PaperlayerView exact revision on arXiv

Source-linked AI summary

Grant-free Non-orthogonal Multiple Access for IoT: A Survey

Muhammad Basit Shahab, Rana Abbas, Mahyar Shirvanimoghaddam, Sarah J. Johnson

arXiv:1910.06529v1eess.SP

TL;DR

Existing NOMA research largely assumes centrally scheduled, already connected users, while mMTC requires autonomous uplink access with limited signaling. This survey organizes NOMA techniques for grant-free connectivity, discusses blind multiuser detection and practical challenges, and outlines future research directions.

  • Problem

    Existing NOMA analyses mainly assume centralized scheduling, predefined system parameters, and perfect synchronization, conditions that may not apply to mMTC.

  • Method

    The article comprehensively surveys NOMA techniques for grant-free uplink connectivity and discusses their design, blind multiuser detection, challenges, and future directions.

  • Results

    The survey categorizes candidate uplink NOMA techniques, explains grant-free designs and blind multiuser detection, and identifies practical challenges and possible solutions.

  • Takeaways & Limitations

    Grant-free NOMA research must address autonomous access, receiver-side activity detection, control-signaling overhead, and practical system-design challenges.

Abstract

from arXiv · show

Massive machine-type communications (mMTC) is one of the main three focus areas in the 5th generation (5G) of mobile standards to enable connectivity of a massive number of internet of things (IoT) devices with little or no human intervention. In conventional human-type communications (HTC), due to the limited number of available radio resources and orthogonal/non-overlapping nature of existing resource allocation techniques, users need to compete for connectivity through a random access (RA) process, which may turn into a performance bottleneck in mMTC. In this context, non-orthogonal multiple access (NOMA) has emerged as a potential technology that allows overlapping of multiple users over a radio resource, thereby creating an opportunity to enable more autonomous and grant-free communication, where devices can transmit data whenever they need. The existing literature on NOMA schemes majorly considers centralized scheduling based HTC, where users are already connected, and various system parameters like spreading sequences, interleaving patterns, power control, etc., are predefined. Contrary to HTC, mMTC traffic is different with mostly uplink communication, small data size per device, diverse quality of service, autonomous nature, and massive number of devices. Hence, the signaling overhead and latency of centralized scheduling becomes a potential performance bottleneck. To tackle this, grant-free access is needed, where mMTC devices can autonomously transmit their data over randomly chosen radio resources. This article, in contrast to existing surveys, comprehensively discusses the recent advances in NOMA from a grant-free connectivity perspective. Moreover, related practical challenges and future directions are discussed.

I. INTRODUCTION … C. Massive Connectivity

The paper frames mMTC as an autonomous, predominantly uplink IoT scenario with tiny device payloads, stringent energy constraints, diverse QoS, and massive scale. It reviews connectivity technologies and identifies massive connectivity as a remaining challenge requiring protocol evolution beyond HTC-oriented designs.

  • I. INTRODUCTION: IoT connects physical devices through wired or wireless networks for autonomous services such as remote monitoring and real-time multi-device control.Examples include connected cars and homes, moving robots, and sensors.
  • A. IoT Traffic Framework: 3GPP and ITU define eMBB, mMTC, and URLLC as network usage scenarios reflecting varied connected-device types and QoS requirements.The framework distinguishes massive machine-type communications from human-oriented communication use cases.
  • A. IoT Traffic Framework: mMTC traffic is mainly uplink, carries very small per-device payloads, requires high energy efficiency, operates partly or fully autonomously, and has diverse QoS requirements.Massive IoT applications may tolerate some reduction in data reliability and latency.
  • A. IoT Traffic Framework: More than 53 percent of projected connected devices are MTCDs and consumer electronics, motivating a dramatic shift from protocols mostly designed for HTC.The scale and QoS diversity of these devices create requirements that current HTC-oriented protocols do not directly address.
  • B. Wireless Connectivity Options: IoT connectivity spans short-range technologies such as WiFi, Bluetooth, and Zigbee and wide-area networks selected according to coverage and performance needs.The passage presents wireless options as serving different IoT use cases across network-coverage ranges.
  • B. Wireless Connectivity Options: 3GPP’s low-power wide-area cellular options include EC-GSM, NB-IoT with 200 kHz bandwidth on existing LTE networks, and LTE-M with power-saving functionalities.These technologies introduce adaptations for massive IoT applications while leveraging cellular-network infrastructure.

1) Orthogonal/non-orthogonal Multiple Access: … A. Channel Access Methods in LTE/LTE-A

The paper motivates grant-free NOMA for mMTC by contrasting OMA’s resource bottleneck and LTE/LTE-A’s grant-based access overhead with autonomous transmission over shared resources. It surveys grant-free NOMA research for uplink IoT communication, alongside theory, challenges, and future directions.

  • 1) Orthogonal/non-orthogonal Multiple Access:: OMA allocates each radio resource to one device, so limited resources and massive device populations create a connectivity bottleneck.NOMA instead overlaps multiple users on a single radio resource block to expand connectivity over scarce resources.
  • 1) Orthogonal/non-orthogonal Multiple Access:: Most NOMA studies assume connected users, centralized scheduling, and capacity maximization, unlike autonomous mMTC traffic.Under these assumptions, users and the base station or eNB are presumed to know extensive system information.
  • 2) Grant-based/grant-free Transmission:: LTE/LTE-A devices request transmission slots through contention-based random access followed by a multi-step handover or granting process.This procedure has been identified as a major performance bottleneck that causes excessive delay and signaling overhead.
  • 2) Grant-based/grant-free Transmission:: Grant-free access allows mMTC devices to transmit as needed without random access or resource granting, or by merging random access with data transmission.The paper presents this shift as necessary for mMTC traffic.
  • D. Contributions and Organization: Unlike prior surveys focused on centralized or grant-based NOMA, this article comprehensively summarizes recent NOMA work from a grant-free perspective.Its primary focus is grant-free access for uplink IoT communication.
  • D. Contributions and Organization: The paper introduces grant-free access and reviews recent grant-free NOMA schemes designed for mMTC.It also discusses grant-free access from an information-theoretic perspective.
  • D. Contributions and Organization: The paper highlights challenges in grant-free mMTC NOMA and proposes future directions for addressing them and designing new schemes.These topics form later parts of the paper’s organization.
  • A. Channel Access Methods in LTE/LTE-A: In LTE/LTE-A, orthogonally allocated time or frequency resources require devices to coordinate through PRACH before accessing the eNB.The LTE standard prescribes a four-way handshake for PRACH access, supporting initial access, uplink synchronization, data transmission or acknowledgment, and handover management.

B. Related Performance Issues · C. Proposed Access Method Modifications for mMTC Traffic

Conventional LTE random access becomes a latency, congestion, reliability, overhead, and energy bottleneck for massive mMTC traffic. Proposed modifications include differentiated and optimized RACH procedures, while grant-free NOMA uplink access addresses scheduling-request overhead and limited resources.

  • B. Related Performance Issues: LTE random access introduces multiple delay sources, including RACH, scheduling requests, transmission intervals, signal processing, and packet retransmissions.These components contribute to the overall latency of accessing and transmitting data.
  • B. Related Performance Issues: Congested PRACH lowers receiver SINR, causing message-detection failures and outage when many MTC or HTC devices attempt access simultaneously.Excessive preamble collisions and retransmissions further degrade access performance.
  • B. Related Performance Issues: 100 B of uplink data requires around 59 B of uplink overhead and 136 B of downlink overhead for signaling.The resulting overhead contributes to network congestion, packet loss, radio-resource wastage, and high energy consumption.
  • C. Proposed Access Method Modifications for mMTC Traffic: Access class barring can temporarily delay or block traffic classes to alleviate congestion-related outage.Different back-off windows and predefined access-attempt slots can further differentiate access among traffic classes.
  • C. Proposed Access Method Modifications for mMTC Traffic: NR RACH optimization includes support for multiple preamble formats with shorter or longer preamble lengths.LTE currently provides five PRACH formats with different preamble lengths.
  • C. Proposed Access Method Modifications for mMTC Traffic: A single LTE RACH timeline may be inefficient for NR services and use cases with different latency requirements.Variable-length RACH timelines are therefore needed for diverse use cases.
  • C. Proposed Access Method Modifications for mMTC Traffic: A 2-step RACH procedure is considered for intermittent small-packet transmissions, unlicensed-spectrum access, and cases without uplink timing advance.These use cases are proposed for consideration in 3GPP NR.
  • C. Proposed Access Method Modifications for mMTC Traffic: Grant-free NOMA-based uplink access is jointly supported by academia and industry to reduce scheduling-request latency and signaling overhead for massive IoT access.It also addresses limited radio resources and related access challenges.

III. NON-ORTHOGONAL MULTIPLE ACCESS … IV. GRANT-FREE ACCESS FOR MMTC

The paper surveys NOMA schemes that multiplex users over shared radio resources and motivates grant-free UL NOMA for mMTC to reduce centralized-scheduling overhead. It organizes NOMA by user-separation mechanisms and describes autonomous transmission with RACH-based or RACH-less operation.

  • III. NON-ORTHOGONAL MULTIPLE ACCESS: NOMA breaks OMA orthogonality by multiplexing different data streams over the same radio resource blocks for 5G and beyond.Its distinguishing principle is shared time-frequency resources combined with multi-user separation.
  • III. NON-ORTHOGONAL MULTIPLE ACCESS: NOMA schemes distinguish multiplexed users through spreading, scrambling, interleaving, or other domains while sharing the same time-frequency resource blocks.The 3GPP RAN WG1 studied several candidate solutions for 3GPP NR.
  • A. Spreading Based: Spreading-based schemes share time-frequency resources through user-specific sequences and divide into low-density spreading and non-low-density spreading categories.LDS sequences are sparse or non-orthogonal low-cross-correlation sequences created by switching off many spreading signature chips.
  • B. Scrambling Based: Scrambling-based NOMA uses distinct scrambling sequences or patterns for user separation, with successive-interference cancellation supporting multi-user detection.Examples include PD-NOMA, RSMA, and LSSA, and low-code-rate FEC may be used with the sequences.
  • C. Interleaving Based: Interleaving-based NOMA distinguishes users with different interleavers and may combine interleaving with low-code-rate FEC and an elementary signal estimator.Prominent schemes include IDMA, IGMA, and RDMA.
  • IV. GRANT-FREE ACCESS FOR MMTC: Centralized analyses predefine spreading sequences, interleaving patterns, or transmission powers at the eNB, creating excessive signaling overhead and motivating grant-free NOMA.The paper presents grant-free communication as necessary for addressing this drawback in mMTC.
  • IV. GRANT-FREE ACCESS FOR MMTC: Grant-free UL NOMA lets MTCDs transmit without dynamic explicit scheduling grants and share the same time-frequency resources through NOMA.RACH-based operation synchronizes users before transmission, whereas RACH-less operation eliminates RACH and starts data transmission when packets are available.

V. GRANT-FREE NOMA SCHEMES · A. Signature based Grant-free NOMA Schemes · 1) Spreading based grant-free NOMA schemes::

The paper organizes grant-free NOMA into MA signature-, compressive sensing-, and compute-and-forward-based classes, then examines spreading-based access through contention-based units (CTUs). Spreading schemes support autonomous multiplexing over shared resources, while code-pool size, collisions, and receiver complexity remain key design considerations.

  • V. GRANT-FREE NOMA SCHEMES: Grant-free NOMA is categorized into MA signature-, compressive sensing-, and compute-and-forward-based approaches.Machine learning applications are discussed separately, and the schemes are summarized in Table V.
  • A. Signature based Grant-free NOMA Schemes: An MA resource combines a physical time-frequency block with signatures such as codebooks, sequences, interleavers, pilots, power, spatial dimensions, or preambles.MTCD signatures may be randomly selected or pre-configured.
  • 1) Spreading based grant-free NOMA schemes::: Spreading-based grant-free NOMA uses long or short spreading sequences selected from a predefined codebook or resource pool, with a CTU defining the contention-based transmission unit.A CTU contains radio resources, reference signals, and spreading sequences, and differs from other CTUs in at least one field.
  • 1) Spreading based grant-free NOMA schemes::: SCMA maps users’ bits to sparse codewords from dedicated codebooks, enabling 6 users to multiplex over 4 subcarriers, or 150 percent loading.The example uses K = 4, N = 2, and J = 6 codebooks, with each codeword’s non-zero-entry location uniquely determined.
  • 1) Spreading based grant-free NOMA schemes::: A contention-based SCMA resource pool contains L × J CTUs, combining time, frequency, codebook, and pilot-sequence choices.MTCDs can select CTUs using an ID-based rule; selecting the same CTU causes a collision that may be resolved by random back-off.
  • 1) Spreading based grant-free NOMA schemes::: For SCMA, changing spreading factor and sparsity can generate 70 codebooks at K = 8 and N = 4, increasing the number of CTUs and improving grant-free access.The contention region can coexist with regularly scheduled uplink transmissions in separate bandwidth portions.
  • 1) Spreading based grant-free NOMA schemes::: Short, relatively low-cross-correlation codes are suitable for grant-free uplink MUSA because long codes combined with SIC increase processing complexity, delay, and error propagation.Complex spreading codes can remain short by using design freedom in their real and imaginary parts.
  • 1) Spreading based grant-free NOMA schemes::: MUSA can generate a larger spreading-code pool than SCMA and PDMA, reducing collision probability, but blind MUD complexity increases as the pool grows.The pool size should therefore balance massive connectivity against receiver complexity, while spreading-sequence design controls inter-user interference and system performance.

2) Scrambling/interleaving based grant-free NOMA schemes:

Scrambling- and interleaving-based NOMA schemes are considered for grant-free mMTC uplink communication, while PD-NOMA with SIC is limited by uncontrolled user power differences without closed-loop power control.

  • Scrambling/interleaving based grant-free NOMA schemes:: PD-NOMA with SIC is constrained in grant-free access because its performance depends on power differences that cannot be controlled without closed-loop power control.Randomly transmitting near-far users and MTCDs create uncontrolled power differences over the same resource block.
  • Scrambling/interleaving based grant-free NOMA schemes:: RSMA supports grant-free and potentially asynchronous uplink transmission for mMTC, while interleaving- and multiple-domain NOMA schemes can likewise be customized for grant-free communication.The related schemes are summarized in Tables III and IV.

3) Other signature based grant-free NOMA schemes:

Signature-based grant-free NOMA schemes combine random access and data transmission over randomly selected subbands, using coded signatures and receiver-side load estimation with SIC. Random NOMA can support more devices than access class barring, but subband-dependent rates create variable slot durations and unique-seed designs impose scalability and detection complexity.

  • Random NOMA: Raptor-code-based Random NOMA lets MTCDs transmit without a separate random-access phase by combining access and data transmission over randomly selected subbands.Each device selects a subband and seed, attaches its terminal ID, and transmits a Raptor-coded message.
  • Random NOMA: The eNB estimates the multiplexed-user load on each subband and then applies successive-interference cancellation to recover superimposed messages.Equal received powers from MTCD power control enable load estimation from total received power.
  • Limitations: Random subband overlap makes each subband’s achievable rate and coded-symbol count variable, so slot duration is determined mainly by the most heavily loaded subband.Two consecutive slots can therefore contain subbands with different durations.
  • Performance: The scheme supported significantly more devices than access class barring in comparisons of average successful MTCD support.Collisions from users selecting the same seed and subband were reported as less probable than conventional random-access collisions.
  • Limitations: Preassigned user-specific seeds could eliminate collisions, but massive seed pools and seed-search detection become complicated as the MTCD population grows.The eNB would need to try many seeds to recover multiplexed users on each subband.

B. Compressive Sensing based Grant-free NOMA Schemes · C. Compute-and-forward based Grant-free NOMA Schemes

The paper presents compressive sensing as a way to exploit sporadic user activity for grant-free uplink NOMA, enabling joint activity and data detection with fewer measurements. It also describes compute-and-forward schemes that recover coded linear combinations from randomly selected sub-blocks before decoding individual messages.

  • B. Compressive Sensing based Grant-free NOMA Schemes: Compressive sensing exploits sparse mMTC activity to recover desired signals from fewer measurements and help the eNB handle more users.The approach addresses multi-user detection by leveraging the fact that active users are usually fewer than the total user population.
  • B. Compressive Sensing based Grant-free NOMA Schemes: CS-MUD supports grant-free uplink NOMA by jointly detecting sporadic user activity and data, while user-specific signatures distinguish active users.This addresses blind multi-user detection at the eNB without requiring prior activity information.
  • B. Compressive Sensing based Grant-free NOMA Schemes: Grant-free NOMA is formulated using SMV-CS for one-shot vector observations or MMV-CS, which reduces sensing-matrix size and mitigates complexity as users increase.Both models represent sparse spreading-based NOMA in uplink mMTC scenarios.
  • B. Compressive Sensing based Grant-free NOMA Schemes: CS-MPA detectors combine compressive sensing with message passing to perform activity and data detection when user activity is unknown and dynamically changes.Related designs use two-stage correlation and CoSaMP detection, or switch between CS-MUD and classical MUD as sparsity varies.
  • B. Compressive Sensing based Grant-free NOMA Schemes: Alternative activity-detection methods include channel-estimation-based detection, FOCUSS, expectation maximization, blind detection, and sparsity-inspired sphere decoding.These methods provide alternatives or complements to CS-based receivers for identifying active pilots and detecting transmitted data.
  • C. Compute-and-forward based Grant-free NOMA Schemes: Compute-and-forward grant-free NOMA uses nested lattice codes so integer combinations of codewords remain codewords, allowing the destination to choose which linear equation to recover.The scheme interprets network coding as converting the network into reliable linear equations.
  • C. Compute-and-forward based Grant-free NOMA Schemes: Each channel use is divided into sub-blocks, and every active user randomly selects one sub-block for transmission using a common concatenated codebook.The codebook contains an inner binary linear code for modulo-2 sums and an outer code for recovering individual messages.
  • C. Compute-and-forward based Grant-free NOMA Schemes: The receiver first decodes the sum of codewords in the compute-and-forward phase, using suitable inner codes and outer codes constructed from T-error correcting BCH codes.Inner-code design targets binary input memoryless output-symmetric channels, for which off-the-shelf codes can be used.

VI. MACHINE LEARNING IN GRANT-FREE NOMA · A. Machine learning in grant-based NOMA: · B. Machine learning in grant-free NOMA:

The section surveys a growing body of machine-learning research in grant-based and grant-free NOMA, motivated by ML’s potential to solve difficult communication optimizations efficiently and robustly. It introduces core learning paradigms before describing applications to channel processing, codebook design, detection, grant-free recovery, latency, and overhead reduction.

  • VI. MACHINE LEARNING IN GRANT-FREE NOMA: ML can solve NP-hard wireless optimization problems faster, more accurately, and more robustly by learning patterns rather than relying on explicit models and equations.The section reviews a growing literature applying ML to both grant-based and grant-free NOMA.
  • VI. MACHINE LEARNING IN GRANT-FREE NOMA: Supervised learning minimizes input–label mapping error, whereas unsupervised learning extracts features from unlabelled data, including through autoencoders.Reinforcement learning instead trains through trial-and-error rewards, and deep learning is an ML subset associated with deep neural networks.
  • A. Machine learning in grant-based NOMA:: In grant-based NOMA, LSTM-based deep learning performs automatic encoding, decoding, and channel detection to address the complexity of traditional methods in fast-changing channels.The system considers randomly deployed users and learns channel characteristics.
  • A. Machine learning in grant-based NOMA:: Deep neural networks design SCMA codebooks that minimize block error rate, adapt to available resources, and support single-shot decoding.The approach targets the high-dimensional, traditionally hand-crafted codebook-design problem.
  • A. Machine learning in grant-based NOMA:: An online ML detector for clustered uplink NOMA uses partially linear beamforming to improve robustness to changing cluster sizes and detect error propagation.The method addresses performance deterioration of nonlinear beamformers in dynamic networks where devices may join sporadically.
  • B. Machine learning in grant-free NOMA:: For uplink grant-free NOMA, statistical and ML cross-validation estimates user sparsity and determines when compressed-sensing multiuser detection should terminate without prior sparsity or noise knowledge.The estimated sparsity is mathematically referred to as the model order.
  • B. Machine learning in grant-free NOMA:: A deep-learning grant-free NOMA design jointly models encoding, user activity, signature-sequence generation, and decoding, with very low latency suitable for tactile IoT applications.Its detailed network structure is presented in Fig. 17.
  • B. Machine learning in grant-free NOMA:: Random and structured sparsity-learning MUD schemes support synchronous and asynchronous transmissions, respectively, achieving low error rates without pilot signals and reducing grant-free access overhead.The proposed algorithms target multiuser detection when users do not use pilot signals.

C. The Way Forward … C. Random Coding Bound

The paper identifies benchmarking, modeling, channel-statistics, and collision-handling gaps that limit comparisons and practical grant-free NOMA research. Its information-theoretic discussion progresses from finite-blocklength MAC bounds to many-access and common-codebook random-coding formulations for massive grant-free systems.

  • C. The Way Forward: Existing studies report faster processing and near-optimal solutions, but their independence prevents clear comparisons, especially among learning methods addressing the same problem.The paper presents these concerns as issues for future work.
  • C. The Way Forward: Benchmarks should move beyond basic, outdated traditional baselines and include recent cutting-edge solutions from the literature.
  • C. The Way Forward: Simple, practical system models could enable comparison with theoretical limits, clarifying proximity to optimal performance when machine-learning results are difficult to verify analytically.The passage emphasizes that theoretical fundamental limits are needed because the best achievable ML performance is difficult to assess analytically.
  • C. The Way Forward: Learning approaches should address unknown and practical channel statistics because current algorithms rely heavily on channel models.This is tied to retaining the ability to tune parameters on the fly.
  • C. The Way Forward: Grant-free NOMA user detection must handle collisions because unique signature sequences are impractical in massive-user settings and collisions bottleneck performance.The paper calls for improved collision detection and resolution.
  • VII. INFORMATION THEORETIC PERSPECTIVE OF GRANT-FREE NOMA: Classical fixed-user, infinitely long-blocklength MAC capacity is well understood, while grant-free channels require bounds for a large transmitter population with a small variable active subset.The latter coding problem was recognized decades ago as a distinct challenge.
  • A. Capacity of Gaussian MAC in Finite Block-Length: For finite blocklength with fixed K, OMA lies strictly below capacity, whereas capacity requires NOMA and joint decoding; evaluating the bounds generally involves 2K-dimensional probability spaces.These bounds are therefore tractable only for small K.
  • B. The Gaussian Many Access Channel: The many-access paradigm allows user count and block length to grow together, including user counts scaling linearly with block length, matching massive-machine-type communication conditions where activity is unknown.This scaling can exceed the block length, making classical theory inapplicable.

D. Achievability of the Random Coding Bound … E. Link adaptation and power control

The paper establishes rateless-code achievability for grant-free random access and outlines practical requirements spanning resource configuration, synchronization, activity detection, collision management, link adaptation, and power control.

  • D. Achievability of the Random Coding Bound: Rateless codes achieve the random coding bound with identical first- and second-order performance whether user activity is known or unknown.For a symmetric multiple access channel, decoding uses a single threshold rather than 2K −1 simultaneous threshold rules.
  • VIII. PRACTICAL CHALLENGES AND SOLUTIONS FOR GRANT-FREE UL NOMA: Grant-free NOMA requires investigation of resource definition, allocation, selection, synchronization, activity and data detection, collision management, HARQ, link adaptation, and power control.These procedures are presented as practical challenges for supporting and optimizing grant-free transmission.
  • A. Resource definition, allocation and selection: Before transmission, grant-free radio resources must be predefined and known to both the UE and BS.A CTU may combine time-frequency resources with pilots for channel estimation or activity detection and MA signatures such as codebooks, sequences, or interleavers.
  • B. Synchronization among devices: Asynchronous grant-free transmission occurs when MTCD timing offsets exceed the cyclic prefix, substantially increasing receiver detection and decoding complexity.Downlink synchronization can enable uplink timing adjustment and often achieve UE offsets within cyclic-prefix length without closed-loop timing advance.
  • C. Blind detection of MTCD activity and data: Joint blind MTCD activity detection and data decoding is crucial because the eNB lacks prior knowledge of transmission timing and must inspect each CTU.The main unresolved issue is determining how and on what basis to perform blind detection.
  • D. Collision management and reliability enhancement: Signature or pattern-vector differences can separate colliding MTCDs, but identical pattern vectors create hard collisions and mutual interference.Potential remedies include efficient signature design, larger resource pools, detection optimization, and MA resource management; power differences can also enable separation.
  • E. Link adaptation and power control: Link adaptation improves resource utilization, detection error rate, energy efficiency, and latency, but grant-free MTCDs may lack exact uplink channel-state information.Link adaptation matches modulation, coding, and other signal or protocol parameters to radio-link conditions.
  • E. Link adaptation and power control: Orthogonal MA blocks can use distinct resources and transmit settings, allowing active MTCDs to select a block before a signature and enabling parallel MUD.Broadcast configurations may limit signatures per block to reduce blind-detection complexity.

F. Hybrid automatic repeat request (HARQ) · IX. FUTURE DIRECTIONS

Grant-free HARQ must detect active users before ACK/NACK feedback and distinguish initial transmissions from retransmissions, while future work targets resource selection, synchronization, collision handling, adaptation, retransmission, and receiver design. These directions continue efforts to address practical challenges in grant-free uplink NOMA for mMTC and other use cases.

  • F. Hybrid automatic repeat request (HARQ): In grant-free uplink transmission, the eNB must detect user activity and data before providing ACK/NACK feedback because transmitting UEs are unknown in advance.Users otherwise wait a fixed period to determine whether retransmission is needed.
  • F. Hybrid automatic repeat request (HARQ): HARQ can merge failed transmissions with previous ones, but the eNB must identify initial transmissions and retransmissions within each HARQ process.Possible approaches include explicitly scheduling retransmissions through downlink control signaling or using distinct pilots.
  • F. Hybrid automatic repeat request (HARQ): Distinct pilots mapped to the initial transmission, first retransmission, and second retransmission can support signal decoding by combining successfully identified uncoded packets.Grant-free HARQ can therefore differ from LTE scheduled HARQ.
  • IX. FUTURE DIRECTIONS: Grant-free uplink NOMA requires further investigation, new designs, and technology integration because IoT use cases present diverse practical challenges.The paper frames the listed study items as a continuation of efforts to solve these challenges.
  • IX. FUTURE DIRECTIONS: Future studies include random or eNB-preconfigured resource selection, uplink synchronization under timing offsets within or beyond the cyclic prefix, and collision handling across resources and signatures.The considered multiple-access signatures include codes, sequences, and interleaver patterns.
  • IX. FUTURE DIRECTIONS: Further directions cover failed-transmission retransmission or repetition with potential combining, link adaptation through MCS or signature assignment, and the relationship between grant-free and grant-based transmissions.These topics also include associated user behavior.
  • IX. FUTURE DIRECTIONS: Research should also examine advanced receiver capabilities, complexity, and the requirement for power control to improve grant-free communication performance.The paper proposes exploring innovative options for mMTC and other use cases.

A. A unified framework for IoT use cases … 1) Relaying based grant-free UL NOMA:

The paper proposes a unified IoT network framework for diverse use cases and QoS requirements, while identifying grant-free UL NOMA directions including slicing, resource partitioning, transmission-mode switching, technology integration, and relaying. Relaying and indirect access are particularly motivated by congestion, power consumption, service life, and poor channel quality.

  • A. A unified framework for IoT use cases: A unified backbone/core system is needed to support eMBB, mMTC, and URLLC despite their extremely diverse QoS requirements.Potential directions toward this framework are highlighted in Fig. 21.
  • 1) Network slicing with grant-free UL NOMA:: Network slicing can partition physical network resources so diverse IoT users or tenants multiplex over a single infrastructure.The approach may provide additional flexibility for allocating resources across varied IoT use cases and QoS requirements.
  • 1) Network slicing with grant-free UL NOMA:: NOMA-based network slicing requires efficient resource management, and slice-based virtual resource scheduling has been proposed to enhance system QoS.Existing network-slicing studies with NOMA focus on scheduling-based rather than grant-free operation.
  • 2) Resource pool partitioning for grant-free UL NOMA:: Multiple NOMA subregions can tailor grant-free resources to different packet payloads, MCSs, transmission block sizes, or coverage enhancement levels.This partitioning aims to use resources efficiently and reduce eNB receiver complexity; OMA-based regions may also be reserved.
  • 3) Switching between grant-free and grant-based:: Dynamic grants can override grant-free transmissions, allowing flexible switching between grant-based and grant-free access for urgent events, resource reconfiguration, or services such as URLLC and eMBB.This gives the eNB scheduler flexibility in a unified network operation.
  • B. Integration of grant-free UL NOMA with other technologies: Grant-free UL NOMA has little existing integration with other cutting-edge technologies, although such integration could further improve performance.Scheduling-based NOMA has been integrated with other technologies in various studies.
  • B. Integration of grant-free UL NOMA with other technologies: Direct access is simple but can cause traffic congestion and excessive signaling overhead as the number of MTCDs increases, motivating gateway and coordinator access scenarios.Gateway access relays group data through a dedicated M2M gateway, whereas coordinator access uses a temporary MTCD gateway with its own traffic.
  • 1) Relaying based grant-free UL NOMA:: Relaying-based grant-free UL NOMA warrants study with half/full-duplex and decode-and-forward or amplify-and-forward mechanisms to address poor channel quality for distant users.Gateway or coordinator group transmission may reduce overall MTCD power consumption and extend service life.

2) MIMO-NOMA: … X. CONCLUSION

The merged sections examine grant-free uplink NOMA across MIMO, energy-harvesting, V2X, UAV, and broader mMTC access scenarios. They conclude by surveying candidate uplink NOMA techniques and their designs for autonomous grant-free connectivity.

  • 2) MIMO-NOMA:: MIMO-NOMA exploits MIMO’s spatial degree of freedom for data reliability through diversity, per-user capacity through spatial multiplexing, and grant-free uplink NOMA transmission.MIMO spatial diversity can facilitate grant-free uplink NOMA transmission.
  • 3) NOMA with energy harvesting:: SWIPT-NOMA can support massive connectivity while offering energy harvesting opportunities that help IoT devices achieve longer battery life.The discussed energy-harvesting solutions primarily concern scheduling-based NOMA schemes.
  • 3) NOMA with energy harvesting:: Figure 24 presents mMTC access scenarios relevant to grant-free NOMA connectivity.The passage identifies the figure as depicting mMTC access scenarios.
  • C. New IoT use cases: Grant-free uplink NOMA should be considered for emerging IoT use cases including vehicle-to-everything and unmanned aerial vehicle communications.These use cases were recently identified or agreed as relevant IoT applications.
  • 1) NOMA for V2X:: NOMA is crucial for cellular V2X because massive access makes orthogonal LTE resource allocation inefficient, causing congestion and access delay, while V2X requires low latency and high reliability.LTE has been considered a promising platform for V2X services.
  • 2) NOMA for UAV:: NOMA can increase connectivity density in UAV communications involving flying base stations serving ground users or multiple flying UAVs connected to a ground-based base station.Existing investigations include NOMA-based UAV communications but focus on scheduling-based approaches.
  • X. CONCLUSION: The survey addresses 5G autonomous, grant-free, contention-based uplink transmission for mMTC by categorizing candidate uplink NOMA techniques and detailing designs that meet grant-free requirements.The article’s stated contribution is a comprehensive survey of NOMA from a grant-free connectivity perspective.
Loading 1910.06529v1…