MASTaR: MAC Protocol for Access Points in Simultaneous Transmit and Receive Mode

Transcription

1 MASTaR: MAC Protocol for Access Points in Simultaneous Transmit and Receive Mode Seongwon Kim, Chan-Byoung Chae, and Sunghyun Choi Department of Electrical and Computer Engineering and INMC, Seoul National University, Korea School of Integrated Technology, Yonsei University, Korea Abstract In-band simultaneous transmit and receive (STR) is expected to increase the efficiency of wireless local area network (WLAN). A major obstacle to implementing STR capability, however, has been insufficient suppression of self-interference. In addition, tackling several problems will require modifications to the current IEEE medium access control (MAC) protocol. In this paper, we propose MASTaR, a novel MAC protocol supporting STR in WLAN, backward compatible with legacy The performance of MASTaR is extensively evaluated via ns-3 simulation. The simulation results demonstrate that significant performance enhancement can be achieved in the current WLANs by an access point (AP) with STR capability and a properly designed MAC protocol. Index Terms IEEE WLAN, medium access control (MAC) protocol, in-band simultaneous transmit and receive (STR), full duplex, scheduling. I. INTRODUCTION The ability of a node to successfully receive signals from another node while transmitting signals to that same node or to other nodes is referred to as in-band simultanous transmit and receive (STR), a.k.a. in-band full duplex. In-band STR is considered a promising technology to increase network capacity in wireless communication systems including IEEE wireless local area network (WLAN) [1]. The main challenge, however, for realizing in-band STR in practice has been self-interference (SI). For a successful reception during transmission, a node should be capable of cancelling or suppressing the SI to the level of noise floor. This capability is known as self-interference cancellation (SIC). Without perfect SIC, the achievable capacity of in-band STR is restricted to less than the full-duplex limit. How SIC may be brought to real radios has been studied in [2 5]. According to the nodes engaged in a STR transmission, the transmission is classified into a symmetric mode and an asymmetric mode. In the symmetric mode, two nodes transmit and receive signals to/from each other simultaneously as shown in Fig. 1(a). On the other hand, in the asymmetric mode, a node transmits and receives to/from different nodes as shown in Fig. 1(b). The node that initiates a transmission primary transmission is called the primary transmitter (PTX); and the corresponding receiver is called the primary receiver (PRX). Similarly, the transmitter and receiver of a subsequent transmission the secondary transmission are called the secondary transmitter (STX) and secondary receiver (SRX). In Fig. 1(b), for instance, if transmits first and AP AP (a) Symmetric mode PTX PRX/STX AP (b) Asymmetric mode Fig. 1. Operational modes of STR in WLANs. SRX subsequently transmits, {PTX, PRX, STX, SRX}={, AP, AP, }. In IEEE WLAN, the asymmetric mode is more feasible because STR capability is not easily implemented in mobile stations, especially small devices such as smartphones. What is more likely, however, to become capable of STR are access points (APs). APs are relatively large and exhibit high performance. Our consideration in this work, therefore, is limited to the asymmetric mode for better feasibility and practicality. On the other hand, to support STR in IEEE WLANs there is needed a modification to medium access control (MAC) protocol. To bring STR to the current IEEE MAC protocol, the following problems should be addressed: 1) With carrier sense multiple access with collision avoidance (CSMA/CA) based channel access, a node that can simultaneously transmit on a busy channel cannot access the channel since its channel status is indicated as BUSY. 2) The current Acknowledgement (ACK) mechanism faces a new problem. Assuming that, in Fig. 1(b), finishes its transmission earlier than AP, ACK timeout happens because AP cannot send ACK due to its ongoing data transmission to. 3) If PTX causes significant interference to SRX in an asymmetric mode, the secondary transmission fails due to low received signal-to-interference-plus-noise ratio (SINR). There is also a fairness issue on the inter-frame spacing (IFS) setting and a backward compatibility issue on network allocating vector (NAV) setting as discussed in [6]. Accordingly, many studies have developed a new WLAN MAC protocol. Proposed in [7] is a centralized MAC protocol scheduling both UL and DL transmissions based on conflict map. Due to the nature of centralized protocol, considerable modifications must be made to the current behavior of WLANs, which generally operate in a distributed manner. Distributed approaches focusing on symmetric modes are proposed in [8 13]. The proposed protocols in [8, 9] enable /16/$ IEEE

2 Decision to adapt transmit (Tx) power block Block Ack Exchange request frame Element ID Length Link margin calculation report frame Element ID Length Tx power Link margin Ack policy = Block Ack S I F S Protocol AP Fig. 2. IEEE h procedure. STR in WLAN based on collision avoidance (by utilizing RTS/CTS) and collision detection (by embedding pseudo noise sequence in preamble), respectively. Proposed in [10 13] are MAC protocols that consider the asymmetric modes. Modified RTS/CTS [10, 11], orthogonal frequency division multiple access (OFDMA)-based on busy tone [12], and power-controlled pseudo noise signature [13] are used to arrange simultaneous transmissions. Those approaches, however, are not compatible with legacy devices. In this paper, we propose a novel MAC protocol for APs supporting STR in WLANs, named MASTaR (MAC protocol for Access points in Simultaneous Transmit and Receive mode). MASTaR solves the aforementioned problems by enabling (i) access to a busy channel after identifying PTX, (ii) bit padding and explicit block ACK, and (iii) transmission scheduling based on link map. The three key components solve the first, second, and third problems mentioned above, respectively. As designed based on the current IEEE standard, moreover, MASTaR s main advantage is compatibility with legacy devices. To the best of our knowledge, this is the first standard-compliant STR MAC protocol in IEEE WLANs, supporting asymmetric mode with legacy STAs. The rest of the paper is organized as follows. In Section II, we introduce IEEE functions utilized in the proposed protocol. In Section III, we present detailed operation of MASTaR. Its performance is evaluated in Section IV. Section V concludes the paper. II. PRELIMINARIES We first introduce preliminary IEEE functions and RF capability, which are utilized in the proposed protocol. A. IEEE h Transmit Power Control () IEEE h is the amendment for spectrum and transmit power management extensions. Within the amendment,, which was originally designed to satisfy maximum transmit power regulation, provides a procedure for a client station (STA) to notify its link margin to the associated AP. As shown in Fig. 2, an AP triggers the procedure by sending a request frame to its STA. Upon the request reception, the STA sends a report frame conveying the STA s link margin information. According to the standard, a STA s link margin is defined as the ratio of the received STA Fig. 3. Explicit block ACK procedure. signal strength of the corresponding request frame to the minimum desired signal strength by the STA. The specific algorithm for computing the link margin is implementation dependent. For example, link margin can be computed as follows. First, with the presence of interference, link margin can be redefined as the ratio of SINR during the reception of a request frame to the desired SINR. Secondly, assuming STAs are aware of their radio frequency (RF) parameters such as receiver sensitivities and noise floor, STA i can estimate desired SINR using its RF parameters as SINR i,m = RxSens i,m NF i, (1) where SINR i,m, RxSens i,m, and NF i are STA i s desired SINR for modulation and coding scheme (MCS) m, Rx sensitivity for MCS m, and noise floor, respectively. All the above values are on the db scale. Then, STA i s link margin with the presence of the interference from STA j is calculated as λ j i,m (Received SINR) i SINR i,m, (2) We assume that the received SINR is measured by comparing the received power before and after the frame detection. B. Explicit Block ACK In baseline IEEE ACK policy, every single unicast frame should be individually acknowledged. IEEE e defines block ACK () policy which allows several data frames to be transmitted before an ACK is returned. With the explicit policy, after sending a data block, an originator sends a block ACK request (R) frame, as shown in Fig. 3. The recipient then responds to the R frame with a frame, which indicates whether each individual data frame is successfully received or not. In this way, explicit policy makes it possible for the transmitter to control when the receiver should send ACK after receiving data frames. C. Second Frame Capture In wireless networks, if more than one frame is being transmitted in the same channel, a node can receive only the strongest frame depending on the relative signal power and the arrival timings of the colliding frames. If the strongest frame arrives at the receiver later than the other frames, the successful reception of the later-arriving frame is referred to as second frame capture. It has been found that approximately 10 db signal-to-interference ratio (SIR) is required for second frame

3 Identify PTX by MAC hdr. 2 nd frame capture Request Broadcast D I F S + Backoff (a) PTX identification by MAC header Identify PTX by RTS RTS CTS 1 st frame capture D I F S + Backoff ACK policy field set to Normal ACK Report Pad (c) Broadcasting request Request Unicast Report (e) Simultaneous (implicit) Pad ACK policy field set to Block ACK R (b) PTX identification by RTS (d) Unicasting request (f) Explicit by R Fig. 4. Basic operation of MASTaR. capture [14], and commercial WLAN devices generally have the second capture capability [15]. Therefore, a later-arriving frame can be decoded if its signal strength is higher than that of an earlier arriving frame by 10 db or more. III. MASTaR: PROPOSED MAC PROTOCOL In this section, we propose a novel MAC protocol, namely MASTaR. We consider only downlink (DL) secondary transmission, i.e., AP starts a secondary transmission after identifying the transmitter of uplink (UL) transmission. 1 We assume that STR capability is implemented in only APs, and STAs are the legacy devices. 2 Based on the functions described in Section II, MASTaR consists of the following components. A. PTX Identification To initiate a secondary transmission, the STX, i.e., AP, should identify that the PTX is one of its client STAs. We propose two PTX identification methods, each for primary transmission initiated with/without RTS, respectively. For data transmission without RTS, AP peeks the MAC header of the frame on the primary transmission as shown in Fig. 4(a). 3 In a strict sense, since bit errors can occur, AP is not convinced of the PTX before checking cyclic redundancy check (CRC). However, the probability that the erroneous MAC address becomes the same as one of client STAs addresses is quite low. If the source address of primary transmission s MAC header does not agree with any of AP s client STAs, PTX identification fails and secondary transmission is not initiated thus preventing possible erroneous operations. If 1 If we modify STA devices, UL secondary transmission is also possible. 2 MASTAR can also consider the STAs with STR capability, by prioritizing the symmetric mode transmissions when it schedules secondary transmission. In the followings, however, we focus on the asymmetric mode. 3 According to IEEE ac [1], VHT-SIG-A field in PLCP header for DL frames contains partial AID which indicates individual client STA. However, for uplink frames, only basic service set ID (BSSID) is contained in VHT-SIG- A field, thus the transmitter of an uplink transmission cannot be identified by PLCP header. If the new IEEE standard defines a new PLCP header conveying transmitter identifier as suggested in [16], AP can identify PTX before decoding MAC header. the source address corresponds to one of AP s client STAs, AP initiates a secondary transmission. 4 If PTX sends RTS before its data transmission, AP can clearly identify PTX before the primary transmission, as shown in Fig. 4(b). In this case, AP schedules a secondary transmission right after CTS in order to make SRX better receive the secondary transmission by first frame capture. 5 B. Training: Link Map Management To choose the best SRX for a given PTX, AP manages link map. For each pair of STAs i and j, link map is defined as L(j, i) : (j, i) (λ j i,0, λj i,1,, λj i,m max ), i, j I, (3) where I is the set of all client STAs and m max is the maximum MCS, e.g., 7 for IEEE n with a single spatial stream. To build a link map, AP sends a request frame after the PTX identification using reference MCS, i.e., m ref. 6 Initially, the request frame is broadcast as shown in Fig. 4(c) since L(PTX, i) is empty for i. At an STA, if the signal strength of the request frame is stronger enough than the frame from PTX, the STA can receive the request frame, calculate the link margin according to the method in Section II-A, and report it via a report frame. If AP receives the report frame before the report timeout, AP estimates link margins for other MCSs and populates the link map. Since the signal power changes due to multi-path fading and the STA s mobility, link map is adapted, depending on the channel condition, as follows. Let the time when AP sends the latest request to STA i during the reception from PTX j be τ j i. AP records the signal power of PTX j at that time as S j (τ j i ), and the signal power of the corresponding 2 Since AP starts transmission during reception in this case, its self interference after only analog cancellation can saturate its automatic gain control (AGC). To prevent this, STR-enabled AP needs to conservatively set the AGC in advance so that the self interference power is within the dynamic range of its analog-to-digital converter (ADC). 5 Strictly speaking, in this case, AP s transmission is not secondary transmission since it precedes the UL transmission. 6 The required SINR of reference MCS should be low so that more STAs can receive the request frame. However, it need not to be lower than the required SIR for second frame capture, which is about 10 db.

4 report frame from STA i as S i (τ j i ). AP also keeps track of subsequent received signal power from all its client STAs, and calculates S i,srx = S i S i (τ j i ) and S j,ptx = S j S j (τ j i ), (4) where S i and S j are the recently measured signal power from STAs i and j, respectively. Then link map is updated as follows: 1) If AP i senses a primary transmission from j and the signal power satisfies S i,srx + S j,ptx thres, it unicasts the request to STA i as shown in Fig. 4(d) to update L(j, i), where thres is the update threshold on the db scale and is set to 5 db for the simulations. 2) If the value is less than the update threshold, AP i updates λ j i,m as λj i,m λj i,m + S i,srx without unicasting the request. The reason why only S i,srx is used for the link margin update in the second case is that if S i changes by S i,srx, S AP i also changes by S i,srx assuming channel reciprocity and no transmit power control at STA i. S j,ptx, on the other hand, cannot be utilized to estimate I j i, since the channel between AP and STA j and the channel between STA i and STA j are not always correlated. C. Secondary Transmission After populating the link map, AP initiates, based on the link map, secondary transmissions based on link map. Various scheduling and rate selection algorithms can be adopted for secondary transmission depending on the purpose, e.g., throughput maximization or achieving fairness, In this work, we present a simple scheduling and rate selection algorithm for reliable secondary transmission, while trying to achieve high throughput gain at the same time. Specifically, MASTaR schedules secondary transmission to an STA the link of which is the strongest with the interference from the current PTX. For a given PTX j, AP finds the best SRX w.r.t. the PTX, i.e., i = argmax i I,i j λ j i,m ref. (5) Used for the secondary transmission is the highest order of MCS satisfying λ j i,m 0. AP then aggregates multiple MAC protocol data units (MPDUs) destined to i so that the aggregated MPDU (A-MPDU) can be transmitted within the remaining duration of the primary transmission. Next, AP sends the A-MPDU with bit padding to make the secondary transmission end within the SIFS time from when the primary transmission ends. This makes the from AP to PTX be transmitted earlier than the from SRX to AP, thus enabling the first frame capture of the frame at PTX. Finally, if PTX can receive AP s with the simultaneous SRX s transmission after PTX s data transmission, i.e., λ i j 0 for data rate, AP sets ACK policy of the A- MPDU to Normal ACK, so that, as shown in Fig 4(e), both transmissions of the responding primary and secondary transmissions overlap. Otherwise, AP sets ACK policy of the A-MPDU to Block ACK, indicating an explicit mechanism TABLE 1: Simulation parameters. Parameter Description Packet size 1460-byte payload MPDU CW size CWmin = 15, CWmax = 1023 A-MPDU bound bytes, 10 ms Pathloss model TGax pathloss model [17] Fading model Jakes model (Doppler velocity= 0.1 m/s) Antenna gain 0 dbi (AP), 8.2 dbi (STA) 7 Transmit power 20 dbm (AP), 15 dbm (STA) CCA threshold 82 dbm Rate adaptation Minstrel using n MCSs as shown in Fig. 4(f). If AP is not responded to by from SRX, or the from SRX indicates that most MPDUs are not successfully received, AP unicasts the request to the SRX at the next primary transmission from the PTX to update the link map. It should be noted that, in MASTaR, STR happens when a STA wins the channel, thus starting a UL transmission. If, on the other hand, AP wins the channel, STR is not possible. Therefore, if AP uses a longer contention window (CW) than do the STAs, the probability that a STA wins the channel increases and as a result more STR opportunities are acquired. The effect of CW size on AP is studied in the following section. IV. PERFORMANCE EVALUATION In this section, we present the performance evaluation results via ns-3 simulation. For reliable simulation results, we elaborately implement STR and second frame capture capabilities in the simulator. The simulation consider a single cell environment with randomly distributed STAs within 10 m of the AP. DL and UL throughput was measured for 1 second by varying the number of STAs (N), the ratio of DL traffic source rate to total traffic source rate (DL traffic ratio), how much AP cancels SI, i.e., SIC performance (SIC), and AP s minimum size of CW (CWmin). Except for the simulation with various DL traffic ratios, each STA has both DL/UL fullybacklogged UDP traffic. Because AP cannot force STAs to use or not to use RTS, we carry out simulations for the following two scenarios: (i) no RTS scenario where STAs do not use RTS and (ii) an RTS scenario where STAs uses RTS for A- MPDU. All results are averaged out for 100 iterations and the detailed simulation parameters are described in Table 1. A. Comparison Protocols We compare MASTaR with the following three protocols: 1) HD: Current half duplex-based MAC protocol. 2) BusyTone: Transmitting busy tone during secondary transmission for hidden terminal resolution as proposed in [2]. 3) RTS/FCTS: Three-way handshake before primary transmission as proposed in [10]. Since it is based on STAs RTS transmission, this is compared only in the RTS scenario. 7 We consider hand-grip loss of mobile STAs as studied in [15].

5 (a) No RTS scenario (b) RTS scenario Fig. 5. Performance depending on N : DL traffic ratio = 0.5, SIC = 110 db, and AP s CWmin = 15. (a) No RTS scenario (b) RTS scenario Fig. 6. Performance depending on DL traffic ratio: N = 5, SIC = 110 db, AP s CWmin = 15, and source rate = 20 Mb/s per STA. (a) No RTS scenario (b) RTS scenario Fig. 7. Performance depending on SIC: N = 5, DL traffic ratio = 0.5, and AP s CWmin = 15. B. Results with Varying Number of STAs (Fig. 5) In WLANs, collisions can occur either when a hidden terminal cannot sense an ongoing transmission or when backoff counters of more than one STA reach zero simultaneously. We call the former collision a hidden collision and the latter a contention collision. Without using RTS, as shown in Fig. 5(a), throughput rapidly decreases as N increases because of the rise in both hidden and contention collisions increase. BusyTone slightly improves throughput, but it does not fully resolve hidden collisions because AP cannot transmit busy tone before identifying PTX. In addition, more UL transmissions fail in the BusyTone protocol owing to residual SI, i.e., SI after SIC. Since the rate adaptation algorithm works, accordingly, STAs are more likely to use a robust data rate, i.e., low data rate, thus occupying the medium for longer during each transmission attempt. This results in lower DL throughput of BusyTone. Despite residual SI and additional MAC overhead, on the other hand, MASTaR achieves much higher throughput in DL thanks to the secondary transmissions. MASTaR also yields more transmission opportunities to STAs because AP resets its backoff counter after secondary transmission. Therefore, using MASTaR also slightly enhances UL throughput. In the other results shown below, MASTaR achieves, for the same reason, much higher throughput than other protocols. Using RTS, the overall throughput increases and becomes less dependent on N since hidden collisions are resolved by CTS, as shown in Fig. 5(b). Still, lower total throughput is achieved with larger N because each STA has fewer transmission opportunities. Thus the rate adaptation algorithm works less optimally. Meanwhile, RTS/FCTS enhances DL throughput by simultaneous transmissions after exchanging one RTS and two consecutive FCTS frames. Since RTS/FCTS does not consider the interference between PTX and SRX, however, secondary transmission fails frequently due to the strong interference between PTX and SRX. Owing to the failures, as well as the additional overhead of three-way handshaking for every UL transmission, RTS/FCTS achieves lower throughput gain than MASTaR. C. Results with Varying DL Traffic Ratio (Fig. 6) Figure 6 shows the aggregate throughput depending on DL traffic ratio. In this case, the source rate is 20 Mb/s per STA, and the number of STAs is five. Therefore, the total source rate is fixed at 100 Mb/s and each link s source rate is given

6 Fig. 8. Performance depending on AP s CWmin: N = 5, DL traffic ratio = 0.5, and SIC = 110 db. by the product of the traffic ratio and the total source rate, e.g., 20 Mb/s and 80 Mb/s for DL and UL, respectively when DL traffic ratio is 20%. For HD and BusyTone, as shown in both Figs. 6 (a) and (b), the ratio of the achieved DL throughput to the total throughput is much smaller, except for when DL traffic ratio = 20%, than the ratio of DL source rate to the total source rate. This is because AP cannot win the channel as often as it needs to if multiple STAs with UL traffic coexist. On the other hand, MASTaR delivers DL traffic more successfully thanks to the secondary transmission during uplink channel access. The portion of MASTaR s DL throughput, therefore, increases proportionally with the given DL traffic ratio. It is also worth mentioning that the throughput gain of MASTaR over HD stands out when DL traffic ratio is equal or greater than 50%, that is, when DL traffic is more congested than UL traffic. D. Results with Varying SIC Performance (Fig. 7) If SIC performance is low, primary UL transmissions fail due to the strong residual SI. When SIC = 80 db, for instance, the residual SI power is about 60 dbm, which is comparable with the received power of the intended signals. The failures cause retransmissions and make the rate adaptation algorithm choose low MCS for UL transmission. Therefore, UL throughput of the STR protocols, i.e., BusyTone, RTS/FCTS, and MASTaR, is worse than that of HD if SIC is low, as shown in Fig. 7. As SIC increases, UL throughput of STR protocols increases, and becomes greater than HD s UL throughput when SIC = 110 db. On the other hand, DL throughput of MAS- TaR depends less on the SIC level. The results demonstrate that if SIC of about 100 db or more is possible, significant performance enhancement for both UL and DL is possible in WLANs with a properly designed MAC protocol. E. Results with Varying CW Size of AP (Fig. 8) Figure 8 shows that the MASTaR s performance further increases if AP uses a larger CW size when N = 5 with fully backlogged traffic. Throughput increases not only in UL by more UL channel access, but also in DL thanks to effective channel utilization with more secondary transmissions. When the channel is heavily occupied by STAs, therefore, AP can adjust its CWmin depending on the the viability of secondary transmissions. For example, if the average link margin in the link map is high, then larger CWmin can be used. V. CONCLUDING REMARKS This paper has introduced MASTaR, a novel MAC protocol for IEEE WLAN. Since MASTaR was designed based on the existing functions of the current standards, it is both standard compliant and backward compatible. The evaluation results from ns-3 simulation under the various conditions confirmed that the proposed protocol was able to achieve up to 1.75 higher throughput when STAs did not use RTS, and up to 1.5 higher throughput when STAs used RTS. The results underscored the notion that a noteworthy performance enhancement may be expected in WLAN with the APs simultaneously transmitting and receiving. Our future work will include transmit power control and adaptive carrier sensing considering multi-cell environments. We also plan to carry out trace-based simulation using the SIC trace obtained from full duplex testbed. ACKNOWLEDGMENT This work was supported by ICT R&D program of MSIP/IITP.B ,Multiple Access Technique with Ultra-Low Latency and High Efficiency for Tactile Internet Services in IoT Environments REFERENCES [1] IEEE ac-2013, Part 11: Wireless LAN Medium Access Control (MAC) and Physical Layer (PHY) specifications. Amendment 4: Enhancements for Very High Throughput for Operation in Bands below 6 GHz, IEEE std., Dec [2] M. Jain et al., Practical, real-time, full duplex wireless, in Proc. ACM MobiCom, Sep [3] E. Aryafar, M. A. Khojastepour, K. Sundaresan, S. Rangarajan, and M. Chiang, MIDU: Enabling MIMO full duplex, in Proc. ACM MobiCom, Aug [4] D. Bharadia, E. McMilin, and S. Katti, Full duplex radios, in Proc. ACM SIGCOMM, Aug [5] M. Chung, M. Sim, J. Kim, D.-K. Kim, and C.-B. Chae, Prototyping real-time full duplex radios, IEEE Commun. Mag., vol. 53, no. 9, pp , Sep [6] Z. Zhang, K. Long, A. V. Vasilakos, and L. Hanzo, Full-duplex wireless communications: Challenges, solutions and future research directions, in Proc. Asilomar, Nov [7] J. Y. Kim et al., Janus: A novel MAC protocol for full duplex radio, Stanford University, Tech. Rep., [8] M. Duarte et al., Design and characterization of a full-duplex multiantenna system for WiFi networks, IEEE Trans. Veh. Technol., vol. 63, no. 3, pp , Mar [9] X. Xie and X. Zhang, Concise paper: Semi-synchronous channel access for full duplex wireless networks, in Proc. IEEE ICNP, Oct [10] W. Cheng, X. Zhang, and H. Zhang, RTS/FCTS mechanism based fullduplex MAC protocol for wireless networks, in Proc. IEEE GLOBE- COM, Dec [11] A. Tang and X. Wang, A-Duplex: Medium access control for efficient coexistence between full duplex and half duplex communications, IEEE Trans. Wireless Commun., vol. 14, no. 10, pp , [12] S. Goyal, P. Liu, O. Gurbuz, E. Erkip, and S. Panwar, A distributed MAC protocol for full duplex radio, in Proc. Asilomar, Nov [13] W. Zhou, K. Srinivasan, and P. Sinha, RCTC: Rapid concurrent transmission coordination in full DuplexWireless networks, in Proc. IEEE ICNP, Oct [14] J. Lee et al., An experimental study on the capture effect in a networks, in Proc. ACM WiNTECH, Sep [15] S. Yoo, Y. Shin, S. Kim, and S. Choi, Toward realistic WiFi simulation with smartphone Physics, in Proc. IEEE WoWMoM, June [16] S. Yoo et al., ProCCA: Protective clear channel assessment in IEEE WLANs, IEEE Commun. Lett., vol. PP, no. 99, [17] IEEE P802.11, TGax Simulation Scenarios, July 2015.