Occupied spectrum Free spectrum 1000 900 800 700 600 500 400 TS DATA 300 ⊙-1Tx-160MHz 200 钟-2TX-80MH2z 100 -4Tx-40 MHz +—8Tx-20MHz 10 0 30 40 50 60 Number of aggregated packets Figure 4.Dynamic spectrum access with and without OFDMA:a)dynamic channel bonding and OFDMA;b)achievable down- Tink throughput using a channel width of 160 MHz and OFDMA.The RTS size is 120 +56.Nx bits,with Nux the number of OFDMA subchannels. protocol are introduced,and several uplink and backoff countdowns at the same time,uplink downlink MAC proposals are reviewed. MU-MIMO transmissions may be started by In the downlink,a challenge for IEEE the AP using a special RTS"packet containing 802.11ax-2019 is to reduce the channel sounding information about the STAs that can transmit overheads given the same explicit approach as in in parallel.Then the selected STAs will simply IEEE 802.11ac-2013 is considered.The overhead start transmitting at the same time and wait to caused by the explicit channel sounding protocol receive the corresponding ACKs.This approach implemented depends on the channel sounding allows all selected STAs to be synchronized,but rate and number of sounded STAs,which can requires knowledge of the STAs's buffer occu- result in an unacceptable overhead in scenari- pancy by the AP,which can be provided by the os with many STAs.Solutions to reduce such a same STAs as in previous transmissions.Figure large overhead,apart from replacing the current 5b shows the throughput achieved by the aP in channel sounding protocol with a more efficient two downlink MU-MIMO configurations(cases solution,will require the use of smart schedul- 16:4:4 and 16:16:1).The obtained throughput ers that consider the current traffic patterns and values are compared with the case where a single the quality of service (QoS)demands of users to spatial stream is transmitted to only one destina- decide when the channel state information (CSI) tion (case 1:1:1). has to be requested,and from which STAs. Massive MIMO:Massive MIMO refers to the To support MU-MIMO transmissions in the case where the aP has many more antennas than uplink,IEEE 802.11ax-2019 must also detail how STAs and uses them to create a nearly identical the uplink CSI from each STA is obtained by number of point-to-point links as the number the AP.introduce a mechanism to signal a group of active STAs [13].In addition to the cost of of STAs to simultaneously start a transmission APs,the extra processing complexity,and high- and include techniques to overcome channel cal- er energy consumption,other open challenges ibration and timing issues in order to efficiently for massive MIMO include obtaining the CSI decode all the simultaneously received packets information;WLANs may require switching to at the AP. an implicit channel feedback approach. In both cases,using the collected CSI,the AP Network MIMO:In coordinated WLAN has to select the specific STAs that will take part deployments,network MIMO can be used to in the next MU-MIMO transmission.Therefore, minimize the interference among simultane- the design of an efficient mechanism to create ous transmissions from different APs.The idea groups of STAs with low spatial channel correla- behind network MIMO is that different APs tion and similar channel quality is still an open can coordinate the transmissions as if they were challenge.Failing to properly create those groups a large array of antennas,which reduces the may prevent next-generation WLANs from fully inter-transmission interference and increases the benefiting from MU-MIMO technology. spatial reuse [14].However,effectively solving Figure 5a shows an AP with three STAs.On the tight synchronization requirements among the left,we have three downlink MU-MIMO the APs remains an open challenge. transmissions.The ones directed to STAs A.B and C contain four,two,and one SU-MIMO WLAN-LEVEL IMPROVEMENTS spatial streams,respectively.To start a downlink The user experience in next-generation WLANs transmission,the AP omnidirectionally sends the will not be simply enhanced by increasing the PHY header with information about the group of achievable network and user throughput as pre selected STAs and the number of spatial streams vious technical features do.To achieve that goal that are transmitted to each STA in SU-MIMO apart from the IEEE 802.11ax-2019 amendment, mode.On the right,we show an uplink MU-MI- there are other IEEE 802.11 amendments in prog- MO transmission.In IEEE 802.11ax-2019, ress.Figure 6 shows the most significant ones, because several STAs are unlikely to finish their including the new features they are targeting. IEEE Wireless Communications.February 2016 43IEEE Wireless Communications • February 2016 43 protocol are introduced, and several uplink and downlink MAC proposals are reviewed. In the downlink, a challenge for IEEE 802.11ax-2019 is to reduce the channel sounding overheads given the same explicit approach as in IEEE 802.11ac-2013 is considered. The overhead caused by the explicit channel sounding protocol implemented depends on the channel sounding rate and number of sounded STAs, which can result in an unacceptable overhead in scenari￾os with many STAs. Solutions to reduce such a large overhead, apart from replacing the current channel sounding protocol with a more efficient solution, will require the use of smart schedul￾ers that consider the current traffic patterns and the quality of service (QoS) demands of users to decide when the channel state information (CSI) has to be requested, and from which STAs. To support MU-MIMO transmissions in the uplink, IEEE 802.11ax-2019 must also detail how the uplink CSI from each STA is obtained by the AP, introduce a mechanism to signal a group of STAs to simultaneously start a transmission, and include techniques to overcome channel cal￾ibration and timing issues in order to efficiently decode all the simultaneously received packets at the AP. In both cases, using the collected CSI, the AP has to select the specific STAs that will take part in the next MU-MIMO transmission. Therefore, the design of an efficient mechanism to create groups of STAs with low spatial channel correla￾tion and similar channel quality is still an open challenge. Failing to properly create those groups may prevent next-generation WLANs from fully benefiting from MU-MIMO technology. Figure 5a shows an AP with three STAs. On the left, we have three downlink MU-MIMO transmissions. The ones directed to STAs A, B, and C contain four, two, and one SU-MIMO spatial streams, respectively. To start a downlink transmission, the AP omnidirectionally sends the PHY header with information about the group of selected STAs and the number of spatial streams that are transmitted to each STA in SU-MIMO mode. On the right, we show an uplink MU-MI￾MO transmission. In IEEE 802.11ax-2019, because several STAs are unlikely to finish their backoff countdowns at the same time, uplink MU-MIMO transmissions may be started by the AP using a special RTS packet containing information about the STAs that can transmit in parallel. Then the selected STAs will simply start transmitting at the same time and wait to receive the corresponding ACKs. This approach allows all selected STAs to be synchronized, but requires knowledge of the STAs’s buffer occu￾pancy by the AP, which can be provided by the same STAs as in previous transmissions. Figure 5b shows the throughput achieved by the AP in two downlink MU-MIMO configurations (cases 16:4:4 and 16:16:1). The obtained throughput values are compared with the case where a single spatial stream is transmitted to only one destina￾tion (case 1:1:1). Massive MIMO: Massive MIMO refers to the case where the AP has many more antennas than STAs and uses them to create a nearly identical number of point-to-point links as the number of active STAs [13]. In addition to the cost of APs, the extra processing complexity, and high￾er energy consumption, other open challenges for massive MIMO include obtaining the CSI information; WLANs may require switching to an implicit channel feedback approach. Network MIMO: In coordinated WLAN deployments, network MIMO can be used to minimize the interference among simultane￾ous transmissions from different APs. The idea behind network MIMO is that different APs can coordinate the transmissions as if they were a large array of antennas, which reduces the inter-transmission interference and increases the spatial reuse [14]. However, effectively solving the tight synchronization requirements among the APs remains an open challenge. WLAN-Level Improvements The user experience in next-generation WLANs will not be simply enhanced by increasing the achievable network and user throughput as pre￾vious technical features do. To achieve that goal, apart from the IEEE 802.11ax-2019 amendment, there are other IEEE 802.11 amendments in prog￾ress. Figure 6 shows the most significant ones, including the new features they are targeting. Figure 4. Dynamic spectrum access with and without OFDMA: a) dynamic channel bonding and OFDMA; b) achievable down￾link throughput using a channel width of 160 MHz and OFDMA. The RTS size is 120 + 56 · Ntx bits, with Ntx the number of OFDMA subchannels. t f Data ACK RTS CTS DATA ACK Occupied spectrum Free spectrum t t Band f f 0 10 20 30 40 50 60 0 100 200 300 400 500 600 700 800 900 1000 Number of aggregated packets AP throughput (Mb/s) 1Tx – 160 MHz 2Tx – 80 MHz 4Tx – 40 MHz 8Tx – 20 MHz