC-ACK
are applied. Let P r{τ = t} be the probability that the hetero- 1
geneous sender (e.g., WiFi) needs t transmissions to deliver a τ ← τ +  τ , ACK = 1;
2 (2)
PHY-CTC packet encoded with Fountain codes to the receiver τ ← max{E{τ }min , τ − 1}, ACK = 0;
(e.g., ZigBee). Here, τ is the expected number of transmissions
required to successfully deliver K linearly independent coded The physical meaning of Eq.(2) is that (i) the sender
packets to the receiver. In the CTC scenario, a sender has to will transmit another  12 τ  packets when the receiver fails
resend packets due to two reasons. First, transmitted packets to decode the K original packets (i.e., one-bit CTC-ACK
are physically lost in the air. Second, although the packets equals one), and (ii) the sender will reduce the number of
are received, the receiver can not decode the packet due to transmissions by one when the receiver successfully obtain
emulation errors. Let qt be the transmission loss rate and the K original packets (i.e., one-bit CTC-ACK equals zero).

263
for further improving the performance of one-to-many CTC.
In the basic design, when the ZigBee receivers have received
enough packets for successful decoding, they send out one bit
CTC-ACK (i.e., ‘0’) to notify the WiFi sender to transmit
a new batch of packets. When there are multiple ZigBee
receivers, these devices may compete with each other for the
channel access to send back one-bit CTC-ACK, which leads
to the ACK broadcast storm problem.
To alleviate this problem, we develop the parallel ACK
technique, where ZigBee receivers send back the CTC-ACKs
Fig. 7. Statistics of receiving probability: the receiver with a smaller τ on different channels simultaneously for improving the ACK
receives about 98% of the batches earlier or at the same time than the receiver efficiency. This is based on the observation that (i) one WiFi
with a higher τ .
channel has the bandwidth of 20MHz while one ZigBee
channel has the bandwidth of 2MHz, and (ii) one WiFi channel
V. N ET CTC W ITH E NHANCED F EATURES (e.g., WiFi channel 6) is completely overlapped with three
In basic NetCTC, we achieve the interaction mechanism ZigBee channels (e.g., ZigBee channel 17, 18, and 19). As
between one WiFi device and one ZigBee device. In this a result, the WiFi sender is able to capture concurrent CTC-
section, we aim to enhance the basic NetCTC by providing ACKs (i.e., RSSI) sent from different ZigBee channels for
extra features including one-to-many CTC and parallel ACKs. detecting multiple CTC-ACKs at the same time. To differenti-
ate CTC-ACKs from parallel ZigBee devices, the WiFi device
A. One-To-Many CTC sends messages to ZigBee devices to allow them to send back
To support CTC from one sender (e.g., WiFi device) to CTC-ACKs with specified transmission durations.
multiple receivers (e.g., ZigBee devices), a direct solution is

 
to assign multiple one-to-one CTC (i.e., basic NetCTC) in  
different time slots. This direct solution becomes inefficient
when the number of receiver increases. The interaction latency
increases since the receiver sends back ACKs one by one after    
random backoff and the sender will not start a new session
until all the ACKs are received.    
We noted that in the one-to-many CTC with Fountain codes, Fig. 8. WiFi detects parallel ACKs
receivers with smaller value of τ almost always obtain the K
original packets before those with larger value of τ . In other Figure 8 depicts an example of using parallel ACKs on two
words, when the receiver with the maximum τ obtain the K different ZigBee channels, i.e., ZigBee channel 17 and 19, for
original packets, it is highly possible that all the other receivers two ZigBee receivers. Since the bandwidth of overlapped WiFi
have successfully obtained the K packets. channel 6 is much wider than that of ZigBee, it senses the con-
To further confirm this observation, we deploy 16 ZigBee current transmission of two ZigBee CTC-ACKs transmitted at
nodes near a WiFi sender, which broadcasts a CTC packet different channels, thus recognizing parallel CTC-ACKs.
(i.e., ZigBee packet) every 10ms. The total number of packet
VI. E VALUATION
broadcasts is 1000. The receivers keep the packet sequence
number and time stamp. The detail setting of this experiment In this section, we evaluate the performance of NetCTC
can be found in Section VI. After collecting the packet across various domains, including (i) CTC performance com-
reception trace, for each packet, we compare the reception parison on throughput, latency, and the number of transmis-
between each link pair. sions, (ii) communication reliability, (iii) support in stationary,
Figure 7 shows that the ZigBee node with smaller value of mobile, duty-cycled, and parallel cross technology communi-
τ receives about 98% of the batches (i.e., K=10 packets in cation scenarios, and (iv) the example application of NetCTC
the experiment) earlier or at the same time than the node with on unicast (i.e., IoT device management/control) and broadcast
larger value of τ . Based on this observation, the sender updates (i.e., neighbor discovery).
the τ value in the initial stage using the direct solution. After
A. Experiment Setting
obtaining all receivers’ τ value, the sender selects the receiver
with maximum τ as the delegate to send back ACK, and thus Figure 9 shows the evaluation setting of NetCTC. We have
achieve one-to-many interaction without causing ACK storm implemented NetCTC as a sender on the USRP-N210 platform
problem. with 802.11 b/g PHY and a commodity WiFi card Atheros
AR2425. At the receiver side, we have tested the NetCTC
B. Parallel ACK on 16 commodity ZigBee receivers (i.e., MICAz). We note
In this section, we introduce the parallel ACK technique, that NetCTC supports direct communication between hetero-
which allows multiple receivers to send back ACKs in parallel geneous commodity devices while USRP-N210 devices are

264
   
   

 濂濙濨澷濈澷澮澔濋濝澺濝 濇濙濢濘濙濦澔








 澜澵濨濜濙濦濣濧澔澵濆澦澨澦澩澝

濂濙濨澷濈澷澮澔濎濝濛澶濙濙澔
濆濙濗濙濝濪濙濦澔澜濁澽澷澵濮澝
濂濙濨澷濈澷澮澔濋濝澺濝
濇濙濢濘濙濦澔澜濉濇濆濄澝
  
 




 濂濙濨澷濈澷澮澔濎濝濛澶濙濙澔

 濆濙濗濙濝濪濙濦澔澜濁澽澷澵濮澝

Fig. 9. The experiment setting of NetCTC. Fig. 10. Lab Fig. 11. Hallway


 
 
濂濙濨澷濈澷澮澔濎濝濛澶濙濙澔   
濆濙濗濙濝濪濙濦澔澜濁澽澷澵濮澝
 
濂濙濨澷濈澷澮澔濋濝澺濝
濇濙濢濘濙濦澔澜濉濇濆濄澝     
  

Fig. 12. Outdoor Fig. 13. Performance In Bit Rate Fig. 14. Reliability

used only for evaluation purpose to obtain low-level physical C. Main Performance
layer information, such as the decoding failure rate (i.e., qd ), The main experiment results is summarized in Figure 13
which are inaccessible by commodity devices. and 14.
In the experiment, the Atheros AR2425 WiFi card (or
1) Comparison with Packet-Level CTC: In Figure 13, we
USRP-N210) transmits 1,000 NetCTC packets to the 16
compare the bit rate of NetCTC with the state-of-the-arts in-
MICAz nodes. The time interval between transmissions is
cluding (i) packet level CTC, i.e., ESense [5] and FreeBee [6]
10ms. The transmit power of WiFi and ZigBee devices is set
and (ii) physical layer CTC, i.e., WEBee [1] with different
to 20dBm and 0dBm respectively. The default transmission
number of transmissions. The bit rate of packet level CTC,
distance is 4m. The maximum transmission unit of each
i.e., ESense and FreeBee, is about 9bps and 17bps with a
transmitted packet is adjust as follows: each NetCTC packet
single CTC transmitter, while the bit rate of NetCTC is about
consists of a 32-bit preamble, and a default frame payload of
180kbps, 10, 000× the bit rate of FreeBee. In the following
25 bytes, resulting in about a 1ms duration, which is the typical
evaluation, we focus on compare the performance of NetCTC
length of payload for ZigBee communications. The batch size
with physical layer CTC.
(i.e., K) is set to 16 in the experiment.
To accommodate the dynamics caused by wireless environ- 2) Comparison with Physical Layer CTC: The bit rate of
ment, we conduct the above experiment on stationary/mobile, WEBee with one transmission is as high as 226kbps, which
indoor/hallway/outdoor, and duty-cycled scenarios, as shown is higher than that of NetCTC (i.e., 180kbps). The interaction
in Figure 10, 11, and 12. To ensure statistical validity, the mechanism in NetCTC introduces some latencies (about 5ms
experimental results of each scenario are the average values on average), which lowers the bit rate. WEBee does not have
of 10 rounds. the interaction mechanism and the reliability of WEBee with
one transmission is as low as 0.4, as shown in Figure 14.
B. Compared Schemes and Metrics The reliability of WEBee with six (re)transmission increases
We compare NetCTC with the following state of the art. to 0.92 while its bit rate drops to 37kbps. With the interaction
1) ESense [5] by K. Chebrolu et al, 2009. mechanism, NetCTC achieves almost 100% reliability.
2) FreeBee [6] by S. Kim et al, 2015. In real-world applications, due to the lack of interaction
3) WEBee [1] by Z. Li et al, 2017. mechanisms, the only way for WEBee to achieve reliable
Among them, ESense [5] and FreeBee [6] are Packet-Level communication is sending packets with the maximum allowed
cross-technology communication techniques while WEBee [1] number of transmissions no matter packets are delivered or
is a physical-layer cross-technology communication technique. not, which can not guarantee either efficiency or reliability.
We use two metrics for the following performance evaluation. In the following, we focus on investigating the achievable
1) Throughput – the total received bits of packets divided throughput of NetCTC in reliable communication (i.e., ACKs
by the time duration of each round of experiment. are required), where throughput is defined as the bits of
2) Reliability – the number of successfully received and successfully delivered packets confirmed by receivers during
decoded packets at receivers over the total number of the experiment. Specifically, we study the impact of (i) net-
transmitted packets. work size, (ii) transmission distance, (iii) payload length, (iv)

265
Fig. 15. Impact of Network Size Fig. 16. Impact of Trans. Distance Fig. 17. Impact of Payload Length

Fig. 18. Impact of Mobility Fig. 19. Impact of Duty Cycle Fig. 20. Parallel Communication

mobility, (v) duty cycle and (vi) parallel communication on


F. Impact of Payload Length
NetCTC. We also evaluate the performance of NetCTC on
one-to-one (i.e., IoT device control) and one-to-many (i.e., Figure 17 plots the throughput of NetCTC with different
neighbor discovery) applications. lengths of payload. As we can see in the figure, the throughput
of NetCTC is 7.27kbps, 6.9ykbps, 6.07kbps, 5.87kbps, and
D. Impact of Network Size 5.096kbps respectively when the length of payload is 15, 20,
25, 30, and 35 bytes. The throughput drops since the decoding
In this experiment, we explore the impact of network size
failure rate (i.e., qd ) increases along with the increase of the
on our design. We evaluate the performance of NetCTC with
payload length. The experiment results suggest using small
different number of receivers, ranging from 1 to 16. Figure 15
packets in the applications of physical layer cross technology
shows the throughput with different number of receivers. From
communication.
the figure, we can see that NetCTC’s throughput is 7.14kbps
when there are only one ZigBee receiver. The throughput G. Impact of Mobility
decreases to 2.83kbps when there are 16 receivers. It’s because
We also evaluate NetCTC with a mobile ZigBee receiver in
that NetCTC tries to reliably deliver a packet to all receivers
the three different scenarios (i.e., lab, hallway, and outdoor). In
and the failure on some receivers causes more transmissions.
this experiment, the moving speed varies from 1m/s to 8m/s.
The total throughput of NetCTC increases gradually with the
Figure 18 shows that the throughput of NetCTC drops from
increase of the number of receivers, indicating that NetCTC
4.60kbps to 3.37kbps, from 4.60kbps to 3.36kbps, and from
works well in one-to-many scenarios.
5.21kbps to 3.72kbps in lab, hallway, and outdoor scenarios.
E. Impact of Transmission Distance The average throughput of NetCTC is 3.64kbps, 3.82kbps,
and 4.29kbps in these three scenarios, indicating that NetCTC
We evaluate the performance of NetCTC with different
works well under different wireless environments and different
transmission distances in indoor (Figure 10), hallway (Fig-
levels of mobilities.
ure 11), and outdoor (Figure 12) scenarios. Figure 16 shows
the throughput of NetCTC between the WiFi sender and H. Impact of Duty Cycle
the ZigBee receiver under varying distances. With increasing NetCTC is also evaluated using low-duty-cycle ZigBee re-
distances from 1 to 10 meters, the throughput drops slowly ceivers, which sleep most of the time and wake up periodically
from 6.58kbps to 6.07kbps in the indoor scenario. Similar to check for radioactivity. In this case, a sender needs to
results are observed in the hallway/outdoor scenario as well, transmit a train of repeated packets to wake up the receiver.
where the throughput changes from 6.53kbps/6.56kbps to In our experiments, the duty-cycle ratios are set as 30%, 15%,
6.15kbps/5.96kbps, indicating that signal attenuation will not 6%, 3%, and 2%. NetCTC keeps on sending out packets
significantly degrade NetCTC’s performance. from the WiFi sender until that the ZigBee receiver’s ACK

266
is received, which means that a ZigBee device is successfully WEBee transmits every control packet 7 times (due to lack
woken up by the WiFi device. We then measure the throughput of interaction mechanism) and achieves a 97% reliability. Our
during this process. The results are shown in Figure 19 where NetCTC design dynamically adjusts the number of retransmis-
we can see that NetCTC works well under a wide range of sions based on the feedback ACK and achieves reliable control
duty-cycle settings. with fewer transmissions.
I. Parallel Communication
Here we show that NetCTC can support two parallel com-
munications between WiFi and ZigBee. In the experiment, two
ZigBee receivers are working on channel 17 and 19, as shown
in Figure 8. One WiFi sender transmits NetCTC packets with
two different subcarrier regions (11 data subcarriers each).
The ZigBee receivers send back one bit CTC-ACK once they
receive CTC-ACK REQ.
Figure 20 compares the throughput of NetCTC on ZigBee
channel 17 and 19. From the figure, we can see that the
throughput of NetCTC on channel is about 5.94kbps. At
Fig. 22. Performance in IoT Device Control: WEBee vs. NetCTC
the same time, the throughput on channel 19 is slightly
lower than that on channel 17, which is about 5.73kbps. The
K. Application II: Neighbor Discovery
reason for this phenomenon is the diversity of channel quality
under different channels. With two parallel channels, the total This section demonstrates the advantage of using one-to-
throughput of NetCTC is doubled. many NetCTC by introducing the application of neighbor
discovery. To reduce the energy consumption, ZigBee devices
SmartPhone adopt low duty cycles, where they turn off wireless radios.
(WiFi BCM4339) Although the low duty cycle strategy saves energy, it leads to
long discovery time since ZigBee devices may miss discovery
packets. With the help of one-to-many NetCTC, the WiFi
device broadcasts ZigBee neighbor information directly to
ZigBee devices. Note that WiFi is always in active state and
thus discover all nearby ZigBee devices’ information, leading
to quick neighbor discovery.
Bulb (ZigBee)
 
     


Fig. 21. Application I: IoT Device Control
      


J. Application I: IoT Device Control       




      

Our NetCTC design is transparent to legacy networks and
compatible to existing standards (e.g., 802.15.4). It calculates          

the WiFi payload bits for emulating ZigBee transmissions


#$ #" #$ #" " $!%
without modification on WiFi PHY and MAC layers. As a
result, NetCTC’s packets are transmitted in the same way as
Fig. 23. Application II: Accelerating Neighbor Discovery
normal WiFi packets, while applications decide WiFi payload
bits respectively, and then send the payload to the WiFi driver For example, in Figure 23, there are three ZigBee nodes,
for WiFi transmissions. with the duty cycle ratio of 1/3, 1/5 and 1/7 respectively. To
This section shows a real-world application of one-to-one speed up the neighbor discovery, WiFi advertises ZigBee’s
NetCTC on IoT device control. We implement both WEBee neighbor information all the time to the ZigBee nodes using
and NetCTC on a Nexus 5X smartphone equipped with the NetCTC technique. When ZigBee nodes wake up, they can
a Broadcom BCM4330 WiFi chip to control ZigBee light get the global nearby devices’ information, thus significantly
bulbs, as shown in Figure 21. The available control includes saving the neighbor discovery time.
the on/off status, light color and intensity. NetCTC achieves Figure 24 depicts the comparison of the time with and
reliable control of ZigBee devices from a WiFi radio without without NetCTC with the number of ZigBee devices increas-
WiFi-ZigBee gateways and without hardware modification on ing from 3 to 7. Following Disco [8], ZigBee nodes adopt
smartphone and smart light bulbs. co-prime numbers, i.e. 3, 5, 7, 11, 13, 17, 19, 23, and 29,
Figure 22 plots the average number of transmissions and the as their duty cycle periods. It is clear that Disco requires
corresponding reliability of WEBee and NetCTC in the IoT significant amount of time with an increasing number of nodes.
device control application. From the figure, we can see that As shown in Figure 24, Disco takes 323 slots for finding all

267
the 7 ZigBee devices, since it discovers neighbors in a pair- Packet Level CTC: Realizing the limitation of gateway-
wise fashion. NetCTC broadcasts the global information, thus based approaches, wireless community invested early efforts to
ensuring that the discovery time is maintained for providing establish direct connectivity for heterogeneous devices through
efficient neighbor discovery. For the discovery of 7 nodes, packet-level CTC. Chebrolu et al. and Zhang et al. propose
NetCTC only takes 19 slots, which is 5.8% of Disco. Esense [5] and Howies [25] which enable direct communi-
cation from commodity WiFi devices to commodity ZigBee
devices by sending out dedicated WiFi packets. Researchers
also establish direct cross technology communication mech-
anisms [6], [26], [27] via existing traffic. For example, Kim
et al. builds direct communication from WiFi to ZigBee via
shifting the timings of the mandatory beacons [6]. Chi et al.
propose B 2 W 2 which builds a direct communication link from
WiFi to Bluetooth via WiFi data traffic [7].
In all these packet-level CTC approaches, packets are u-
tilized as minimal units to construct special energy pattern
Fig. 24. Performance in Neighbor Discovery: Disco vs. NetCTC (i.e., delivered information) on the shared spectrum which
VII. R ELATED W ORK can be detected by other technologies through carrier sensing.
Since the duration of a wireless packet is in the range of
This section first introduces approaches for addressing het- milliseconds, embedding communication symbols with sparse
erogeneity to present the context of our work. Then we review packet-level information limits the transmission rate.
existing studies on (i) cross technology communication and (ii) Physical Layer CTC: To eliminate the limitation of packet-
interaction mechanism which are closely related to our design. level CTC, researchers propose physical layer CTC [1], [2]
A. Approaches for Addressing Heterogeneity which establish direct physical layer communication from
Many of today’s wireless technologies (e.g., WiFi, Blue- WiFi and Bluetooth to ZigBee via software based signal
tooth and ZigBee) are sharing unlicensed/open spectrum (e.g., emulation. These physical layer CTC techniques achieve the
ISM bands). To address the coexistence issue among hetero- maximum transmission rate defined by the standard (e.g.,
geneous technologies, many techniques have been proposed, 250kbps for ZigBee), which is thousands of times of the state-
such as (i) CSMA and RTS/CTS in IEEE 802.11, DSSS in of-the-art packet-level CTC (e.g., 7.5bps@99% for ZigBee).
IEEE 802.15.4, FHSS in Bluetooth, (ii) cognitive radio [9], The PHY-CTC approach faces transmission failure (due to
(iii) interference cancellation [10]–[14], (iv) spectrum underlay emulation error) and asymmetric link (due to emulation infea-
and overlay [15]–[19], and (v) white space networking [20]– sibility) issues. In this paper, we propose networking support
[23]. All these techniques address heterogeneity based on im- for PHY-CTC to address these issues to meet the requirement
plicit information since heterogeneous devices are assumed to (e.g., efficiency and reliability) from upper layer protocols.
be unable to exchange their message (e.g., spectrum utilization C. Interactive Models
information) directly.
In this paper, we focus on addressing heterogeneity with Extensive interactive models have been proposed to provide
cross-technology communication which allows explicit real- networking support in homogeneous networks [28], [29].
time information exchange among heterogeneous devices. The The interactive models, which are designed with Fountain
capability of direct communication brings great opportunity to codes [3], [4] or network coding [29]–[31], are the closest
either support/improve existing approaches or establish new to our design. Chieochan et al. develop interactive models for
approaches for addressing heterogeneity. delay-sensitive media streaming by modifying the traditional
802.11e block acknowledgment scheme to work with fountain
B. Cross Technology Communication coding [28]. Hagedorn et al. propose interactive mechanisms
Existing cross technology communication (CTC) technolo- for rateless Deluge and ACKless Deluge using random linear
gies can be divided into two categories. One is indirect network coding.
approach via gateways. The other is direct approach which Different from these traditional interactive models, our
includes packet-level CTC and physical-layer CTC. design establish interactivity between heterogeneous devices.
Gateway Based CTC: In the indirect approach, multi-radio In heterogeneous networks, due to the difference of wireless
gateways are deployed as translators to bridge the communi- standards at sender and receiver (e.g., MAC and channel access
cation gap between heterogeneous devices [24]. The gateway- precision), the receiver may not have chance to access the
based approach has several limitations. Gateways incur not channel when the sender is sending packets. In addition, ACKs
only extra hardware costs but also extra traffic overhead and may not be available simply because that the communication
time delay due to protocol conversions. In addition, gateways from the receiver to the sender is not supported. In this paper,
have to be deployed in advance whenever the communication we address these issues with a new interactive model which
among heterogeneities is required, making it difficult to sup- provides efficient and reliable cross technology communica-
port mobile and ad hoc scenarios. tion for heterogeneous devices.

268
VIII. C ONCLUSION [13] S. Gollakota and D. Katabi. Zigzag decoding: combating hidden
terminals in wireless networks. In ACM SIGCOMM, 2008.
In this work we present NetCTC, a new interactive mech- [14] Yan Yubo, Yang Panlong, Li Xiangyang, Tao Yue, Zhang Lan, and You
anism between heterogeneous devices. This interactive mech- Lizhao. Zimo: Building cross-technology mimo to harmonize zigbee
smog with wifi flash without intervention. In Proceedings of the 19th
anism provides networking support for PHY-CTC to address Annual International Conference on Mobile Computing (MobiCom),
the transmission failure and asymmetric link issues, thus pro- 2013.
viding efficient and reliable cross technology communication [15] TQ Duong, Vo Nguyen Quoc Bao, and H-J Zepernick. Exact outage
probability of cognitive af relaying with underlay spectrum sharing.
to upper layer protocols. We evaluate NetCTC on both USRP Electronics letters, 47(17):1001–1002, 2011.
and commodity WiFi/ZigBee devices. The experiment results [16] Steven Hong, Jeffrey Mehlman, and Sachin Katti. Picasso: flexible rf
show that NetCTC achieves bidirectional cross technology and spectrum slicing. In SIGCOMM, 2012.
[17] Liping Luo, Ping Zhang, Guangchi Zhang, and Jiayin Qin. Outage per-
communication, while providing reliability under stationary, formance for cognitive relay networks with underlay spectrum sharing.
mobile and duty cycled scenarios. IEEE Communications Letters, 15(7):710–712, 2011.
[18] Shravan Rayanchu, Vivek Shrivastava, Suman Banerjee, and Ranveer
ACKNOWLEDGEMENT Chandra. Fluid: Improving throughputs in enterprise wireless lans
This work was supported in part by the NSF CNS-1444021, through flexible channelization. IEEE Transactions on Mobile Com-
puting, 11(9):1455–1469, 2012.
NSF CNS-1525235, NSF CNS-1718456, and NSF China [19] Lei Yang, Wei Hou, Lili Cao, Ben Y. Zhao, and Haitao Zheng.
61672196. We sincerely thank our shepherd Vijay Gopalakr- Supporting demanding wireless applications with frequency-agile radios.
ishnan and the anonymous reviewers for their valuable com- In NSDI, 2010.
[20] Jun Huang, Guoliang Xing, Gang Zhou, and Ruogu Zhou. Beyond co-
ments and suggestions. existence: Exploiting wifi white space for zigbee performance assurance.
R EFERENCES In Proceedings of the 18th IEEE International Conference on Network
Protocols (ICNP), 2010.
[1] Zhijun Li and Tian He. Webee: Physical-layer cross-technology commu- [21] Min Li Huang and Sin-Chong Park. A wlan and zigbee coexistence
nication via emulation. In Proceedings of the 23rd Annual International mechanism for wearable health monitoring system. In Proceedings of
Conference on Mobile Computing and Networking (MobiCom), 2017. the 9th international conference on Communications and information
[2] Wenchao Jiang, Zhimeng Yin, Ruofeng Liu, Zhijun Li, Song Min Kim, technologies (ISCIT), 2009.
and Tian He. Bluebee: a 10,000x faster cross-technology communication [22] Byoung Hoon Jung, Jo Woon Chong, Chang Yong Jung, Su Min
via phy emulation. In SenSys, 2017. Kim, and Dan Keun Sung. Interference mediation for coexistence of
[3] D. J. MacKay. Fountain codes. In IEE Proceedings-Communications, wlan and zigbee networks. In Personal, Indoor and Mobile Radio
2005. Communications (PIMRC), 2008.
[4] Wan Du, Jansen Christian Liando, Huanle Zhang, and Mo Li. When [23] Yufei Wang, Qixin Wang, Zheng Zeng, Guanbo Zheng, and Rong Zheng.
pipelines meet fountain: Fast data dissemination in wireless sensor Wicop: Engineering wifi temporal white-spaces for safe operations of
networks. In Proceedings of the 13th ACM Conference on Embedded wireless body area networks in medical applications. In Real-Time
Networked Sensor Systems (SenSys), 2015. Systems Symposium (RTSS), 2011.
[5] Kameswari Chebrolu and Ashutosh Dhekne. Esense: Communication [24] Tao Jin, Guevara Noubir, and Bo Sheng. Wizi-cloud: Application-
through energy sensing. In Proceedings of the 15th Annual International transparent dual zigbee-wifi radios for low power internet access. In
Conference on Mobile Computing and Networking (MobiCom), 2009. Proceedings of IEEE International Conference on Computer Communi-
[6] Song Min Kim and Tian He. Freebee: Cross-technology communication cations (INFOCOM), 2011.
via free side-channel. In The 21st Annual International Conference on [25] Yifan Zhang and Qun Li. Howies: A holistic approach to zigbee
Mobile Computing and Networking (MobiCom), 2015. assisted wifi energy savings in mobile devices. In Proceedings of IEEE
[7] Zicheng Chi, Yan Li, Hongyu Sun, Yao Yao, Zheng Lu, and Ting Zhu. International Conference on Computer Communications (INFOCOM),
B2w2: N-way concurrent communication for iot devices. In Proceedings 2013.
of ACM Conference on Embedded Network Sensor Systems (SenSys), [26] Zhimeng Yen, Wenchao Jiang, Song Min Kim, and Tian He. C-morse:
2016. Cross-technology communication with transparent morse coding. In
[8] Prabal Dutta and David Culler. Practical asynchronous neighbor discov- Proceedings of IEEE International Conference on Computer Commu-
ery and rendezvous for mobile sensing applications. In Proceedings of nications (INFOCOM), 2017.
the 6th ACM conference on Embedded network sensor systems, pages [27] Wenchao Jiang, Zhimeng Yen, Song Min Kim, and Tian He. Transparent
71–84. ACM, 2008. cross-technology communication over data traffic. In Proceedings of
[9] Carl R Stevenson, Gerald Chouinard, Zhongding Lei, Wendong Hu, IEEE International Conference on Computer Communications (INFO-
Stephen J Shellhammer, and Winston Caldwell. Ieee 802.22: The COM), 2017.
first cognitive radio wireless regional area network standard. IEEE [28] Surachai Chieochan and Ekram Hossain. Wireless fountain coding
communications magazine, 47(1):130–138, 2009. with ieee 802.11e block ack for media streaming in wireline-cum-
[10] Shyamnath Gollakota, Fadel Adib, Dina Katabi, and Srinivasan Seshan. wifi networks: A performance study. IEEE Transacrion on Mobile
Clearing the rf smog: Making 802.11n robust to cross-technology Computing (TMC), 2011.
interference. In Proceedings of ACM SIGCOMM, 2011. [29] Andrew Hagedorn, David Starobinski, and Ari Trachtenberg. Rateless
[11] Daniel Halperin, Thomas E. Anderson, and David Wetherall. Taking deluge: Over-the-air programming of wireless sensor networks using
the sting out of carrier sense: interference cancellation for wireless random linear codes. In IPSN, 2008.
lans. In Proceedings of the 14th Annual International Conference on [30] R. Ahlswede, Ning Cai, S.-Y.R. Li, and R.W. Yeung. Network informa-
Mobile Computing and Networking, MOBICOM 2008, San Francisco, tion flow. IEEE Transactions on Information Theory, 2000.
California, USA, September 14-19, 2008, pages 339–350, 2008. [31] T. Ho, M. Medard, R. Koetter, D. R. Karger, M. Effros, J. Shi, and
[12] Yubo Yan, Panlong Yang, Xiang-Yang Li, Yafei Zhang, Jianjiang Lu, B. Leong. A random linear network coding approach to multicast. IEEE
Lizhao You, Jiliang Wang, Jinsong Han, and Yan Xiong. Wizbee: Wise Transactions on Information Theory, 2006.
zigbee coexistence via interference cancellation with single antenna.
IEEE Trans. Mob. Comput., 2015.

269

You might also like