1514 IEEE TRANSACTIONS ON INSTRUMENTATION AND MEASUREMENT, VOL. 57, NO. 8, AUGUST 2008 Experimental Study of Coexistence Issues Between IEEE 802.11b and IEEE 802.15.4 Wireless Networks Leopoldo Angrisani, Member, IEEE, Matteo Bertocco, Member, IEEE, Daniele Fortin, Student Member, IEEE, and Alessandro Sona, Member, IEEE Abstract—Coexistence issues between IEEE 802.11b wireless communication networks and IEEE 802.15.4 wireless sensor networks, operating over the 2.4-GHz industrial, scientific, and medical band, are assessed. In particular, meaningful experiments that are performed through a suitable testbed are presented. Such experiments involve both the physical layer, through measurements of channel power and the SIR, and the network/transport layer, by means of packet loss ratio estimations. Different configurations of the testbed are considered; major characteristics, such as the packet rate, the packet size, the SIR, and the network topology, are varied. The purpose of this paper is to gain helpful information and hints to efficiently face coexistence problems between such networks and optimize their setup in some real-life conditions. Details concerning the testbed, the measurement procedure, and the performed experiments are provided. Index Terms—IEEE 802.11b, IEEE 802.15.4, packet loss ratio (PLR), signal-to-interference ratio (SIR), wireless local area network (WLAN), wireless sensor network (WSN). I. I NTRODUCTION W IRELESS communication networks are receiving increasing interest in the scenario of digital communications. With respect to the wired counterpart, they avoid the need for expensive cabled infrastructure, improve network scalability and reconfigurability, and allow the efficient use of portable and mobile terminals. A key issue of wireless networks is their typical susceptibility to radio interference and the fact that they are commonly deployed in areas that are already crowded of radio disturbances, generated by nearby operating electrical and electronic appliances and other wireless apparatuses. The presence of interference over the same frequency band may even lead to disruptive effects in the transmission of data packets among wireless stations (STs). Typical interference effects are loss of data packets, transmission delay, false commands, false alarms, jitter, and loss of synchronization. A critical case of interference involves IEEE 802.11b wireless local area networks (WLANs) [1] and IEEE 802.15.4 wireless sensor networks (WSNs) [2]. Two such networks exploit the same frequency band, i.e., the 2.4-GHz industrial, scientific, Manuscript received July 5, 2007; revised April 14, 2008. L. Angrisani is with the Department of Computer Science and Control Systems, University of Naples Federico II, 80125 Naples, Italy (e-mail: angrisan@ unina.it). M. Bertocco, D. Fortin, and A. Sona are with the Department of Information Engineering, University of Padova, 35131 Padova, Italy (e-mail: matteo. bertocco@unipd.it; daniele.fortin@unipd.it; alessandro.sona@unipd.it). Digital Object Identifier 10.1109/TIM.2008.925346 and medical (ISM) band, and are widely deployed in a number of common applications in which they have to coexist in close proximity [3]. In these cases, some interference effects should always be expected, unless one would take into account specific coexistence issues and adopt proper interference avoidance solutions. Coexistence issues between IEEE 802.15.4 WSNs and IEEE 802.11b WLANs have been studied in the past [3]–[8]. In particular, in [3] and [4], the coexistence impact of an IEEE 802.15.4 network on IEEE 802.11b devices is analytically investigated, and a predicting model is proposed. The model is powerful but limited to the specific network configuration analyzed and to the environmental conditions considered. In [5]–[7], the interference effects of IEEE 802.15.4 networks over IEEE 802.11b WLANs and vice versa are analyzed, both analytically and through simulations. Simulation results are given in terms of packet error rates, throughput, and transmission time at different distances between the two networks. In [8] and [9], coexistence issues between wireless networks (IEEE 802.15.4, IEEE 802.11b, and IEEE 802.15.1) are investigated through measurements. In particular, in [8], several coexistence scenarios are analyzed, including the case of IEEE 802.15.4 and IEEE 802.11b coexisting networks. However, the few reported results, although very interesting, do not provide a comprehensive experimental overview of the phenomenon under examination, which, to the best knowledge of the authors, has yet to be reported in the literature. Helpful information about how to arrange the testbed and to perform cross-layer measurements is finally provided in [9] and [10], even if it does not specifically refer to the case of an IEEE 802.11b WLAN coexisting with an IEEE 802.15.4 WSN. In this paper, coexistence problems between an IEEE 802.11b WLAN and an IEEE 802.15.4 WSN operating within a real-life environment are experimentally analyzed. In particular, an original comprehensive set of experiments based on packet loss ratio (PLR) estimates is presented, involving the cases of a WLAN under the interference of a WSN and a WSN under the interference of a WLAN. A final overview of the deduced coexistence conditions is also provided. Stemming from the past experience documented in [11], the purpose is to deduce helpful information and hints for designers and technicians, which are to be efficiently used to optimize the network design and setup in common reallife coexistence conditions. A number of experiments are conducted through a proper testbed, and a measurement approach that is similar to that proposed in [9] and [10] is followed. 0018-9456/$25.00 © 2008 IEEE Authorized licensed use limited to: Georgia Institute of Technology. Downloaded on February 21, 2009 at 00:43 from IEEE Xplore. Restrictions apply. ANGRISANI et al.: COEXISTENCE ISSUES BETWEEN IEEE 802.11b AND IEEE 802.15.4 WIRELESS NETWORKS Fig. 1. WLAN frequency channels. II. N OTES F ROM S TANDARDS A. IEEE 802.11b WLAN A WLAN is generally a set of computing devices and access points (APs) communicating with each other through radio waves and over common frequency channels [1], [12]. The IEEE 802.11b standard defines a total of 14 frequency channels, each of which characterized by a 22-MHz bandwidth (Fig. 1). In the USA, channels 1–11 are allowed by the Federal Communications Commission, whereas in Europe, channels 1–13 can be used. In Japan, one choice is allowed—channel 14. Fig. 1 shows that only three channels at a time (e.g., 1, 6, and 11) can be used without overlap. In the design of a WLAN, the assigned channel and spatial position of each AP should be fixed in such a way as to avoid frequency overlap over the whole coverage area. In a typical communication, an AP periodically transmits a frame called beacon, which contains information such as a network identifier (ID), the beacon and channel parameters, and other traffic information. Each network ST receives the beacon, and if it is aimed at accessing the network, it sends a request for authentication. A series of tests is then performed to ensure the identity of the ST. Once authenticated, an ST may communicate to the AP or vice versa according to a carrier sense multiple access with collision avoidance (CSMA/CA) protocol. In particular, an ST (or the AP) wishing to transmit senses the channel, and, if no activity is detected, it waits for an additional randomly defined back-off period and then transmits if the medium is still free. If the packet is received without corruption, the AP issues an acknowledge (ACK) frame that completes the process. If the ST does not detect the ACK frame within a fixed timeout period, a collision is assumed, and the data packet is retransmitted after another random back-off period. Usually, to assess the channel status, the clear channel assessment (CCA) technique in mode 1 is implemented [1]. In this mode, the channel is assumed as idle if the channel power level is below a given user selectable threshold, which is called the CCA threshold; otherwise, it is assumed to be busy. A schematic representation of the described communication protocol is shown in Fig. 2(a). A packet fragmentation is performed before the transmission, which allows breaking of large packets into smaller parts of length d, which is useful in disturbed environments since large 1515 Fig. 2. Time diagrams of (a) WLAN and (b) IEEE 802.15.4 frames. packets can be more easily corrupted. Two overhead frames are then added to each data packet—a header h, which contains information such as the ST identifier number, the packet sequence number, the sender and receiver addresses, and the total packet length (d + h + c); and a control frame c, which allows errors to be handled. In Fig. 2(a), the time interval covered by the packet is denoted as τwlan , whereas the time interval between two packets is denoted as Tpck . The two delays td1 and td2 between packets and the occurrence of ACK vary from one packet to another, mostly depending on channel characteristics, network topology, and back-off time. To indicate the percentage of Tpck that is specifically dedicated to the transmission of the whole packet (d + h + c), the duty cycle λwlan is commonly introduced [9]. It is defined as follows: λwlan = τwlan /Tpck . B. IEEE 802.15.4 WSN IEEE 802.15.4 is a standard designed for a low-rate wireless personal area network (PAN) characterized by limited coverage area, reduced cost, and low energy consumption. It is one of the most promising technologies for implementing WSNs [2]. A variety of network topologies is known; an interesting one is the star configuration, where a PAN coordinator (the master) cyclically queries a set of slaves (sensor nodes), one by one, and slaves reply by transmitting a short data packet (usually tens of bytes) containing the information acquired by the associated sensor. The transmission between the master and the slaves exploits the 2.4-GHz ISM band at 250 kb/s and, in particular, one of the 16 available channels. As shown in Fig. 3, the channels have a 3-MHz bandwidth and are uniformly distributed within the ISM band. In Fig. 3, three nonoverlapped WLAN channels are also shown to highlight that WSNs and WLANs can occupy the same frequencies and that the hazard of in-channel interference is very high whenever a WSN operates in the nearness of one or more active WLANs. The master dedicates a time interval, called the polling window, of fixed time length Tpoll to each sensor node. As schematically represented in Fig. 2(b), the master starts waiting a back-off period td1 and then senses the air according to the CSMA/CA protocol. If the channel is free, it transmits a frame (query) q throughout the network, containing information such as the destination node address, the sequence number of the performed cycle, and the time stamps. After a pre-fixed short Authorized licensed use limited to: Georgia Institute of Technology. Downloaded on February 21, 2009 at 00:43 from IEEE Xplore. Restrictions apply. 1516 IEEE TRANSACTIONS ON INSTRUMENTATION AND MEASUREMENT, VOL. 57, NO. 8, AUGUST 2008 Fig. 3. Frequency channels peculiar to (a) IEEE 802.15.4 and (b) IEEE 802.11. Fig. 4. Adopted testbed. delay td2 , the queried sensor node replies with an ACK and waits for a second back-off period td3 . It then senses the air and, if it is free, transmits the data packet d with a header frame h and a tail field c. In the case of correct reception, the master issues an ACK, and at the end of the polling window, it passes to the subsequent sensor node. If the sensor does not receive the ACK from the master, it waits for another back-off period and then retries to transmit. At the expiration of the polling window, if the ACK from the master has yet to be issued, the packet from that queried sensor is considered lost. III. T ESTBED Experiments have been carried out by using the testbed sketched in Fig. 4, which consists of an IEEE 802.15.4 WSN and an IEEE 802.11b WLAN operating in close proximity. The WLAN is simply realized through a standard AP and an ST placed at a distance d = 13 m from it. Although limited to only two terminals, the configuration can be considered general and representative of situations with more than two WLAN STs. In fact, according to the CSMA/CA mechanism, only one WLAN ST may transmit at a time and, thus, interfere with the WSN regardless of the number of WLAN nodes that are involved. The AP is a D-link 524 device connected to a PC via a wired link, whereas the ST is a notebook Toshiba Satellite (Intel Centrino 1.60 GHz) with an Intel Pro/Wireless integrated WLAN transceiver module. Traffic is generated through the tool Distributed Internet Traffic Generator (D-ITG) [13], which allows the setting of the transport layer protocol (Transmission Control Protocol, User Datagram Protocol, Internet Control Message Protocol) and application layer functionality—Domain Name System, Telnet, voice, or custom. In the custom mode, the packet rate, the packet size, and the interdeparture time statistical distribution can be varied and set according to the desired configuration. Also, it allows measurements of quality-of-service parameters, such as the packet rate, the PLR, the packet average delay, and the packet jitter, both at the receiver and transmitter sides. The IEEE 802.15.4 network is a set of TmoteSky sensor nodes (motes), available from Moteiv, randomly distributed inside a circumference of radius r = 2 m and centered around the ST. The motes operate according to the general-purpose high-layer protocol based on a master–slave relationship. The network includes one master M and N = 10 slaves, each of which is equipped with the following basic elements: 1) one microprocessor; 2) one or more analog input sections, which are to be interfaced with one or more sensing elements; 3) a buffer for acquired data; and 4) a radio communication module. M is connected to a PC via a wired link for storing the collected data and postprocessing. As described in Section II, it executes periodical polling of each slave for receiving data from monitoring sensors. All transmissions are handled by the medium access control layer, which allows the radio channel to be accessed and the data packets to be retransmitted in case of losses or missing acknowledgments. Traffic is generated and monitored through the tool WSN [14], which allows the setup of the master and slaves and the display of the data collected by the sensors. Coexistence analysis has been performed in the following two scenarios. 1) WLAN is a victim of WSN interference. 2) WSN is a victim of WLAN interference. In both scenarios, the two networks have been forced to operate on overlapped frequency channels, i.e., channel 11 for the WLAN and channel 22 for the WSN. Measurements have been conducted both at physical and network/transport layers [9], [10]. In particular, at the physical layer, the channel power and the SIR have been estimated by using a National Instrument PXI 5660 module RF signal analyzer and a receiving antenna Seibersdorf PCD 8250 biconical dipole. At the network/transport layer, the PLR, i.e., the ratio between the number of lost packets and the number of transmitted packets, has been estimated by using D-ITG [13] and the WSN tool [14], running on the two PCs. Different configurations of WSN and WLAN parameters have been considered. In particular, two values of Tpoll have been taken into account—Tpoll = 30 ms, which is the minimum allowed value, below which the number of lost packets in the absence of interference abruptly worsens (about 100% [15], [16]), and Tpoll = 100 ms. In the two analyzed scenarios, a different meaning has been given to the SIR. In scenario 1, SIR = Pwlan /Pwsn , where Pwlan is the WLAN useful power, and Pwsn is the WSN interference power, both measured over the WLAN channel and at the space position occupied by the ST. In scenario 2, SIR = Pwsn /Pwlan , where Pwsn and Pwlan are measured over Authorized licensed use limited to: Georgia Institute of Technology. Downloaded on February 21, 2009 at 00:43 from IEEE Xplore. Restrictions apply. ANGRISANI et al.: COEXISTENCE ISSUES BETWEEN IEEE 802.11b AND IEEE 802.15.4 WIRELESS NETWORKS the WSN channel and by averaging the power readings achieved at any sensor node position. For the WSN and the WLAN, experiments have been executed in CCA mode 1, with a CCA threshold level Pth equal to −76 dBm (the default value). Moreover, measurements have been performed inside a nonshielded and nonanechoic room, hence, with the potential occurrence of external interference and collateral phenomena such as reflections from near metallic objects and structures. To also properly account for this phenomenon, different positions of sensor nodes have been considered. IV. E XPERIMENTS IN THE C ASE OF THE WLAN V ICTIM The WLAN behavior under the effect of WSN interference has been investigated both in the case of the AP transmitting to the ST and vice versa. A. Transmission From the AP to the ST The conducted experiments have aimed to assess the behavior of the WLAN under the effect of the WSN when the ST (located in the proximity to the WSN) receives data packets from the AP (located far way from the WSN). This configuration has emulated the typical situation in which the WLAN transmitter (here, the AP) senses an interference power level (from the WSN) below its CSMA/CA threshold, i.e., the WLAN channel is sensed free, thus enabling the transmission. The first experiments have been conducted measuring the PLR of the WLAN, i.e., PLRwlan , in the case of SIR = −5.6 dB, for different values of the WLAN duty cycle λwlan and the duration of the transmission window equal to 60 s. PLRwlan has been estimated as the ratio PLRwlan = 1 − Nrp /Ntp , where Nrp is the number of correctly received packets, and Ntp is the number of packets delivered by the WLAN transmitter (here, the AP) in ideal conditions (i.e., in the absence of interference). The quantity Ntp − Nrp can, in turn, be considered as the sum of two contributions, i.e., Ntp − Nrp = Nerp + Nntp , where Nerp is the number of packets erroneously gathered by the WLAN receiver (here, the ST) due to signal collisions with WSN packets and/or channel errors, whereas Nntp is the number of packets not delivered by the WLAN transmitter (here, the AP) due to channel occupation. According to the above assumption about the AP location with respect to the WSN position, Nntp is expected to be null in the analyzed configuration. The obtained results are summarized in Fig. 5, in which three diagrams are shown, where each referred to a specific size of WLAN packets pswlan —600 B [Fig. 5(a)], 1112 B [Fig. 5(b)], and 1563 B [Fig. 5(c)], with a payload of 512, 1024, and 1475 B, respectively. Different operative conditions of the testbed have been analyzed—WSN off and WSN on with Tpoll = 30 ms and Tpoll = 100 ms. The solid lines represent the PLRwlan estimates, whereas the dashed lines give the percentage of packets that are not transmitted by the AP due to channel occupation, i.e., Nntp /Ntp . Vertical bars also indicate the standard deviation experimented under different positions of the sensors. 1517 From the analysis of the results, the following conclusions can be drawn. • In Fig. 5(a) and (b), as expected, the percentage of nottransmitted packets Nntp /Ntp is always null, irrespective of the chosen λwlan . This confirms that the AP always finds the channel free, and it can transmit without interruption. • In Fig. 5(c), Nntp /Ntp is not always null and increases upon the increase of λwlan . This means that despite the fact that the channel is always assessed free, the AP encounters problems when managing very long packet sizes. • PLRwlan assumes quite the same values for pswlan = 600 and 1112 B, whereas much higher losses are experienced with pswlan = 1563 B. For instance, with λwlan = 75%, WSN on, and Tpoll = 30 ms, the resulting PLRwlan is 15%, 10%, and 100%, respectively. • In the absence of WSN interference, nonnegligible degradation of WLAN performance is visible for any considered λwlan . In particular, PLRwlan is constant and nearly equal to 5% up to a given threshold level of λwlan and is equal to 75%, 75%, and 53% for Fig. 5(a)–(c), respectively, whereas it increases even abruptly [see Fig. 5(c)] for higher values. • In the presence of WSN interference and Tpoll = 30 ms, WLAN performance degrades further. This is particularly true in the case of pswlan = 1563 B and λwlan ≤ 53% for which PLRwlan changes from nearly 5% (without interference) up to 70% (for λwlan = 53%). • In the presence of WSN interference, the threshold values of λwlan , beyond which PLRwlan increases, reduce from 75%, 75%, and 53% to 65%, 60%, and 30% for Fig. 5(a)–(c), respectively. • With a longer Tpoll of 100 ms, the effect of the WSN on the WLAN is much weaker. In fact, with Tpoll = 100 ms, the estimated PLRwlan values are very close to the results achieved with WSN off. • The dispersion of PLRwlan estimates around the corresponding mean values is almost always negligible. The only exception is given in Fig. 5(c), with WLAN on and Tpoll = 30 ms. Other experiments have been carried out with different levels of the SIR from −13 up to −3 dB, two values of λwlan = 56% and 75%, two values of Tpoll = 30 and 100 ms, and pswsn = 1112 B. The SIR has been changed by acting only on Pwlan . The obtained results are summarized in Fig. 6, and the following considerations arise. • In the transition range −13 ≤ SIR ≤ −7 dB, PLRwlan abruptly decreases from 100% to very low levels, i.e., below 13%, upon the increase of the SIR. • In the same range, the effect of Tpoll on PLRwlan is clearly visible; stretching Tpoll from 30 to 100 ms, PLRwlan considerably decreases. For instance, with SIR = −11 dB and λwlan = 56%, PLRwlan is lowered from 34% to 10% by increasing Tpoll from 30 to 100 ms. • In the upper range, i.e., SIR > −7 dB, the effect of Tpoll and λwlan on PLRwlan is negligible. In fact, the same values of PLRwlan , within 5%–12%, are achieved, regardless of the chosen Tpoll and λwlan . Authorized licensed use limited to: Georgia Institute of Technology. Downloaded on February 21, 2009 at 00:43 from IEEE Xplore. Restrictions apply. 1518 IEEE TRANSACTIONS ON INSTRUMENTATION AND MEASUREMENT, VOL. 57, NO. 8, AUGUST 2008 Fig. 5. PLR of the WLAN versus λwlan for different packet sizes. (a) 600 B. (b) 1112 B. (c) 1563 B. • For levels of the SIR below −13 dB, WLAN packets are completely lost, i.e., PLRwlan = 100%, regardless of the chosen value of λwlan and Tpoll . This is due to the reduced level of Pwlan that, for the considered SIR values, is comparable with or lower than the sensitivity level of the ST, which is typically on the order of −84 dBm (in the case of data rate = 11 Mb/s). • The experimental standard deviation is almost always negligible, i.e., below 10%, with the exception of the highest PLRwlan . B. Transmission From the ST to the AP The conducted experiments have aimed at assessing the behavior of the WLAN under the effect of the WSN interference when the ST (located in the proximity to the WSN) is transmitting data packets to the AP (located far way from the WSN). This configuration has emulated the typical case in which the WLAN transmitter (here, the ST) senses a high level of the WSN signal beyond its CCA threshold Pth . In this case, the presence of the WSN signal saturates the WLAN channel, Authorized licensed use limited to: Georgia Institute of Technology. Downloaded on February 21, 2009 at 00:43 from IEEE Xplore. Restrictions apply. ANGRISANI et al.: COEXISTENCE ISSUES BETWEEN IEEE 802.11b AND IEEE 802.15.4 WIRELESS NETWORKS Fig. 6. PLR of the WLAN versus the SIR for different λwlan . Fig. 7. PLRwlan versus λwlan for a packet of 1112 B. and the ST, according to the CSMA/CA protocol, is forced to defer the transmission. Consequently, data packets are delivered from the ST to the AP with an amount of delay, and increased values of PLRwlan can be experimented with. The same procedure described in Section IV-A has been applied in a number of experiments, in which the only case of pswlan = 1112 B has been considered. The obtained results are summarized in Fig. 7, and the following considerations from their analysis are of concern. • The percentage of not-transmitted packets Nntp /Ntp , represented by dashed lines in the figure, is not always null and increases upon the increase of λwlan . • The three dashed lines completely overlap the corresponding solid lines that are obtained with the same value of Tpoll and with WSN on/off, i.e., PLRwlan = Nntp /Ntp for any considered λwlan . • PLRwlan increases significantly for higher λwlan values. For instance, with λwlan = 87%, PLRwlan is nearly 27% and 31% for Tpoll = 100 and 30 ms, respectively. 1519 • With respect to Fig. 5(b), quite the same values of PLRwlan have been obtained in the presence of interference. C. Comments and Hints Relevant comments and hints deriving from the obtained results are given in the following. 1) Both with and without WSN interference, the WLAN PLR strictly depends on λwlan and increases for λwlan values beyond a given threshold. It is advisable to set λwlan below such a threshold, which represents a key parameter in the design stage, to be accurately known and, if necessary, measured, as shown above. 2) Both with and without WSN interference, the WLAN PLR also depends on the packet size. In particular, whereas very similar results have been achieved with pswlan = 600 and 1112 B, much higher values have been experienced with pswlan = 1563 B and for λwlan > 55%. The use of a packet size of 1112 B is suggested; it allows the improvement of PLRwlan in the presence of Authorized licensed use limited to: Georgia Institute of Technology. Downloaded on February 21, 2009 at 00:43 from IEEE Xplore. Restrictions apply. 1520 IEEE TRANSACTIONS ON INSTRUMENTATION AND MEASUREMENT, VOL. 57, NO. 8, AUGUST 2008 Fig. 8. PLRwsn versus λwlan for different WLAN packet sizes—600, 1112, and 1563 B—and Tpoll = 30 ms. 3) 4) 5) 6) 7) interference with respect to pswlan = 1563 B and higher throughput with respect to pswlan = 600 B. This result agrees with the values of the packet length obtained by simulations in [6] for the same case of an IEEE 802.11b WLAN under the interference of an IEEE 802.15.4 WSN. The WLAN PLR also strictly depends on the SIR. In particular, PLRwlan may abruptly worsen for SIR values that are below a given threshold, whereas for upper values, it assumes low levels and becomes quite independent of WSN characteristics such as polling time. The SIR should be always greater than this threshold, which can, in turn, be efficiently measured as described. For the deployed testbed, the effects of WSN interference, in terms the WLAN PLR, are the same, regardless of the communication direction between the AP and the ST. A different interference phenomenon arises depending on the case of the AP transmitting to the ST or vice versa. In particular, in the case of the AP transmitting to the ST, i.e., when the WLAN source is located far way from the WSN, interference comes out as collisions between WSN and WLAN data packets and/or channel errors. Conversely, in the case of the ST transmitting to the AP, i.e., when the WLAN source is closely located to the WSN, interference comes out as channel occupation due to the presence of WSN emissions in the WLAN channel and the use, in the WLAN transmitter, of the CSMA/CA mechanism. The increase in WSN polling time from 30 to 100 ms allows the mitigation of the interference effects of the WSN on the WLAN, particularly for λwlan and SIR values corresponding to when the estimated PLRwlan is greater. In all the analyzed cases, very short standard deviations have been experienced. External interference or collateral phenomena, such as reflections, have not introduced significant perturbations on conducted measurements. V. E XPERIMENTS IN THE C ASE OF THE WSN V ICTIM A second set of experiments has been carried out with the WSN victim and the WLAN interferer. Only the configuration of the AP transmitting to the ST has been considered. It allows a better investigation into the behavior of the WSN under WLAN interference, avoiding the effects of the WSN on the farawaylocated AP. In fact, because of the low power level received by the AP from the WSN, i.e., below its CCA threshold, the AP always finds the channel free, and consequently, it can continuously transmit, interfering the WSN without interruption. This is not true in the other case, i.e., the ST transmitting to the AP, because of the near operation of WSN nodes, which periodically require accessing the same channel, hence with the hazard of disturbing the ST itself. A. Results Measurements have been conducted with three different WLAN packet sizes—pswlan = 600, 1112, and 1563 B—at the same level of sensor node output power Pmote = 0 dBm and Tpoll = 30 ms, which guarantee, in the absence of radio interference, a null WSN PLR (denoted here as PLRwsn ) [15]. The first experiments have been carried out with SIR = −3 dB and Pwlan greater than the CCA threshold of WSN motes Pth . In Fig. 8, the obtained results in terms of the WSN PLR, i.e., PLRwsn , upon varying the WLAN duty cycle λwlan are summarized. The diagram clearly shows that, in the presence of WLAN interference, PLRwsn abruptly degrades from 0% (with WLAN off) up to values that are always greater than 50% for λwlan beyond 40%. Moreover, PLRwsn increases quite linearly and quickly toward a saturation level comprised within 75%–85%. In these cases, the probability for a WSN to find the channel free, hence to begin transmission, is very low, and it worsens for shorter WLAN packet sizes. This phenomenon is clearly visible for values of λw up to 35% and is simply due to the fact that, for a given λw , a shorter time interval between two consecutive WLAN packets (Tpck − τw ) is available for WSN transmission. To avoid such critical degradation of WSN performance in the presence of WLAN interference, an efficient solution has consisted of increasing Tpoll , for instance, from 30 to 100 ms. Some tests have been performed with the suggested solution, Authorized licensed use limited to: Georgia Institute of Technology. Downloaded on February 21, 2009 at 00:43 from IEEE Xplore. Restrictions apply. ANGRISANI et al.: COEXISTENCE ISSUES BETWEEN IEEE 802.11b AND IEEE 802.15.4 WIRELESS NETWORKS Fig. 9. 1521 PLRwsn versus λwlan for two values of Tpoll —30 and 100 ms. Fig. 10. PLR of the WSN versus the SIR for different polling window values—30 and 100 ms. and the obtained results are summarized in Fig. 9; only one packet size, i.e., 1112 B, and two values of Tpoll , i.e., 30 and 100 ms, have been taken into account. Fig. 9 highlights that in the case of Tpoll = 100 ms, reduced values of PLRwsn (below 10%) have been achieved with λwlan ≤ 35%. Further results are given in Fig. 10, which are obtained with psw = 1110 B, with λwlan = 60%, and upon varying the value of the SIR. The diagram shows that PLRwsn proportionally decreases upon the increase of the SIR, which is always below a given threshold. Such a threshold value is nearly 16 and 8 dB for Tpoll = 30 and 100 ms, respectively. Beyond the threshold, PLRwsn assumes quite constant values, i.e., nearly 10% and 0% for Tpoll = 30 and 100 ms, respectively. 2) Interference effects of a WLAN on a WSN can be mitigated increasing Tpoll and reducing λwlan below a suitable threshold. Therefore, a WSN may coexist with a near-operating WLAN but to the detriment of the throughput of both the WLAN (lower λwlan ) and the WSN (higher Tpoll ). 3) The maximum λwlan and the minimum Tpoll to guarantee the desired levels of PLRwlan are essential parameters to be taken into account in a design stage. They can be determined through measurements such as those described above. 4) PLRwsn strictly depends on the SIR, which should not be lower than a given threshold. Such a threshold varies upon varying Tpoll and can be experimentally determined, as shown above. B. Comments and Hints Relevant comments and hints deriving from the obtained results are given in the following. 1) A WLAN may considerably degrade the performance of a nearby operating WSN, increasing the values of PLRwsn , even up to 80%. VI. C OEXISTENCE I SSUES The obtained results have underlined the importance of the two parameters pswlan and λwlan in the optimization of the performance of both IEEE 802.15.4 and IEEE 802.11 networks when operating in the same environment and Authorized licensed use limited to: Georgia Institute of Technology. Downloaded on February 21, 2009 at 00:43 from IEEE Xplore. Restrictions apply. 1522 IEEE TRANSACTIONS ON INSTRUMENTATION AND MEASUREMENT, VOL. 57, NO. 8, AUGUST 2008 Fig. 11. (Dashed lines) PLRwsn and (solid lines) PLRwlan for a different setup of a WLAN. sharing the same frequencies. Their choice can be simplified and optimized by using comprehensive diagrams such as that in Fig. 11, which is obtained by gathering the results of Figs. 5 and 8. The diagram shows two sets of curves—the dashed ones give the values of pswlan and λwlan that guarantee a given level of PLRwsn (indicated through vertically oriented numbers), whereas the solid ones represent the values of pswlan and λwlan that guarantee a given level of PLRwlan (indicated through horizontal-oriented numbers). The two sets of curves refer to Tpoll = 30 ms. The diagram fixes the couple of pswlan and λwlan values allowing the increase in the WLAN throughput (which can be pursed increasing pswlan and/or λwlan ) and the minimization of PLRwsn and PLRwlan . For instance, with a maximum PLRwsn of 24.4% and a maximum PLRwlan of 9.0%, the diagram shows that the two parameters pswlan and λwlan must be in the region on the left of the curve with PLRwsn = 24.4%, which means that a very low λwlan must be chosen, and consequently, low WLAN throughput is obtained. Instead, in the case of a maximum PLRwsn of 79.0% and a maximum PLRwlan of 9.0%, a much greater region of pswlan and λwlan values is available. In this case, it can be noted that to increase the WLAN throughput, a solution is to reduce the WLAN packet size, which allows greater values of λwlan to be set. In general, the graph confirms that the predominant effect of interference is from the WLAN to the WSN. In fact, with pswlan ≤ 1112 B, all the PLRwsn curves are on the left with respect to the corresponding ones (having the same PLR level) of PLRwlan . Therefore, in this case, for any chosen λwlan , PLRwlan PLRwsn . VII. C ONCLUSION The experimental analysis has confirmed that the WLAN and the IEEE 802.15.4 WSNs may coexist when operating in close proximity but to the detriment of some performance parameters such as the PLR and the throughput. In the case of the WLAN victim, the WLAN duty cycle must always be set lower than a given threshold value, which depends on the WLAN packet size, decreases in the case of WSN interference, and can be determined as described above. Moreover, in the presence of WSN interference, the WLAN PLR does not significantly depend on the number and the position of WSN nodes inside a fixed circumference that is centered around the WLAN receiver. In the opposite case, i.e., the WSN victim, much more visible loss of WSN packets arising when the WLAN transmits has been observed. In particular, at the beginning of the WLAN operation, the PLR abruptly increases from 0% to values beyond 70%. In this case, to enhance the WSN reliability, a solution may be either to increase the time duration of the polling window in such a way as to have more possibilities to correctly retransmit or to reduce the WLAN duty cycle. Unfortunately, this solution causes lower throughput of the WSN, if a longer polling window is set, or of the WLAN, if the duty cycle is reduced. R EFERENCES [1] Part 11: Wireless LAN Medium Access Control (MAC) and Physical Layer (PHY) Specifications, IEEE Std. 802.11b 1999/Cor 1-2001, 2001. [2] Part 15.4: Wireless Medium Access Control (MAC) and Physical Layer (PHY) Specifications for Low-Rate Wireless Personal Area Networks (LRWPANs), IEEE Std. 802.15.4, Oct. 2003. [3] I. Howitt, “WLAN and WPAN coexistence in UL band,” IEEE Trans. Veh. Technol., vol. 50, no. 4, pp. 1114–1124, Jul. 2001. [4] I. Howitt and J. Gutierrez, “IEEE 802.15.4 low rate-wireless personal area network coexistence issues,” in Proc. Wireless Commun. Netw. Conf., Mar. 2003, vol. 3, pp. 1481–1486. [5] K. J. Myoung, S. Y. Shin, H. S. Park, and W. H. Kwon, “IEEE 802.11b performance analysis in the presence of IEEE 802.15.4 interference,” IEICE Trans. Commun., vol. E90-B, no. 1, pp. 176–179, Jan. 2007. [6] D. G. Yoon, S. Y. Shin, W. H. Kwon, and H. S. Park, “Packet error rate analysis of IEEE 802.11b under IEEE 802.15.4 interference,” in Proc. Veh. Technol. Conf., May 7–10, 2006, vol. 3, pp. 1186–1190. [7] S. Y. Shin, H. S. Park, S. C. Choi, and W. H. Kwon, “Packet error rate analysis of IEEE 802.15.4 under IEEE 802.11b interference,” in Proc. WWIC, 2005, pp. 279–288. [8] A. Sikora and V. F. Groza, “Coexistence of IEEE802.15.4 with other systems in the 2.4 GHz-ISM-band,” in Proc. IMTC, Ottawa, ON, May 16–19, 2005, pp. 1786–1791. Authorized licensed use limited to: Georgia Institute of Technology. Downloaded on February 21, 2009 at 00:43 from IEEE Xplore. Restrictions apply. ANGRISANI et al.: COEXISTENCE ISSUES BETWEEN IEEE 802.11b AND IEEE 802.15.4 WIRELESS NETWORKS [9] N. Golmie, R. E. Van Dyck, A. Soltanian, A. Tonnere, and O. Rebala, “Interference evaluation of bluetooth and IEEE802.11b systems,” Wirel. Netw., vol. 9, pp. 201–211, 2003. [10] L. Angrisani and M. Vadursi, “Cross-layer measurements for a comprehensive characterization of wireless networks in the presence of interference,” IEEE Trans. Instrum. Meas., vol. 56, no. 4, pp. 1148–1156, Aug. 2007. [11] L. Angrisani, M. Bertocco, D. Fortin, and A. Sona, “Assessing coexistence problems of IEEE 802.11b and IEEE 802.15.4 wireless sensor networks through cross-layer measurements,” in Proc. IMTC, Warsaw, Poland, May 1–3, 2007. [12] Part 11: Wireless LAN Medium Access Control (MAC) and Physical Layer (PHY) Specifications, ANSI/IEEE Std. 802.11, 1999. [13] [Online]. Available: http://grid.unina.it/software/ITG [14] [Online]. Available: http://www.dei.unipd.it/ricerca/gmee [15] M. Bertocco, G. Gamba, A. Sona, and S. Vitturi, “Performance measurements of CSMA/CA-based wireless sensor networks for industrial applications,” in Proc. IMTC, Warsaw, Poland, May 1–3, 2007. [16] S. Vitturi, I. Carreras, D. Miorandi, L. Schenato, and A. Sona, “Experimental evaluation of an industrial application layer protocol over wireless systems,” IEEE Trans. Ind. Informat., vol. 3, no. 4, pp. 275–288, Nov. 2007. Leopoldo Angrisani (M’08) was born in Nocera Superiore, SA, Italy, on April 16, 1969. He received the M.S. degree (cum laude) in electronic engineering from the University of Salerno, Salerno, Italy, in 1993 and the Ph.D. degree in electrical engineering from the University of Napoli Federico II, Napoli, Italy, in 1997. Since 2002, he has been an Associate Professor with the Dipartimento di Informatica e Sistemistica, Università degli Studi di Napoli Federico II. His research activity is focused on typical topics of electrical and electronic measurements and, in particular, on the definition and the implementation of new digital signal processing methods for the performance assessment of telecommunication systems and apparatuses; the definition and the implementation of new methods for communication network test and measurement; the definition and the implementation of new digital signal processing methods for the detection, measurement, and classification of nonstationary signals; design, realization, and metrological characterization of automatic measurement instruments based on digital signal processors; and the definition and the implementation of alternative methods for uncertainty evaluation in indirect measurements. 1523 Matteo Bertocco (M’92) was born in Padova, Italy, in 1962. He received the Laurea degree and the Ph.D. degree from the University of Padova, Padova, both in electronics engineering in 1987 and 1991, respectively. In 1994, he joined the Department of Electronics and Informatics, University of Padova, as a Researcher, where he became an Associate Professor with the Electronic Instrumentation and Measurement in 1998 and, since 2005, has been a Full Professor. He has taken part in a number of national and international research projects, including the European Union fourth and fifth framework programs. He is the author of over 140 scientific papers and textbooks for academic courses. His research interests include digital signal processing, estimation, virtual systems, automated instrumentation, and electromagnetic compatibility. Daniele Fortin (S’07) was born in Padova, Italy, in 1977. He received the Laurea degree in telecommunications engineering and the Ph.D. degree in information engineering, with a course in bioelectromagnetism and electromagnetic compatibility, from the University of Padova in 2004 and 2008, respectively. He is currently with the University of Padova. His research interests include wireless communications (DVB, WLANs, and WSNs), which are focused on measurement methodologies for the performance assessment of digital communication systems. Alessandro Sona (M’05) received the Laurea degree (summa cum laude) in electronics engineering from the University of Padova, Padova, Italy, in 1999 and the Ph.D. degree in electronic instrumentation from the University of Brescia, Brescia, Italy, in 2002. Since then, he has been with the Department of Information Engineering, University of Padova, where he is currently an Assistant Professor of instrumentation and measurement. His main interests are in the field of measurement uncertainty, electromagnetic compatibility, and wireless communication systems and networks. In particular, his research is focused on new measurement techniques and tools for the analysis of electromagnetic compatibility phenomena and on performance evaluation of digital communication systems in the presence of interference. He is currently responsible for several internal projects on electromagnetic compatibility, digital communications, and sensor networks. Authorized licensed use limited to: Georgia Institute of Technology. Downloaded on February 21, 2009 at 00:43 from IEEE Xplore. Restrictions apply.