1 Introduction

Wireless traffic demands have been continuously increasing. Ericsson’s latest mobility report shows that from 2020 to 2024, the mobile traffic will grow by 30 percent compound annual growth rate (CAGR) [1]. Wireless local area network (WLAN) is one of the most important and popular wireless networks. Cisco’s statistics shows that the traffic carried by WiFi was accounted for 43 percent of global IP traffic in 2017, and the ratio will rise to 51 percent by 2022 [2]. Thus, both industry and academia focus on the evolving technical features and standardization of WLAN. IEEE 802.11ax or named high-efficient WLAN (HEW), considered as WiFi 6, will be released in 2020 [3,4,5,6].

Improving maximum throughput is the most important evolving goal of wireless networks including cellular network and WLAN for a long time. Especially in recent years, the throughput requirements of wireless services are becoming increasingly stringent. For example, the emergence of 4K/8K video, virtual reality (VR), augmented reality (AR), mobile cloud computing, and other services in recent years poses great challenges to the throughput of WLAN. Therefore, before the formal release of IEEE 802.11ax, industry and academia immediately regard extremely high throughput as the core technical objective of the new generation WLAN. In May 2019, the IEEE 802.11be Task Group (TGbe) was established [7]. The IEEE 802.11be is named as extremely high throughput (EHT) WLAN, probably considered as WiFi 7.

To achieve the technical objective, IEEE 802.11be introduces a series of key technologies. We classify them as five classes: physical layer (PHY) enhancements, multi-band operation, multi-AP coordination, enhanced link reliability, and latency & jitter guarantee [8]. Physical layer enhancements include larger bandwidth such as 320MHz and 240MHz, multiple input multiple output (MIMO) enhancement such as 16 spatial streams, new tone-plan, new preamble puncture method, and etc. Multi-band operation enables the access points (APs) and non-AP stations (STAs for short in the sequel) to simultaneously work in multiple bands in order to fully use the frequency resources. Multi-AP coordination enables the cooperation among APs so that the “win-win” situation can be achieved. Enhanced link reliability tries to further improve the link adaptation and to introduce some new technologies such as the hybrid automatic repeat quest (HARQ). Latency & jitter guarantee means IEEE 802.11be aims to improve the worst case latency feature to satisfy the real-time applications such as online game and VR.

Since the TGbe has just been established, the research of these key technologies is still in the preliminary stage of exploration. Therefore, this article investigates and analyzes the key technologies of IEEE 802.11be, and further provides our perspectives and insights on it. Specifically, this article gives a brief overview on IEEE 802.11be, including the target scenario and technical objective, key technologies overview, and the standardization process. After that, we further investigate, analyze and provide perspectives on the key technologies of IEEE 802.11be including multi-band operation, multi-AP coordination, enhanced link reliability, and latency & jitter guarantee in detail. The investigation and analysis contain the progress of standardization, academia solutions, and our perspectives. To the best of our knowledge, this is the first work to fully investigate, analyze and provide insights on IEEE 802.11be.

The structure of this article is summarized as follows. Section 2 provides a brief overview of IEEE 802.11be. Sections 3–6 investigate and analyze multi-band operation, multi-AP coordination, enhanced link reliability, and latency & jitter guarantee, respectively. Finally, Section 7 concludes this article.

2 Brief overview of IEEE 802.11be

2.1 Target scenario and technical objective

Video traffic is the dominant traffic type of Internet and mobile Internet at present and in the near future. Video traffic accounted for 60 percent of all mobile traffic in 2018, and this proportion is expected to reach 74 percent in 2024 [1]. Especially in recent years, the demands for ultra-high definition video such as 4K and 8K have become increasingly intense, thus it can be predicted that ultra-high definition video traffic will become the killer scenario of the next generation wireless network [9, 10].

Furthermore, real-time application has emerged in recent years and attracts increasing attention [11]. For example, VR, AR, wireless video meeting, online games and mobile cloud computing are typical cases of real-time applications. Especially, some applications like VR and AR require both high speed and low latency. These kinds of services require the network not only to provide extremely high throughput, but also to guarantee low latency (especially worst case latency and deterministic latency) and jitter. This puts forward very stringent requirements and challenges for wireless networks.

Therefore, the ultra-high definition video traffic and real-time traffic are the target scenarios of IEEE 802.11be. To efficiently support the target scenarios, the key technical objective of IEEE 802.11be is to achieve extremely high throughput (at least 30Gbps) and to improve the worst case latency and jitter.

2.2 Key technologies of IEEE 802.11be

IEEE 802.11be focuses on a series of technical enhancements of PHY and media access control (MAC) layer, including PHY enhancements, multi-band operation, multi-AP coordination, enhanced link reliability, and latency & jitter guarantee.

2.2.1 PHY enhancements

The enhancement of PHY technology is the most direct means to improve throughput. Among them, the expansion of resource domain capability and the improvement of modulation mode are the main methods. For frequency domain, IEEE 802.11be intends to extend the maximum transmission bandwidth from 160MHz to 320MHz (240MHz is also an optional) [12,13,14]. Besides, we should rethink and re-design the tone-plan [15,16,17] and the preamble puncture scheme [18,19,20]. For spatial domain, IEEE 802.11be intends to increase the maximum number of spatial streams from 8 to 16 [21, 22]. Then, adopting multiple resource units (RUs) per STA is one important enhancement for IEEE 802.11be since only one RU can be used for each STA in IEEE 802.11ax. Moreover, the PHY protocol data unit (PPDU) format should be re-designed to support a series of new capabilities. For example, in recent months, the TGbe passed a motion to introduce universal signal field (U-SIG)—containing version-independent signal field (SIG) and version-dependent signal field (SIG)—and EHT SIG, which is quite different to the IEEE 802.11ax and older versions. Finally, multi-band operation, HARQ and other MAC layer protocols also require to improve the PHY technologies.

In addition to the above mentioned candidate PHY technologies discussed in TGbe, we believe that the following PHY technologies are also possible candidates. In terms of modulation mode, we consider higher order modulation such as 4096 quadrature amplitude modulation (4096-QAM) as a possible candidate technology. In the early 2020, TGbe also pays attention to 4096-QAM. Moreover, full-duplex communication technology can significantly improve throughput, thus we believe that full-duplex communication can also help achieve the goal of extremely high throughput.

It is worth noting that considering that most of the PHY enhancements of IEEE 802.11be belong to engineering technology, and in addition, due to the limited space, the following sections will not analyze the physical layer technology in detail. Readers can track TGbe’s proposals on PHY to get the latest progress.

2.2.2 Multi-band operation

The IEEE 802.11be operates between 1 and 7.250 GHz. Specifically, IEEE 802.11be supports three bands: 2.4GHz, 5GHz and 6GHz. More importantly, IEEE 802.11be intends to introduce multi-band operation [23,24,25,26]. The core idea is that either AP or STAs can work in multiple bands at the same time. Efficient operations among multiple bands are introduced. Therefore, AP and STAs can use multiple band resources efficiently and dynamically, thereby achieving the goal of improving throughput and guaranteeing delay. Multi-link operation is the most important features of multi-band Operation that will be detailed in Section 3.

2.2.3 Multi-AP coordination

In order to improve area throughput, IEEE 802.11ax introduces spatial reuse technology. However, due to the lack of coordination between APs, the interference in multi-AP scenarios becomes more serious and uncontrollable. Thus, IEEE 802.11be intends to introduce multi-AP coordination technology by enabling multiple APs to exchange information, coordinate and serve STAs consistently, so as to optimize network performance from a more global perspective [27,28,29,30,31,32]. Section 4 investigates and analyzes multi-AP coordination in detail.

2.2.4 Enhanced link reliability

Since WLAN works in unlicensed band and adopts carrier sense multiple access with collision avoidance (CSMA/CA) method based on random contention, the state and quality of communication links are constantly changing, and the reliability of links is difficult to be guaranteed. Traditional WLAN uses link adaptation technology to dynamically adjust transmission rate and ARQ technology to deal with transmission failure. In order to further take into account both transmission efficiency and link reliability, IEEE 802.11be intends to introduce HARQ technology [33,34,35,36,37]. It is worth noting that although HARQ has been adopted in cellular networks for years, it is the first time that HARQ has been introduced into WLAN. Section 5 investigates and analyzes enhanced link reliability in detail.

2.2.5 Latency & jitter guarantee

WLAN’s lack of support for delay and jitter has been criticized for a long time, which is also an important shortcoming restricting the development of WLAN. Therefore, IEEE 802.11be pays attention to the support of delay and jitter, and intends to introduce at least one mode to improve the performance of worst case delay and jitter [38,39,40,41,42]. Section 6 investigates and analyzes latency & jitter guarantee in detail.

2.3 Standardization roadmap of IEEE 802.11be

The standardization roadmap of IEEE 802.11be [43] is shown in Fig. 1. Before the formal establishment of the working group, the IEEE 802.11 Standards Committee first established the EHT Study Group (SG) in 2018. After thorough research and discussion of EHT SG, with the output of PAR files, the IEEE 802.11be Working Group (TGbe) was established in May 2019. Specification framework document (SFD) documentation is expected to start in 2020, and draft version 1.0 of IEEE 802.11be in 2021 will be completed. After that, it is expected that draft 3.0 and 5.0 will be completed in 2022 and 2023, respectively. Finally, it is planned to officially release IEEE 802.11be in May 2024. It can be seen that the current research of IEEE 802.11be and its key technologies is in progress.

Fig. 1
figure 1

Standardization Roadmap of IEEE 802.11be

Full size image

In early 2020, TGbe reaches a consensus that the standardization process will be divided into two phases, thereby publishing two releases. The technologies in Phase 1 include multi-link operation, wider band such as 320MHz and 240MHz, 4096-QAM, multiple RUs per STA, and some easy designed AP coordination scheme, while other features will be studied in Phase 2.

Table 1 shows the key technologies and their phases.

Table 1 Key technologies of IEEE 802.11be
Full size table

3 Multi-band operation

Table 1 summarizes the different types of AP coordination.

3.1 Standards progress introduction

IEEE 802.11be operates in several bands, including 2.4 GHz, 5 GHz and the newly introduced 6 GHz. The APs and STAs of IEEE 802.11be can install multiple radios, thus enabling APs and STAs to work simultaneously on multiple bands. In fact, only one radio can also support multi-band capability, as long as the interference between adjacent bands can be eliminated sufficiently well, which is out of the space of this article. It is worth noting that it is not new to allow multiple bands to work at the same time. Legacy IEEE 802.11 has also been supported. However, the proposed multi-band capability of IEEE 802.11 is different from the multi-band capability supported by legacy IEEE 802.11.

3.1.1 Flow-level and packet-level multi-link aggregation

First of all, it is necessary to introduce the concept of “link”, which refers to the communication link that can send and receive frames. Often a link corresponds to a radio frequency (RF) link. Different links can work on one band or on multiple bands, thus the concept of multi-link is a little broader than multi-band. Without special explanation, the following text will no longer distinguish between multi-band and multi-link.

IEEE 802.11be allows a logical multi-link (ML) device (MLD) to consist of a series of physical entities. As shown in the Fig. 2, a logical AP MLD consists of multiple APs, and the logical non-AP MLD (STA MLD for short in this paper) consists of multiple non-AP STAs. Different from legacy IEEE 802.11, the ML function of IEEE 802.11be proposes that only one entity connecting to the upper layer (such as the network layer), while the ML function of legacy IEEE 802.11 has several independent entities connecting to the upper layer. In other words, legacy IEEE 802.11’s ML function is equivalent to several relatively independent APs or STAs encapsulated in one device.

Fig. 2
figure 2

Multi-band/multi-link aggregation of IEEE 802.11be

Full size image

As shown in Fig. 2, flow-level multi-link operation means that an AP or STA can allocate its multiple traffic streams to different links, but flow-level multi-link operation does not allow data frames from one traffic stream to be migrated to other links unless a complex processes named fast session transfer (FST) is adopted. Packet-level multi-link operation refers to that an AP or STA can dynamically transfer any of its frames to different links. This means that data frames belonging to the same traffic stream can be flexibly hosted on different links. Even as shown in the Fig. 2, if the transmission of the one frame (frame 4) fails on one link, the retransmission process can be arranged on another link. Similarly, the block acknowledgement (BA) corresponding to the transmission in one link can be sent in other links. Obviously, packet-level multi-link operation is more flexible and can further improve system performance, but more complex MAC control is needed.

Generally, there are two types of ML aggregation named synchronous and asynchronous aggregation. Wherein the synchronous ML aggregation means the different links should simultaneously transmit and simultaneously receive. This type is more suitable for the APs or STAs who cannot support simultaneous transmission and reception, named non-STR capability, due to the adjacent channel interference problem. The asynchronous ML aggregation means different links can transmit and receive independently, named STR capability enabled.

A special case of synchronous ML aggregation refers to aggregating the bandwidth of multiple bands into a larger bandwidth to transmit only one PPDU. For exapmle, see PPDU 5 in Fig. 3, sender aggregates the bandwidth of multiple bands into a larger bandwidth (e.g. 320MHz) and sends only one PPDU in this larger bandwidth. This mode is similar to channel bonding (CB). The normal mode of ML aggregation is to send independent PPDUs on multiple bands, e.g. PPDU 1 & 2 as well as PPDU 3 & 4 in Fig. 3.

Fig. 3
figure 3

Synchronous ML aggregation of IEEE 802.11be

Full size image

3.1.2 Multi-band management

Management function is an important function of IEEE 802.11, which affects the availability and efficiency of WLAN. IEEE 802.11be introduces more fine-grained multi-band operation, thus it will involve how to efficiently interoperate between multiple bands, such as multi-band scanning, multi-band association, multi-band authentication, multi-link setup and update, exchange status information, rapid traffic stream migration, and etc.

3.2 Investigation of possible technologies

To the best of our knowledge, there are few studies on multi-band operation in WLAN system. Our latest work in [8] directly faces to IEEE 802.11be and concludes that only introducing multiple bands without elaborate designs can hardly achieve the expected gain. After that, we describe several possible schemes of multi-band operation and investigate the feasibility of combining full-duplex with multi-band operation. Furthermore, we propose a unified MAC framework to support all possible multi-band operation with/without full duplex, i.e., one-fit-all. Specifically, the master AP sends an improved trigger frame (TF) which has been specified in IEEE 802.11ax to trigger the coordination between APs and/or schedule the resources between APs and STAs. This MAC framework can not only support all types of multi-band operation presented in the standard proposals, but also support AP coordination assisted and full-duplex assisted multi-band operation proposed in this study.

The authors in [44] propose a lower frequency band, i.e., 2.4GHz and 5GHz, assisted mmWave frequency transmission scheme. Operating in the 2.4/5GHz, if one STA is aware that its peer STA has been located in the same mmWave BSS, it will switch to mmWave to deploy the further data transmission process. In this case, the transmission capability is high since mmWave WLAN possesses much larger bandwidth. Furthermore, the overhead in mmWave WLAN such as beam scanning and neighbor discovery can be mitigated.

In [45], the authors consider the multi-band assembled WLAN system, wherein each band corresponds to one independent MAC layer. They formulate a resource allocation problem and further propose a packet scheduling algorithm in upper layer in order to minimize the transmission delay and the reorder delay.

3.3 Perspectives and open issues

  1. 1)

    Fast ML scanning, association, setup, and update

    Because the AP or STA of IEEE 802.11be can work in three bands at the same time, it contains lots of available channels (each channel has a bandwidth of 20 MHz). STA needs to scan the channel before it associate with one BSS, thus it will bring great challenges and costs for scanning and association. How to scan and association more efficiently becomes an important issue. After that, if STA needs to associate with each link independently, it will increase the overhead when setting up the ML relationship between AP and STAs. Therefore, how to efficiently associate and setup the ML relationship and keep fine backward compatibility is also an important issue. Finally, the ML relationship between AP and STAs may need to be updated caused by some reasons such as link ability update and link to traffic identifier (TID) mapping rule change.

  2. 2)

    Channel Access for synchronous and asynchronous ML operation

    If sending and receiving data packets in multiple links independently, the channel access problem is relatively simple. However, in many cases, multiple links are needed to coordinate with each other to perform channel access procedure. For example, if one PPDU needs to be sent by aggregating multiple bands, or if multiple links need to send and receive at the same time, we need to design a specific channel access method. Furthermore, if self-interference occurs between multiple adjacent links, i.e. non-STR scenario, the problem becomes more challenging.

  3. 3)

    Transmission Problem

    For packet-level ML operation, the frames from the same traffic stream can be sent in multiple links, even one frame can be retransmitted in other links. In this case, the acknowledgement policy and the reordering scheme need to be designed. Meanwhile, the packet allocation algorithm should be carefully designed.

  4. 4)

    ML Power Save

    Introducing multiple links significantly improves the maximum throughput, but it also results in high power consumption. If the power efficiency problem is not effectively solved, the quality of user experience will become worse. Therefore, IEEE 802.11be needs to find an efficient way to improve the power efficiency.

4 Multi-AP coordination

4.1 Standards progress introduction

IEEE 802.11be intends to introduce multi-AP coordination so that the network resources can be used more orderly and optimally. So far, five multi-AP coordination schemes have been proposed by IEEE 802.11be proposals, which are described as follows.

4.1.1 Coordinated spatial reuse (co-SR)

As shown in Fig. 4a, co-SR refers to the downlink (DL) transmission or scheduling of uplink (transmission) to their own STAs on the same channel through information exchange and negotiation between multiple APs. Since co-SR does not use joint MIMO, there is interferences between multiple links. In order to suppress the interference, the interaction information between APS is needed to ensure that the interference between the two links is acceptable. For example, as shown in the Fig. 4a, AP1 sends DL data to STA1 on its left side, while AP2 sends data to STA2 on its right side. The interference between the two links is light, thus the data of both links can be successfully received.

Fig. 4
figure 4

Multi-AP coordination of IEEE 802.11be

Full size image

4.1.2 Coordinated OFDMA (co-OFDMA)

UL and DL OFDMA was introduced by IEEE 802.11ax [46]. As shown in Fig. 4b, co-OFDMA refers to a master AP that successfully accesses the wireless channel coordinates with other APs called slave AP and allocates frequency resources among them. After that, both master AP and slave AP send DL data or schedules UL transmission simultaneously by using OFDMA. Each AP occupies a part of resource units (RU). The advantage of co-OFDMA is that there is no interference in theory before multiple links. As shown in Fig. 4b, AP sends data in RU 1 to STA 1 while AP 2 sends data in RU 2 to STA 2 by using OFDMA simultaneously.

4.1.3 Coordinated beamforming (co-BF)

As shown in Fig. 4c, co-BF refers to a master AP that successfully accesses the wireless channel coordinates with other slave APs. After that, both master AP and slave AP send DL data to their own STAs by using MIMO. As shown in Fig. 4c, AP 1 and AP 2 serve STA1 and STA2 respectively in MIMO mode. Because of the use of MIMO, the interferences of AP 1 to STA 2 and AP 2 to STA 1 have been eliminated theoretically.

4.1.4 Coordinated joint transmission (co-JT)

Co-JT is also named distributed MIMO (D-MIMO), as shown in Fig. 4d. Co-JT refers to a master AP that successfully accesses the wireless channel coordinates with other slave APs. All antenna resources from multiple APs and all users from multiple BSSs are pulled through together. After that, both master AP and slave AP send DL data to multiple STAs by using MIMO. It is noteworthy that co-JT allows AP to serve OBSS STAs and enables multiple AP to serve the same STA. As shown in Fig. 4d, AP 1 and AP2 pull-through antenna resources together and serve all STA: STA 1, STA 2 and STA 3 from two BSSs by MIMO.

4.1.5 Coordinated TDMA (co-TDMA)

Coordinated time division multiple access, as shown in Fig. 4e, refers to a master AP that successfully accesses the wireless channel coordinates with other slave APs to share the following transmission opportunity (TXOP). Different APs arrange their own AP UL/DL transmission at different time period in the TXOP. as shown in Fig. 4e, in the same TXOP, AP 1 deploy DL transmission first while AP 2 following.

Table 2 summarizes the different types of AP coordination.

Table 2 Summary and comparison between different types of AP coordination
Full size table

4.2 Investigation of possible technologies

4.2.1 Coordinated spatial reuse

Many studies focus on the improvements of spatial reuse by AP coordination. The first class is centralized solutions. The authors in [47] proposes a cooperative scheduling scheme for uplink OFDMA of IEEE 802.11ax in high-dense OBSS scenario. It divides all BSSs into several groups. The interferences between BSSs in the same group is very low. Therefore, the scheme controls the trigger frames of different groups to be sent in different time, and the trigger frames of the same groups can be sent in the same time. This enhances the spatial reuse capability, thereby improving the area throughput. Jiang et al. [48] assumes a software-defined network (SDN) like environment, and proposes a game theory based scheme for APs to control their transmitting power. The propose scheme achieves less interferences between BSSs. Similarly, [49] designs a SDN based WiFi network architecture, and further proposes a global downlink scheduling scheme.

For the distributed solutions, in [50], the authors propose a joint CCA and transmitting power adjusting scheme based on AP coordination. Specifically, APs exchange statistical information carried in the Beacon frame with each other. After that, the APs adjust their CCA level and transmitting power to guarantee the throughput fairness between APs. Chen et al. [51] propose a distributed coordination opportunistic transmission scheme for mmWave WLAN. To avoid interfering ACK frame caused by different simultaneous transmission links, the sender of the first link sends a short synchronization frame. After the sender of the second link receives the synchronization frame, it can transmit simultaneously with the first link, but the data transmission time should be aligned to avoid the ACK frame conflict. This scheme improves the spatial reuse capability.

4.2.2 Coordinated joint transmission

Coordinated joint transmission is also named as D-MIMO. Some studies focuses on the algorithms and protocols for D-MIMO. The authors in [52] assume that several APs form a group, within which there is a centralized controller. A user selection algorithm and a aggregate rate maximizing scheduling algorithm are proposed, respectively. After that, the authors extend their study in [53] by considering the scheduling for multiple time-slots. In [54], to minimize the CSI feedback overhead and to maximize the overall throughput, several D-MIMO groups are established. The APs in the same group utilize D-MIMO while different groups utilize orthogonal resources to avoid interferences. After that, joint AP grouping and STAs association algorithm are proposed.

Some studies pay attentions to the system design of D-MIMO. Hamed et al. [55] designs and implements a real-time D-MIMO system. Specifically, the authors propose a mathematical model and a protocol to achieve real-time channel update by using channel reciprocity. After that, coordinated power control scheme is proposed. The authors in [56] propose a D-MIMO system and focus on solutions to synchronization, channel status feedback, and STAs grouping problem. Hamed et al. [57] designs a fully distributed D-MIMO system by introducing self-organized architecture. Shepard et al. [58] proposes a design and implementation of D-MIMO system and corresponding beamforming algorithm. Gan et al. [59] proposes an OFDMA based synchronization protocol for D-MIMO. A D-MIMO protocol and device are designed in [60].

4.2.3 Coordinated beamforming

Coordinated Beamforming is not a quite new topic, and a series of related studies focus on it and achieve some fundamental results [61,62,63]. However, applying coordinated beamforming into WLAN has only attracted attention in recent years. Osama et al. [64] proposes a coordinated beamforming method for IEEE 802.11. The primary AP sends feedback trigger frame first, and receives the feedback information including the channel status from the candidate STAs. After that, the coordinated beamforming based transmission can be deployed. Yang et al. [65] details a coordinated beamforming method that after measuring the channel status and service requirements of STAs belonging to multiple BSSs, Each AP beamforms signals to a specific area while canceling other APs’ signals. The authors in [66] utilize coordinated beamforming to solve hidden terminal problem in WLAN which is an inherent problem of CSMA/CA based wireless system.

4.2.4 Coordinated network management

Some studies focus on the network management such as operating channels management and association management based on AP coordination. Bouhafs et al. [67] proposes a centralized solution based on SDN architecture. The controller collects and evaluates the interference information of each channel. Joint considering the interference information and users’ requirement, the controller optimized determine the operating channels and the transmitting power of each AP. Yang et al. [68] proposes a multiple AP association method for IEEE 802.11ax to improve the edge users’ throughput. Ma et al. [69] enables two-hop distance STAs associating with AP and assigning different operating channels for the first-hop and the second-hop. Furthermore, it allows the STAs to associate with multiple APs for the purpose of load balance. Karowski et al. [70] proposes a coordination scheme for neighbor discovery in WiFi networks by using multiple transceivers.

Some researchers try to use AP coordination to support mobile handoff. The authors in [71] face to the handover problem of WiFi, which is an important problem for high-dense deployed scenario. To improve the handover performance when STAs move from BSS to BSS, a network coding scheme is propose by sending additional coded frames. In [72], the authors allow the STAs to associate with multiple BSSs and introduce a AP clustering based handoff protocol for IEEE 802.11ay.

4.3 Perspectives and open issues

  1. 1)

    AP coordination grouping and master AP selection

    Several APs should probably form a coordination group, otherwise the coordination information is untrusted. Thus, how to automatically and efficiently form a coordination group is a fundamental problem to be addressed. Furthermore, master AP is necessary for many multi-AP coordination modes, thus the master AP selection is also an important problem to be solved.

  2. 2)

    Access and transmission for AP coordination

    For different multi-AP coordination modes, the process of negotiation between multiple APs is different, and the information of interaction between multiple APs is also different. Therefore, the cooperation process between multiple APs is the technical premise of multi-AP coordination, which needs to be solved first. Moreover, it is an important question that needs to be answered and carefully designed whether each coordination mode uses an independent protocol process or all coordination modes use an unified process.

  3. 3)

    Inter-AP Backhaul

    Co-JT needs one AP to send STAs’ buffer data to other APs, resulting in very huge amount of backhaul data. How to transfer data to other APs more quickly without incurring much air interface overhead is an important issue. Wired network backhaul is a solution, but it puts forward additional requirements for network deployment.

  4. 4)

    Sounding and measurement

    Each multi-AP coordination scheme needs to measure and acquire some information. For example, co-SR needs to measure the interference information between different links, and co-BF and co-JT need to deploy channel sounding. Therefore, how to collaborate and efficiently accomplish measurement tasks in multiple AP environments is a challenge. Moreover, the sounding and measurement feedback is overhead for IEEE 802.11. Therefore, how to find a tradeoff between measurement performance and feedback overhead is also an open problem to be solved.

  5. 5)

    Synchronization

    Different multi-AP coordination schemes have different requirements for synchronization between APs. The most stringent one is co-JT, which requires strict time, frequency and phase synchronization. In addition, due to the rapid channel changes in wireless environment, how to deal with out-of-step and resynchronization is also an important issue to be concerned.

5 Enhanced link reliability

5.1 Standards progress introduction

Link reliability is an important factor affecting throughput. Legacy IEEE 802.11 introduces two mechanisms. a) Retransmit mechanism. When a data frame transmission fails, the sender needs to re-contend the channel and retransmit the data frame until the transmission succeeds or the maximum retransmissions time is reached. b) Link adaptation (LA) mechanism or algorithm. The sender or the receiver (usually the sender) adjusts the modulation and coding scheme (MCS) level adaptively according to the historical transmission results or some information such as Signal to Interference and Noise Ratio (SINR) to adapt to the current air interface environment. However, the traditional retransmit mechanism and LA mechanism will lead to low efficiency and throughput suppression. As shown in the Fig. 5, when a MAC protocol data unit (MPDU) transmission fails, according to the LA algorithm, the sender may reduce the MCS, so that all MPDUs transmitted at the next transmission have to use the lower MCS. Moreover, and every information bit of the failed data has to be retransmitted, which is costly.

Fig. 5
figure 5

HARQ enabled transmission for IEEE 802.11be

Full size image

HARQ enables the retransmission frame to just contain part of information bits of failed frame and coded bits with or without the MCS decrease. In this case, as shown in the Fig. 5, for the retransmission, the sender transmit all the MPDUs by using the same MCS as the last transmission, while an HARQ redundant version of the failed MPDU 2 is transmitted. Therefore, the throughput can be guaranteed.

5.2 Investigation of possible technologies

HARQ has been adopted by cellular network for years. However, the technical details in WLAN and cellular network are quite different. Specifically, a) cellular network is scheduling based system while WLAN is contention based system, resulting in that the MAC layer procedures including HARQ of these two wireless networks are quite different. b) The frame format of these two systems are quite independent and different. c) The WLAN works on unlicensed band, thus the air interface environment is quite complicated comparing with cellular network. The reasons of unreliable transmissions are also complex. For example, channel fade loss, collision, and etc. This further increased difficulty in HARQ design for WLAN. Therefore, only in recent years, some scholars began to pay attention to HARQ technology in WLAN.

In [73], the authors propose an incremental redundancy HARQ (IR-HARQ) MAC protocol with fine backward compatibility. Specifically, the receiver replies clear-to-send frame (CTS) instead of acknowledgement frame (ACK) to declare that the transmission is error, and the network allocation vector (NAV) in the CTS indicate the retransmission time. Here the CTS frame represents NAK. After the sender successfully receive CTS/NAK, it should retransmit the data according to the retransmission time carried in NAV field. The authors extend the scheme into multi-band scenario in [74] to further improve the latency. The authors in [75, 76] combine chase combining HARQ (CC-HARQ) with NACK based channel reservation to improve the transmission reliability for WLAN. The NACK frame carries NAV and reserve the channel for HARQ retransmission. Kim et al. [77] proposes a detailed HARQ transmission method and design the frame format based on IEEE 802.11ac. Wang et al. [78] proposes a general HARQ method including the transmission scheme and the information indication scheme based on IEEE 802.11ax. It allows the ACK/NAC information to be aggregated with data frame. Furthermore, it also details the frame aggregation and block acknowledgement (BA) method.

Some related works study the HARQ scheme for unlicensed band networking. For example, [79] and [80] propose scheduling based uplink HARQ scheme in unlicensed band. Liu and Song [81] focuses on the ACK feedback scheme for HARQ and channel status feedback for unlicensed band in cellular network. Fang and Babaei [82] proposes an upplink HARQ scheme for listen before talk system. Based on the above, however, these related works are more suitable for cellular network such as Long Term Evolution Unlicensed (LTE-U) rather than WLAN.

5.3 Perspectives and open issues

  1. 1)

    The reason of transmission failure

    Since WLAN works in unlicensed band, the reasons for transmission failure are complex, such as channel deep fading and random interference. Different failure reasons have different coping methods. If the packet loss is caused by channel deep fading, it is not appropriate to use HARQ only, and LA should also be used together. If the packet loss is caused by random interference, HARQ is more appropriate, and it may not need to reduce the MCS. Therefore, how to accurately identify the cause of transmission failure in WLAN is an important issue.

  2. 2)

    Chase combining (CC) or incremental redundancy (IR)

    HARQ has two basic implementation approaches: CC and IR, each with its own advantages. The advantage of CC lies in its simple implementation, but its disadvantage lies in its limited performance gain. The advantage of IR is that it has better performance gain, but the disadvantage is that the implementation is more complex. Therefore, IEEE 802.11be needs to further study the implementation complexity and performance gain of these two approaches, and seek a tradeoff.

  3. 3)

    Protocol and PHY and MAC frame design

    Since it will be the first time for IEEE 802.11 to introduce HARQ, the MAC protocol supporting HARQ needs to be designed carefully. For example, for acknowledgement, legacy IEEE 802.11 only has positive ACK, that is, to reply to ACK after the recipient successfully receives the data frame. If the frame is not received successfully, no frame is replied. However, HARQ requires the receiver to reply ACK after receiving the data frame correctly and NAK after not receiving the data packet correctly. Therefore, IEEE 802.11be needs to support ACK and NAK functions. Moreover, the PHY frame also needs to be designed especially the HARQ information indication in the signal field in EHT preamble.

  4. 4)

    HARQ and link adaptation algorithm

    HARQ and link adaptation is highly correlative. The sender should adjust the HARQ mode and MCS dynamically according to the channel status, packet loss status, and QoS requirements. But it is quite difficult to obtain the global and realtime information. Operating in unlicensed band exacerbates this problem. Therefore, designing an efficient joint HARQ and link adaptation algorithm is quite important for implementation.

6 Latency & jitter guarantee

6.1 Standards progress introduction

Since legacy IEEE 802.11 has little corresponding supports for latency and jitter guarantees, and IEEE 802.11be has just started research on latency and jitter, no consensus has yet been reached.

Time-sensitive network (TSN) is a series of standards specified by IEEE 802.1. In order to guarantee delay and jitter, it consists of a series of functions. a) Time synchronization. All nodes involved in time-sensitive services need strict time synchronization. b) Resources Reservation. In order to guarantee delay, TSN reserves resources. c) Time gated queue and priority queue. Adding a gate to the queues can restrict or open the function of queue output packet by controlling ON/OFF. In addition, different queues have different priority. However, since TSN is standardized in IEEE 802.1, its design does not consider compatibility and adaptability with IEEE 802.11, thus how to adapt TSN to IEEE 802.11 is still a challenge [83,84,85]. Some studies focus on the TSN solutions for IEEE 802.11 [86, 87].

Moreover, a series of other potential technologies are proposed in TGbe to guarantee the low latency and jitter. For example, some proposals consider that multi-link operation is one potential solution for low latency, some proposals try to introduce preemption scheme, and some others focus on priority guarantee and channel access based solutions. It is worth noting that the solutions for worst case latency in IEEE 802.11be are still unclear.

6.2 Investigation of possible technologies

6.2.1 Redundancy transmission and preemption

Redundancy is an important option to enhance the transmission reliability, thereby improving the average latency. Rentschler and Laukemann [88] inroduces multi-radio redundant schemes to enhance the reliability and further improve the latency of WLAN by connecting the single wired network interface of one device to its doble IEEE 802.11 channels and adopting Parallel Redundancy Protocol (PRP). Halloush [89] proposes an interference (named jamming) detection scheme to detect the transmitting jamming before the time deadline. If the jamming is detected for a packet, it will be retransmitted in other channel named safe channel.

In [90], the authors introduces a narrow-band channel for the STAs to send busy tone to report the real-time requirements. If other STAs successfully receive the requirements, they will stop transmitting in the main channel, while the STA who sends requirements should preempt channel resource in the main channel.

6.2.2 Scheduling based solutions

Scheduling is an efficient technology since AP is the core element of WLAN especially IEEE 802.11ax enhances the schedule capability. The authors in [91, 92] focus on the real-time application support in hybrid OFDMA based transmission including explicit scheduling access and random access. Specifically, the authors propose a cyclic resource allocation algorithm for IEEE 802.11ax and a group resource allocation algorithm for IEEE 802.11be. The necessary protocol modification is also described. VR is the killer application for the near future wireless network. To decrease the UL delay of VR device, [93] introduces a scheduling based scheme by using UL OFDMA protocol of IEEE 802.11ax. The UL contention of VR devices are disabled by the AP. After that, AP periodically schedules UL transmission of multiple VR users. Li et al. [94] focuses on the stream video traffic in the hign-dense scenario, and proposes a QoE aware scheduling scheme. A new metric named packet importance index carried in the ToS field is introduced to indicate the importance of one packet, and then the authors improve the gradient based scheduling by considering the metric. The authors in [95] propose a hybrid scheme combining transmission and retransmission scheduling, redundancy, and bandwidth allocation to guarantee the real-time requirements.

Some studies focus on the static or quasi static configurations of WiFi. The authors in [96] proposes a scheme which regulate the traffic from MAC to the queue in PHY based on contention situation. The scheme guarantees that the delay-sensitive frame can avoid the queuing process. The authors in [97] introduce a decision tree to study the WiFi factors that influence the latency. After that, AP can change some configurations such as the transmission power, operating channel, and AP location to improve the latency. The authors in [98] focus on the load balancing for several BSSs. Packet deadline is considered for the AP change determination, and a game theory based AP selection algorithm is proposed.

The authors in [99] studies the scheduling scheme in polling or named contention-free based IEEE 802.11 MAC and propose a two-level priority queues to reduce the delay. Vishalakshi and Nagaraja [100] proposes a highest urgency first scheduling algorithm in contention-free mode of WLAN.

For WLAN based mesh or multi-hop network, the authors in [101] propose a hybrid coordination function (HCF) controlled channel access (HCCA) based scheduling scheme. It firstly estimates a series of metrics, e.g., the packet loss rate, future queue size, and etc. After that, the most urgent packets in the queue are scheduled in the next TXOP. Similarly, [102] proposes a packet scheduling scheme by adjusting the packet length according to the estimated link conditions.

6.2.3 Channel access optimization

Lin et al. [103] introduces neural networks to model the relationship between STAs’ performance, such as throughput and delay, and STAs’ EDCA channel access parameters. Then, the AP can optimally adjust and broadcast the EDCA parameters to all STAs to guarantee the differentiated QoS. Nguyen et al. [104] faces two types of services: delay-sensitive services and throughput-sensitive and proposes a proportional tradeoff scheme by adjusting the STAs’ TXOP limit and the minimum contention window to satisfy both types’ QoS. The authors in [105] propose a fast retransmission scheme for VoIP service. Specifically, if a VoIP frame encounters collision, the node will send a non-data packet named energy burst immediately after AIFS to let other nodes know such collision. After that, the collided frame can be retransmitted after a short backoff period while other nodes increase their backoff counter to guarantee the retransmission. Syed and Roh [106] proposes an algorithm to dynamically change the contention window of each AC based on the STA number. Similarly, the authors in [107] also adjust the contention window by Q-learning algorithm. In [108], the authors propose a new channel access protocol scheme to mitigate the tail latency of WiFi. The proposed protocol introduces a new concept named “Round”, during which only one frame is allowed to sent for each STA by modifying the backoff and contention window strategy.

Some works focus on the combination of energy efficiency and delay guarantee. Liu et al. [109] optimizes the sleep time and the wake-up time under the constraints of STA’s delay requirements to maximize energy-efficiency and to avoid the access contention when any STA wakes up. Meanwhile, [110] differentiates delay-sensitive service and other service. Immediate transmission is adopted for delay-sensitive service, while packets should be buffered for other services.

6.2.4 Transmission optimization

Optimizing the frame aggregation scheme is one important direction in transmission optimization. Zheng et al. [111] studies the performances of video traffic with the variation of per MPDU size and the retransmission time when frame aggregation is adopted. Furthermore, the authors find the optimal parameters under specific channel condition. Similarly, the authors in [112, 113], and [114] try to find the optimal aggregation sizes towards different QoS types. In [115], the authors study the retransmission scheduling algorithm for A-MPDU considering the delay-sensitive requirements.

Bobarshad et al. [116] studies the interaction relationship between expired frame dropping rate and the buffer overflow dropping. Furthermore, the authors propose a retransmission time adaption algorithm to improves the video traffic performance. The authors in [117] classify the packet loss situation into three types: collision, congestion, and channel errors. After that, a link adaptation algorithm is proposed. Khavasi and Suk [118] studies the link adaptation scheme for MU-MIMO in IEEE 802.ac to gurantee the real-time video services. Tramarin et al. [119] formulates an optimizing problem considering the deadline of packets, and then solve the problem by optimizing the transmission rate. The authors in [120] propose a deadline-constrained backoff scheme for periodic real-time services. Specifically, the STAs double their backoff counters not only when the collisions occur, but also when the channel is busy. Furthermore, a intra-AC_VO differentiation is introduced by modifying the retransmission time of periodic real-time flow. In [121], the authors propose an online reinforcement learning algorithm for the STAs to determine their MCS according to the channel status and buffer status in order to guarantee the energy efficiency and the delay-sensitive requirements.

6.2.5 Admission control

Charfi et al. [122] and Inaba et al. [123] propose frame aggregation based admission control algorithm and fuzzy-based admission control algorithm, respectively, to guarantee the existing STA’s latency and throughput. The authors in [124] propose a model to estimate the performance if a new flow is admitted to access. After that, admission control result is decided by the AP.

6.3 Perspectives and open issues

  1. 1)

    TSN adapted to IEEE 802.11

    TSN can effectively guarantee latency and jitter, but the technologies of TSN itself do not match 802.11 well. Thus, how to extract some appropriate features from TSN technologies and design appropriate mechanisms to make it naturally adapt to 802.11 is a great challenge.

  2. 2)

    Multi-band supported latency guarantee

    Since APs and STAs of IEEE 802.11be have multi-band capabilities, it is worth studying whether and how multi-band capabilities can support delay and jitter.

  3. 3)

    Enhanced AP scheduling

    An important reason for the poor latency and jitter guarantee of IEEE 802.11 is that distributed random access makes conflict and interference uncontrollable. The global scheduling capability of AP has been greatly enhanced since IEEE 802.11ax. Therefore, how to utilize AP’s powerful scheduling capability to support latency and jitter is a noticeable direction.

  4. 4)

    5G inspired latency guarantee scheme

    Since 5G significantly improves the latency performance by a series of solutions such as improved protocol stack and flexible slot configuration. Therefore, the solutions or the research thought of 5G might be the inspirations for WLAN.

7 Conclusions

In this article, we investigates and analyzes the key technologies of the next generation WLAN: IEEE 802.11be, established in May 2019, and further provides our perspectives and insights on it. We summarize that the target scenario of IEEE 802.11be is the ultra-high definition video traffic and real-time traffic. To efficiently support the target scenario, the key technical objective of IEEE 802.11be is to achieve extremely high throughput (at least 30Gbps) and to improve the worst case latency and jitter. After that, we further investigate and analyze the technical characteristics and challenges of multi-band operation, multi-AP coordination, enhanced link reliability, and latency & jitter guarantee, respectively. To the best of our knowledge, this is the first work to investigate, analyze and provide insights on IEEE 802.11be. Future work is to study the key technologies of IEEE 802.11be