Grant-Free NOMA-OTFS Paradigm: Enabling Efficient Ubiquitous Access for LEO Satellite Internet-of-Things

The GB-RA protocols, which were originally conceived for human-type communication, have dominated the existing cellular networks’ standards nowadays, e.g., long term evolution (LTE)/LTE-advanced (LTE-A) and 5G new radio (NR). Besides, they have been adopted by the standardized narrow-band IoT (NB-IoT) to support machine-type communications for co-existing with cellular networks [3]. These types of protocols stipulate that the active entities initialize the access procedures through a physical random access channel (PRACH) in a contention-based manner to set up the connection with the base station (BS). After the typical four-step handshake procedure is completed, the connected terminals are allowed to initiate a grant acquisition (GA) request for transmitting their payload data to the BS on the allocated orthogonal or non-orthogonal RBs.

In general, these solutions developed for terrestrial cellular networks are generally not applicable to LEO-SATs systems due to the harsh channel conditions in the TSLs, such as long propagation delay, severe Doppler effects, strong channel spatial correlation, and etc. For this purpose, some modifications have been made in recent endeavors [6, 7, 5, 4]. For LEO-SATs employing massive multiple-input multiple-output (MIMO) and full frequency reuse techniques, [4] proposed a space angle-based user grouping algorithm to schedule the served terminals into different groups, where the terminals having different angles of arrival (AoAs) are assigned to the same group and share the same TF resources. While the different groups are allocated to the orthogonal RBs to alleviate the potential interference caused by channel correlation impairment and to maximize the average signal-to-interference-plus-noise ratio. Without requiring any modification to the current NB-IoT waveform, [5] addressed the issues of MA caused by the typical satellite channel impairments, including long round-trip delays, significant Doppler effects, and wide beams. Furthermore, [6, 7, 8] integrated the OTFS modulation with the GB-RA protocols and formulated new resource allocation schemes to accommodate the heterogenous mobility demands of differnt IoT applications.

Despite the aforementioned efforts, some critical issues arise with the sophisticated access procedures in the context of LEO-SATs-based ubiquitous massive connectivity:

• •

  The access latency (e.g., 14.5 ms [3]) caused by standard handshakes and grant requests may far exceed the round-trip latency of TSL (e.g., 3 ms [5]) and constitutes as the main latency source.

• •

  The scheduling signaling overheads are prone to exceed the length of the short-packet payload data generated by IoT applications, leading to excessive waste of resources.

• •

  With the exponential growth of IoT devices, collisions and network congestion would occur with a high probability due to the limited pools of the orthogonal preamble sequences and the following endless retransmission attempts.

Consequently, these drawbacks of existing GB-RA protocols call for a new paradigm shift to adapt the pivotal features of LEO-SATs and to enable more efficient connections.

In this case, the GF-RA protocols arise as a better option to cope with the escalating IoT traffic demand over satellite communications, where the representative case is ALOHA [2]. The original ALOHA protocol allows the terminals to transmit the data packets without any prior coordination, thus only performs well at modest traffic intensity. To alleviate the network congestion and the potential network instability issues, more advanced GF-RA techniques have been developed, such as Contention Resolution Diversity ALOHA (CRDSA) and Enhanced Spread Spectrum ALOHA (E-SSA), which integrates diversity or spreading techniques with successive interference cancellation (SIC) [9]. In fact, Globalstar has deployed the simplex data network based on the SSA-RA protocol along with the concept of sending multiple replicas of the same burst to increase reception probability [9]. Meanwhile, some satellite communication standards, such as digital video broadcasting return channel via satellite (DVB-RCS) and DVB-satellite services to handhelds (DVB-SH), have widely adopted ALOHA-based GF-RA protocols. Nevertheless, the existing protocols commonly depend on OMA technique and require excessively increasing resources as the number of terrestrial IoT terminals to be connected grows exponentially.

Recently, an efficient GF-NOMA protocol has been proposed for terrestrial massive machine-type communication (mMTC) [10]. This protocol advocates to merge the RA process with data transmission by removing the GA request. For instance, when a terminal wakes up, it initializes the pre-configuration and synchronization by exploiting the beacon signals periodically broadcast from the BS. Before delivering the uplink payload data, the awakened terminals only need to transmit short non-orthogonal preambles for facilitating the detection of active terminals’ set (ATS) at the BS, rather than going through the complicated GA-related procedure. Furthermore, by integrating the NOMA technique and allocating unique signature patterns in the power, code, and/or spatial domain to different terminals, the GF-NOMA schemes can efficiently separate simultaneously served terminals from the assigned overlapped TF RBs via SIC or multi-user detection (MUD) algorithms [10]. At the expense of the complexity at the receiver, the GF-NOMA schemes can significantly reduce the overall latency and signaling overheads with less resource consumption and enhanced access capability.

The typical GF-NOMA schemes can be divided into two categories based on channel state information (CSI) acquisition: passive [11] and active types [13, 12]. The former passively utilizes the periodically broadcast beacon signals to update the CSI and performs equalization at the terminals or the BS [11]. This mechanism is effective in quasi-static channels since it saves extra resources to perform channel estimation (CE). The latter actively transmits pilot sequences to jointly estimate the ATS and CSI [12], [13]. As a result, it can better capture the variation of wireless channels, while at the expense of transmission efficiency.

Nevertheless, unlike terrestrial cellular systems, the wireless TSLs have high more dynamic characteristics due to the high mobility of satellite stations and thus robustness to fast time-varying channel conditions is an essential indicator for efficient MA paradigms. In this context, the aforementioned exsiting GF-NOMA schemes still have their limitations:

• •

  For the passive-type schemes, Doppler shift can be precompensated at the terminals by exploiting the satellite ephemeris. Yet, it is not a trivial task for the terminals to track the variation of Doppler shift precisely, since beacon signals tend to be broadcasted occasionally to limit the required resource consumption. Meanwhile, Doppler precompensation is only precise for terminals located at the beam centre. Inevitably, there will be Doppler residual for others [5].

• •

  For the active schemes, the pilots and data are usually multiplexed in the TF domain, such as orthogonal frequency division multiplexing (OFDM) systems. However, the severe Doppler effect would generally destroy the orthogonality between OFDM subcarriers, which makes it difficult to accurately estimate the ATS and Doppler shift. In this case, traditional techniques could be invalidated as well [13, 12].

To tackle these pressing issues and motivated by the strengths of burgeoning waveform–OTFS [14], we intend to incorporate it into the GF-NOMA protocols and heuristically propose a GF-NOMA-OTFS paradigm for LEO-SATs empowering efficient ubiquitous connections in the next section.

As an amalgamation of the GF-RA mechanism with the OTFS waveform, the transmission procedure of the proposed GF-NOMA-OTFS paradigm can be summarized as follows: Step 1) After initializing the pre-configuration and the time synchronization, the active IoT terminals are allowed to directly transmit RA uplink signals without complicated access and resource scheduling request; Step 2) The satellite stations collect signals received from all active terminals and perform active terminal identification (ATI) and CE, where the estimated ATS and CSI are used for the following DD domain data demodulation in the absence of scheduling information; Step 3) The satellite stations feedback the demodulated data through inter-satellite links (ISLs) and backhaul to the gateway and broadcast acknowledgment character to the active terminals. In order to achieve more efficient ATI and CE, we first propose an improved RA signal frame structure for OTFS waveform.

To compare the existing MA schemes surveyed in Section II and demonstrate the superiority of the GF-NOMA-OTFS paradigm and the frame structure design, we study a representative case in the LEO-SATs-based massive IoT access and provide simulation results. For comparison, we consider the following schemes:

• •

  GF-NOMA with OFDM waveform [13]: the uplink RA signals utilize frequency-domain pilots and space-frequency structured sparsity of access channel matrices to perform ATI and CE. Its effective pilot length is defined as the time slots occupied by frequency-domain pilots.

• •

  GF-NOMA-OTFS with traditional frame structure [14]: the traditional OTFS RA signals utilize DD domain embedded pilots and its 2-D circular convolution interaction with discrete DD structured-sparse access channel matrices to perform ATI and CE, cf. Fig. 2. Its effective pilot length is defined as the size of pilot block along delay dimension.

• •

  GF-NOMA-OTFS with virtual oversampling lattice [14]: the virtual oversampling representation is further utilized to improve virtual Doppler and delay resolutions and mitigate access channel spreading due to discretization, cf. Fig. 2.

• •

  GF-NOMA-OTFS with proposed TS: the uplink RA signals utilize proposed TS and DD domain sparsity of access channel matrices to perform ATI and CE. Its effective pilot length is defined as the non-ISI region of TS.

The simulation parameter settings are summarized in TABLE II. It is noteworthy that the LEO satellite is expected to be equipped with a large phased array to enhance beamforming capability, which requires numerous radio frequency components including highly linear power amplifier and analog-to-digital converter (ADC). Therefore, we consider to employ low-resolution ADCs for receiver of LEO satellite, which enjoy low power consumption and simple hardware implementation, to meet the on-board power and cost constraints.

Fig. 3 and Fig. 3 respectively depict the activity error rate (AER) and normalized mean square error (NMSE) of different schemes as a function of effective pilot length. Specifically, the severe Doppler effect incurs ICI for the GF-NOMA scheme employing OFDM waveform and thus dominates significant performance loss regardless of pilot overhead. In contrast, the proposed GF-NOMA-OTFS paradigms enjoy performance superiority by handling the Doppler effect with their DD domain signal processing manner and the exploitation of the DD domain channel’s property. Unfortunately, due to the impact of discretization, the channel spreading and the limited resolution destroy the beneficial features of TSLs in the DD domain, degrading its optimal performance. The utilization of virtual oversampling lattice can compensate the performance loss to some extent while at the cost of higher computation complexity. Moreover, the proposed TSs enable the combination of CS-based estimation in the delay domain and super-resolution estimation in the time domain, ensuring superior performance while keeping lower pilot overheads.

On this basis, we further compare the performance of bit error rate (BER) in Fig. 3 based on the estimated ATS and CSI, where a straightforward DD domain LS multi-user signal detector is utilized to perform SD. The performance gain achieved by the proposed GF-NOMA-OTFS paradigm suggests the noticeable benefits from more accurate ATS and CSI, and the full channel diversity extraction provided by OTFS modulation. Furthermore, it is noteworthy that with the limited resolution ADC, i.e., $B^{\rm adc}=2,3\,\rm bits$, the performance of the GF-NOMA-OTFS paradigms degrades sluggishly in contrast to ideal ADC, which indicates the proposed schemes remain robust in this case, making hardware-efficient implementation possible for satellite payloads.

The IoT terminals are expected to be power-constrained for low carbon footprint and cost-efficient deployment, which yet may result in failure to close the link considering the significant path loss. At the cost of achievable data rate, the repetition code spreading specified in NB-IoT standard [5], and the direct sequence spread spectrum widely used in satellite communications [15], can be integrated with the OTFS modulation to improve link SNR and enhance robustness with low transmit power. In addition, imperfect hardware components remain another challenge. Practical satellite communication systems usually work within the saturation point of power amplifiers to provide the most efficient power output and compensate the significant path loss, yet results in severe non-linear behaviors. In particular, most multi-carrier waveforms are with inconstant signal envelopes, which is likely to lead to non-linear distortion in this case. On the other hand, low-resolution digital-to-analog converters and ADCs, have attracted increasing research attention to be equipped for terminal and satellite apparatus. In the presence of non-ideal characteristics, undesired interference is often inevitable. It is suggested that the joint optimization of performance taking these on-board and terrestrial constraints into account is an underlying trend in the future research of MA paradigms [13].

In addition to SDMA, other cutting-edge NOMA paradigms, e.g., power-domain NOMA [7] and sparse code MA [8], have been considered to integrate with OTFS waveform to enhance spectral efficiency and alleviate the spatial correlations caused by overloaded SDMA. Therefore, it is rewarding to investigate their amalgamation with the proposed MA scheme for more efficient payload data multiplexing, striking a tradeoff between the spectral efficiency and transmission reliablity. In this case, SIC and MUD algorithms are pivotal but challenging in the presence of more complicated multi-dimensional (Doppler, delay, spatial domains) interweaving and inter-users interference. In fact, under the framework of single-input single-output OTFS, a myriad of SD schemes have been proposed from different perspectives [14], including but not limited to execution in various domains, e.g., the DD domain, the TF domain, and the cross-domain iteration; different optimization objective, e.g., lower complexity and superior BER performance; and different techniques, e.g., zero forcing, minimum mean square error, ML, and message passing. On this basis, their extension to more sophisticated multi-user NOMA systems should be further studied.

Benefiting from the ultra-dense network topology, the concept of multi-connectivity arises in LEO-SATs networks [2]. In this case, each terminal can be covered by multiple satellite stations simultaneously, and thus more available connection options arise: an IoT terminal may communicate with the preferred satellite for the shortest link or connect to multiple satellites concurrently with the aid of ISLs for exploiting diversity gain and avoiding frequent handover. Besides LEO-SATs, assortments of non-terrestrial infrastructures, such as high altitude platforms and unmanned aerial vehicles are expected to be integrated with the terrestrial interfaces to constitute an integrated space-air-ground-sea network for 6G communications [13]. Therefore, how to coordinate the intra-network and inter-network cooperation with the NGMA scheme to facilitate the seamless integration of such multi-dimensional heterogeneous networks should be further investigated for QoS enhancement. To this end, key technologies, such as interference management, cognitive radio, and software-defined networking [2], which show promise to cope with more complicated interference scenarios among different interfaces and contribute to more flexible network architectures, are worthy of further exploration.