Co-existence Issue in IoT Deployment using Heterogeneous Wireless Network (HetNet): Interference Mitigation using Cognitive Radio

—In today’s era of Machine-to-Machine (M2M) communication, where a tiny embedded device is having enough functionality and computational power to be network enabled for various applications aided with sensors, relays, and actuators to make Internet-of-Things (IoT) a reality. IoT is not a technology by itself rather a vision to interconnect every single piece of intelligent electronics using existing LAYER-I and II data communication standards and the internet based on OSI Layer Architecture. In the context of deployment of such IoT networks, it is imperative; a heterogeneous wireless network will be most viable for the need of different application-specific scenarios. The primary goal of this paper is to propose a Cognitive Radio (CR) Algorithm for mitigation of interference based on the analytical and empirical model of packet error rate (PER) for IEEE 802.15.4 network in the presence of IEEE 802.11 b/g/n network operating in the 2.4 GHz unlicensed industrial scientific medical (ISM) frequency band. The power spectral density (PSD) of three standards widely deployed IEEE 802.11 b/g/n are considered for determining the interference power of overlapping channels of IEEE 802.11 b/g/n. To create a test-bed for an Empirical Model Digi’s XBEE® 802.15.4 Zigbee modules is used for IEEE 802.15.4 and Intel/ Qualcomm (Atheros) WiFi Alliance adaptors are considered and IEEE 802.11 to validate the analytical model.


I. INTRODUCTION
The scope of this paper is to propose an interferenceaware Cognitive Radio (CR) algorithm. To propose such an algorithm, two models have been considered: (1) Analytical Interference Model (2) Empirical Model based on IEEE 802.11 and IEEE 802.15.4 Zigbee (XBEE®) interference assessment Test-bed. The inspiration for such work is taken from the deployment of the Internet of Things (IoT) based on a Heterogeneous wireless network (HetNet) [1]. Fig.1 shows a practical scenario of such IoT deployment using existing IEEE standard architecture in LAYER-I and LAYER-II of TCP/IP stack to enable the embedded devices to have data Network functionality; to be seamlessly connected on Internet [2]- [4]. The primary motivation for deploying IoT using HetNet is the luxury of using the architectural benefit of using both WPAN and WLAN, depending on the scenario specific requirements [5] such as for interconnection of necessary sensors and actuator a low data-rate. Zigbee Mesh would be best suitable with low Power applications and for interconnecting such WPAN clusters to the Internet backbone a relatively higher data-rate and higher radio coverage IEEE 802.11 WLAN would be best suited.Further such WLAN adaptor can be added to gateway devices mostly wall powered with not much power restrictions. However, at the same time this advantage of using WLAN and WPAN together as HetNet can become a necessary evil since such spectral overlapping of IEEE 802.15.4 Zigbee Mesh (WPAN) and IEEE 802.11 (WLAN) would give rise to interference issue as they both use 2.4 GHz ISM band and the channels are overlapping in nature [6]- [8] . Thus to minimize the impact of interference of IEEE 802.11 on IEEE 802.15.4 this paper has proposed a Cognitive Radio Algorithm [9] aided with Analytical Interference Model of IEEE 802.11 [8] on Zigbee IEEE 802.15.4 [7]. Further, the Analytical Model has been experimentally validated by a test-bed using Digi XBEE® IEEE 802. 15

.4 and Intel/Qualcomm (Atheros) WiFi Alliance adaptors.
A. "Co-existence" Impact of IEEE 802.11 b/g/n on IEEE 802. 15.4(ZigbeeTM) for IoT deployment Scenario IoT architectural model [1] suggests the heterogeneity of under-laying technologies. In any scenario of IoT deployment, it gives an M2M backbone network supporting LAYER-I/II functionalities enabling the smart wireless devices to have seamless interoperability by the use of various technologies. This further provides a common platform from Transport/Application Layer socket for various emerging applications because of layering architecture. Due to the heterogeneity of IoT networks, it is inevitable to have coexistence issues in the ISM 2.4 GHz unlicensed band as many PHYSICAL Radio standards are using the same spectrum. The scope of this paper is to evaluate and quantify the severity of the inference of co-existing radio channels using an analytical model. Further, a hardware testbed to support the analytical model. In this paper, IEEE 802.15.4 (Zigbee TM ) is taken as the technology of interest for deployment of IoT M2M network. The impact of interference from IEEE 802.11 b/g/n network is evaluated and quantified. The motivation behind taking IEEE 802.11 b/g/n network as a major contributor for interference for IEEE 802.15.4 (Zigbee TM ) IoT deployment is because of two reasons. 1) In the suggested architecture [4] of IoT network, it is most likely that an IEEE 802.11 b/g/n network might interoperate with IEEE 802.15.4 (Zigbee TM ) for any IoT deployment for various application specific scenarios. 2) IEEE 802.11 b/g/n has a more significant Spectrum footprint in terms of Power (which is up to 20dBm Transmitted Power) as compared to any typical IEEE 802.15.4 (Zigbee TM ), which is a very low-power LR-WPAN. (which is up to 0dBm Transmitted Power). The overlapping is shown in Fig.2 15.4(ZigbeeTM) in the presence of interference from coexisting IEEE 802.11b/g/n An Analytical model [10] can be used to quantize the performance degradation of IEEE 802.15.4 (Zigbee TM ) IoT deployment contributed by the interference from coexisting IEEE 802.11 b/g/n network. This model has especially considered two key parameters for the performance model: 1) As this paper is motivated from IoT deployment scenarios, the deployment of smart devices must have key parameters of the distance of deployment from the potential IEEE 802.11 b/g/n interferer. 2) Secondly, from the perspective of interference model PSD of the Channel is considered to play a major role as the distribution of Power over the spectrum of the band-limited channel is non-uniform [11]. The Transmission mask of various IEEE 802.11 b/g/n network is given in Figure 3.  [8] As the PSD of IEEE 802.11 interferer is non-uniform, the frequency-offset of center-frequency of IEEE 802.15.4 (Zigbee TM ) and center-frequency of IEEE 802.11 is the key factor for calculation of power ratio desired IEEE 802.15.4 power to the IEEE 802.11 interferer power, which further quantifies SINR. In this paper, the measurement for interference is SINR, which is the parameter for quantifying interference.
The performance deterioration of IEEE 802.15.4 in the presence of IEEE 802.11b/g/n interference can be quantified by PDR, which can be calculated involving LAYER-I and LAYER-II parameters. 1) PHYSICAL LAYER (LAYER-I) PARAMETER: the First parameter is related to the BER, or the Probability of Error Q(x) based on the SINR, which is calculated by the ratio of Zigbee received power to that of interference power from multiple IEEE802.11 interferers added with a Noise power PN 0 . 2) DATA-LINK LAYER / LAYER-II (MAC SUB-LAYER): For calculation of PDR, here the Tc: Collision Time has been considered, which is calculated by the Timeoffset of the inter-arrival times of the data-link layer PDU. From the above two LAYER parameter BER and Tc respectively, the PER is calculated. PDR is obtained by the ratio of the number of successful packets received without error to the number of packets sent.

A. Analytical Model 1) Physical Layer:
The physical layer parameter for performance in the presence of interference is BER, which is the Probability of Error function Q(x) of O-QPASK, modulation scheme used in IEEE 802.15.4 (Zigbee TM ) PHYSICAL LAYER standard [7].
Probability of Error Q(x) / Bit-Error-rate (BER) for O-QPASK modulation is given by equation (1).
Two models are specified as follows: • Model for non-interference Scenario: •

2) Path Loss Model:
To obtain the received power from desired IEEE802.15.4 (Zigbee TM ) source and the received interference power from IEEE 802.11 source at the IEEE802.15.4 (Zigbee TM ) receiver an indoor propagation model has been used [10]. The motivation behind using an indoor propagation model is IoT deployment in many household and industrial application typically in urban indoor radio condition. The following equation (5) is the path loss based on the distance from the source to IEEE802. 15 The PSD can be obtained by summation of Power of all 52 sub-carriers ranging from .
Fig.5(a) shows the PSD for 52 OFDM sub-carriers. The simulation for OFDM envelope alongside the TX Mask for IEEE 802.11g is done in MATLAB. The major difference between IEEE 802.11b DSSS and the IEEE 802.11g OFDM is that in the band-limited channel of 22MHz for the prior and 20 MHz, which is from fc-10MHz to fc+10MHz the later shows an almost flat PSD distribution unlike the prior which shows a distribution. Thus, the frequency offset between center frequencies of IEEE 802.15.4 to that of IEEE 802.11 plays a key role in IEEE 802.11b DSSS interference. However, the same is not that significant in the case of OFDM i.e. IEEE 802.11g interference case. Fig.5 shows the flat top or a more-over uniform distribution pattern of power in OFDM case. This can be further reasoned with the summation of 52 individual sub-carriers. Each sub-carrier of OFDM shows a non-uniform distribution of PSD. However, when all the 52 sub-carrier power is summation the resulting PSD distribution result in a more over uniform distribution. Sub-carrier frequency spacing 20MHz/64=0.3125MHz .where, symbol duration Transition time [12].Technically IEEE 802.11n HT 20MHz is similar to that of IEEE 802.11g. The IEEE 802.11n HT 20MHz use the same OFDM technique. Thus, equation (8) holds good for calculation of power and the distribution of PSD HT 20 MHz.
The PSD can be obtained by summation of Power of all 56 sub-carriers ranging from .  The PSD can be obtained by summation of Power of all 114 sub-carriers ranging from .  at . From the power has a negative slope and reaches -28dBr. The power further decreases to -45dBr at    For modeling interference, it is primarily required to map the channels of IEEE 802.15.4 (Zigbee) with the overlapping channels of IEEE 802.11. Secondly and most importantly the Frequency offset is an important especially in case of IEEE 802.11b DSSS interference as the PSD is non-uniformly distributed.   Inter-arrival rate of WLAN IEEE 802.11 Packets: (12) Inter-arrival rate of Zigbee IEEE 802.15.4 Packets: (13) In this paper, the IEEE 802.11b/g/n interference source is transmitting 200 packets per sec. Thus makes each packet arrives per 5000 µsec. This makes only one IEEE 802.11b/g/n packet will overlap with each IEEE 802. 15 11 Network two laptops equipped with Intel® Centrino IEEE 802.11 b/g/n Adaptor is separated by 1 meter works as an interference source. Further "d" is the distance between Network-1 and Network-2 is variable in the range of 1-10 meters for finding out PER. The above-mentioned test bed is considered for finding the impact of interference on the IEEE 802.15.4 network in the presence of IEEE 802.11 b/g/n network. To validate the analytical model this testbed in taken into consideration to find out packet Error Rate (PER). The distance between XBEE® PAN and IEEE 802.11 b/g/n network is varied to find out the comparative analysis of PER. Fig.10 shows the picture of the experimental setup as per the block diagram in Fig.9. Fig.10a Shows the Network-1 XBEE® IEEE 802.15.4 PAN Coordinator and End-Device and receiver and transmitter respectively. Fig.10b shows the complete test bed comprising of the above mentioned XBEE® PAN and the Network-2, which is an IEEE 802.11 b/g/n network implemented using to laptop computers having an Intel® Centrino Network Interface Card (NIC) works an interference source for the XBEE® PAN. Fig. 10c shows the Size of the XBEE® Zigbee Module.

B. Cognitive Radio Model For Interference Mitigation
Result analysis of the interference experiments shows a comparative study of an analytical and empirical model for the derogative impact on the performance of IEEE 802.15.4 network in the presence of IEEE 802.11 b/g/n nodes as interfering sources. Further these IEEE 802.11 b/g/n nodes are having significantly higher PSD over the IEEE 802.15.4 making it a serious matter of concern for performance especially when the distance between both the networks is less than 2 meters, which is highly probable for a Internet of Things deployment scenario, where tiny low power Sensor/ Actuator nodes would be deployed using IEEE 802.15.4 Zigbee TM network and such networks would be connected to the internet backbone networking using IEEE 802.11 network a typical case of heterogeneous network. The scope of this paper is further extended by not only to give an account of interference in above-mentioned scenarios but to propose a Cognitive Radio Architecture to mitigate the interference for IEEE 802.11 b/g/n sources and to enhance the performance of IEEE 802.15.4 network for an Internet of Things deployment scenario. The IoT deployment would be done us heterogeneous network as discussed above, and such networks can be highly dynamic based on the application. Dynamic in terms of the scalability of the network i.e. more number of IoT smart devices may be added to the network in real-time or on the contrary the device might log out or go to a sleep mode based on application requirement. Further such IoT networks may be a complex combination of both static of mobile nodes again based on application, in such case again the network becomes highly dynamic. In case of such dynamic networks, it is challenging to do spectrum planning or optimization of radio channel allocation before network deployment. Thus a real-time and self-optimizing approach is the most pragmatic solution in such IoT network deployment.

1) Spectrum Sensing
Spectrum Sensing is the technique adapted from the concept of Cognitive Radio. This technique will come handy especially for real-time interference mitigation. The proposed Cognitive Radio Algorithm would be implemented on the IEEE 802.15.4 nodes to mitigate the interference from any IEEE 802.11 b/g/n source present in the proximity. Spectrum sensing technique would enable the IEEE 802.15.4 nodes to scan the radio channels using CSMA to detect any presence of IEEE 802.11 b/g/n network in the area. If it detects any such network, then it will further examine other essential parameters of such network like RSSI (-dBm) to get the severity of interference from such source. In addition to that, it will try to read the PLCP to get information about the type as b/g/n and data-rate such s 11/54/72/144 Mbps. Once all the parameters are gathered by spectrum sensing it will be fed to the Cognitive Radio Algorithm to take the decision of selecting the optimum radio channel by avoiding spectral overlapping to minimize interference on IEEE 802.15.4 from the IEEE 802.11 source. Spectrum sensing mechanism can be periodic or event-driven based on the application.

2) Interference Index
The proposed Cognitive Radio Algorithm based on realtime spectrum sensing use the Interference Index as its primary parameter to mitigate interface. Interference Index is a relative mathematical calculation based on the parameters obtained from spectrum sensing as discussed above. Interference Index would be calculated based on the interference analytical model fine-tuned by the empirical model to find BER from SINR in equation (4) and selecting the appropriate by finding the type of interference network present. If the IEEE 802.11 b is present as an interference source then to calculate Interference Index its will use the analytical model of equation (7) similarly for IEEE 802.11 g, 802.11n 20 MHz and 802.11n 40 MHz it will use equation (9), (10) and (11) respectively. Fig.11 shows the proposed Cognitive Radio Algorithm based on Real-time Spectrum Sensing as discussed above.The Interference Index can be calculated. The proposed Cognitive radio Algorithm Fig. has 3 stages of operation. Stage 1 is real-time spectrum sensing and gathering parameters like type of network b/g/n, data rate 11/54/72/144 Mbps, RSSI (-dBm) and Radio Channel about IEEE 802.11 interference source. Stage 2 of the CR Algorithm is to calculate Interface Index based on the interference power of . Further SINR(ch z ), BER(ch z ) and PER(ch z ) would be calculated, where ch z means z th channel. The 3 rd and final stage of the algorithm is the decision making a stage to find out the minimum interference radio channel deployment of IEEE 802.15.4 based on the interference source present in the proximity. The minimum interference radio channel is given as Best Channel of deployment as given in equation (19).

4) Cognitive Radio Algorithm implementation of low-
power IoT nodes From the perspective of a deployment Internet of Things Sensor/Actuators nodes based on of IEEE 802.15.4 Zigbee Network, the nodes are low data rate and low power devices specially designed for longer service time powered by limited battery life. The Cognitive Radio algorithm proposed above is having some significant amount of computational requirement especially to calculate SINR(ch z ), BER(ch z ) and PER(ch z ) based on received , which in terms consume of vital battery life and reduce the service time of such IoT motes. To deal with the above problem the proposed architecture in the scenario of low power IoT devices can be further optimized by implementing the modified CR algorithm.
The modified CR algorithm is to make the algorithm less computationally regressive by aiding all the possible channel overlapping scenario interference pattern as per Table 1 & 2. The SINR(ch z ), BER(ch z ) and PER(ch z ) can be precalculated and stored in a database of the IoT motes. The database file size is also minimum possibly three files b/g/n 20MHz and one file for n 40Mhz overlapping scenario. Each file would content PER(ch z ) for IEEE 802.11 b/g/n 20 MHz 4 Overlapping pattern x 14 = 56 float data type i.e 672 bytes total for three files. The IEEE 802.11 n 40Mhz overlapping pattern 8 x 8 channel =64 float i.e. 256 bytes Table. Total required database size is 928 bytes. Thus the whole set of learning data set can be implemented no more than 1 KB database, which can be easily implementable in IoT motes.
The IoT motes aided with spectrum sensing can determine the radio channel overlapping scenario between the IEEE 802.15.4 network and the interference source as IEEE 802.11. One the overlapping pattern is finding as per Table 1 & 2, the expected PER(ch z ) would be taken from the look-up table of the databases just been discussed above and finally implementing equation (19) the best channel with least interference can be selected and the IoT mote will select that as its radio channel to communicate in the PAN.

III. RESULTS AND DISCUSSION
The decisive parameter for performance here is considered as Packet Error Rate (PER) [17]. PER is having more significance than BER as it includes the data-link layer timing parameters as given in Table.4 for IEEE 802.11 b/g/n and in Table. Fig.14. The peak PER value is around 0.025 for the analytical model and 0.015 for empirical test-bed values for the maximum spectral overlapping scenario with the minimum frequency offset of 3MHz. This scenario is almost analogous to the IEEE 802.11n HT 20MHz case with little peak PER change of 0.020 can be explained by the further less Collision time as shown in Figure 12, which is by equation (17).
It can be inferred from the test results that, IEEE 802.11b network poses the most severe interference to adjacently located IEEE 802. 15 Zigbee PAN particularly when the IEEE 802.11 interference source is present in the 2-meter proximity for PAN. This 2meter proximity case is highly possible especially in the IoT deployment scenario as shown in Fig.1, which might be an area of concern for IoT deployment using a heterogeneous wireless network. The scope of this paper includes not only to quantify the matrices for interference but also to suggest a mitigation technique specially designed for the IoT deployment environment using heterogeneous wireless networks. In the further research, a Cognitive Radio algorithm should be proposed based on the real-time spectrum sensing technique. This algorithm would be implemented based on the interference index obtained from the testbed as well as the analytical model. The Cognitive Radio Algorithm will take decisions to select an optimum radio channel for IEEE 802.15.4 Zigbee PAN network, which will help in mitigating the interference from adjacently deployed IEEE 802.11 b/g/n network. The Cognitive Radio Algorithm will consider the BER for all possible spectral overlapping conditions and Collision time to finally get PER to decide least PER channel deployment to mitigate interference.

IV. CONCLUSION
This paper has suggested a Cognitive Radio Algorithm for IoT motes deployment scenario using IEEE 802.15.4 Zigbee WPAN to mitigate interference from IEEE 802.11 b/g/n sources. The Cognitive Radio Algorithm is proposed based on the impact of such interface using analytical model, corroborated by experiments performed using XBEE® IEEE 802.15.4 Zigbee motes in the presence of IEEE 802.11 b/g/n interference source.