Extended HARQ Processes

• Extended HARQ processes are advanced retransmission protocols that utilize parallelization to reduce delay and optimize spectral efficiency in 5G/6G systems.
• They integrate cross-packet, multi-packet, and variable-rate coding schemes to adapt to finite-blocklength and high-QoS wireless conditions.
• Polar code-based HARQ variants leverage low-complexity decoding and incremental redundancy for enhanced reliability and deterministic latency in URLLC.

Extended HARQ Processes

Extended Hybrid Automatic Repeat reQuest (HARQ) processes denote advanced retransmission protocols in wireless systems that generalize and expand upon classical, stop-and-wait HARQ through parallelization, cross-packet coding, flexible scheduling, and joint coding paradigms. These protocols are fundamental to the architecture and performance of modern 5G and emerging 6G systems, where stringent delay, reliability, and throughput requirements coexist. Extended HARQ encompasses multi-process parallel HARQ, cross-packet and multi-packet HARQ, and advanced polar code-based HARQ mechanisms, each providing specific structural and performance advantages in the finite-blocklength, high-QoS wireless regime.

1. Motivation and Architecture of Parallel HARQ

In 5G New Radio (NR) and related standards, HARQ is realized not as a singular stop-and-wait loop but through a bank of N parallel processes—typically with N = 8 or 16 per user. Each HARQ process maintains its own state machine, counter, and resource mapping. This parallelization ensures continuous transmission while awaiting decoding and feedback from previous rounds, amortizing the delays induced by non-negligible receiver processing and uplink/downlink ACK/NACK feedback (Moothedath et al., 12 Feb 2025). A core insight is that if only a single HARQ process is active, the downlink or uplink link idles for the entire HARQ round-trip time (RTT), wasting valuable spectral opportunities. Extended HARQ thus aligns protocol behavior with hardware-limited feedback cycles, maximizing link utilization under real-world latency constraints.

2. Queueing-Theoretic and Finite-Blocklength Delay Models

The performance of extended HARQ is rigorously modeled via multi-server queueing systems. The system is described as a discrete-time queue with arrival probability $f$ per slot, buffer size $Q_{max}$, and $N$ parallel HARQ servers. Each (re)transmission occupies one slot, with the overall state captured by the tuple $(q, m)$—queue length and the current retransmissions. Service completion depends on a finite-blocklength error process, where decoding and feedback latency are explicitly incorporated. For short packets (finite $n$), decoding error for ARQ is modeled as $p(S) = Q((\log_2(1+S) - \eta) \sqrt{L/V})$, and for HARQ-IR as $p_m(S) = Q((\log_2(1+S) - \eta/m) \sqrt{mL/V})$, with subsequent averaging over Rayleigh fading (Moothedath et al., 12 Feb 2025). The packet service time is $D_{serv}(m) = m (1 + D_{dec}) + (m-1) \delta$, and the packet delay (including waiting for a free server) admits a tractable Markov chain analysis, yielding tight bounds on Delay Violation Probability (DVP) under strict delay targets.

3. Cross-Packet, Variable-Rate, and Multi-Packet HARQ Extensions

Beyond classical IR-HARQ, several extended HARQ schemes dynamically incorporate new information bits and coordinate the transmission of multiple packets within a single HARQ process.

• Cross-Packet HARQ (XP-HARQ): Here, each HARQ round injects new information bits in addition to redundancy, with joint coding/decoding across rounds. The decoding success at round $k$ depends on the aggregate mutual information exceeding the accumulated rate, $I_1 + \ldots + I_k \geq R_1 + \ldots + R_k$ (Jabi et al., 2016). Adaptation of the rate vector $Q_{max}$0 according to feedback or past channel realization yields substantial throughput improvements and reduced SNR gap to ergodic capacity.
• Multi-Packet HARQ: Permits two (or more) packets to share a channel block, e.g., via time-sharing or superposition coding. Decisions on scheduling and resource splitting are optimized using Markov Decision Process (MDP) or Partial-State Information MDP when feedback is limited (Jabi et al., 2014). This closes much of the SNR gap to ergodic capacity that plagues conventional HARQ at high rates.
• Variable-Rate HARQ (VR-HARQ): Each retransmission may vary code rate, with rates $Q_{max}$1 selected to maximize effective capacity under latency and reliability constraints (Shi et al., 2018). XP-HARQ and VR-HARQ provide nearly identical effective capacity and both outperform fixed-rate schemes.

The table summarizes key process classes and performance metrics:

HARQ Class	Rate Adaptation / New Bits	Performance Gain
Multi-Process HARQ	No	Removes transmitter idling; strict DVP control (Moothedath et al., 12 Feb 2025)
XP-HARQ	Yes (joint coding)	1.5–2.5 dB SNR gain, $Q_{max}$2 VR-HARQ (Jabi et al., 2016Shi et al., 2018)
Multi-Packet HARQ	Yes (joint scheduling)	3–8 dB SNR gain at high rate (Jabi et al., 2014)
VR-HARQ	Yes (per-round rate)	Nearly matches XP-HARQ; both optimal for effective capacity (Shi et al., 2018)

4. Polar Codes and Incremental Redundancy Mechanisms

Extended HARQ processes for polar codes address key challenges in low-complexity, high-throughput scenarios. Prominent mechanisms include:

• Rate-Compatible Polar (RCP) HARQ: Constructs a family of nested codes via puncturing and repetitions built atop a polar base code, with a fixed information set (Chen et al., 2013). Soft information is naturally accumulated by SC decoding, requiring no change to the decoder algorithm.
• Matrix-Extension IR-HARQ for Polar Codes: Extends the length of the mother polar code in retransmissions, with each new block hosting new information bits and “parity-check frozen” (pc-frozen) bits mapped to previous ones (Ma et al., 2017Jalaleddine et al., 4 Dec 2025). Specialized modifications to SC decoders enable 70–72% reductions in decoding latency via special node skipping, while preserving FER (Jalaleddine et al., 4 Dec 2025).
• ARUM/Bit Relocation Schemes: Utilize inter-transmission polarization with XOR masking and dynamic relocation of weakly protected bits to highly reliable positions in later rounds. Each round may use distinct code lengths and rate-matching schemes, providing per-round design flexibility. The approach achieves within 0.2 dB of direct polar code design and strictly increases polarization depth per round (Chen et al., 2018).

Polar-based extended HARQ offers hardware-friendly $Q_{max}$3 complexity with deterministic latencies, which is significant for ultra-reliable, low-latency (URLLC) contexts.

5. Performance Metrics: Delay, Throughput, and Effective Capacity

Analysis of extended HARQ focuses on metrics such as delay violation probability (DVP), effective capacity $Q_{max}$4, and SNR gap to ergodic capacity:

• Delay Violation Probability (DVP): The probability that the sum of waiting and service delay exceeds a target, $Q_{max}$5, is derived via closed-form bounds (for ARQ) and efficient algorithms (for HARQ-IR) in Markovian multi-server models (Moothedath et al., 12 Feb 2025). Realistic decoding and feedback latencies (e.g., RTT = 4 slots) must be explicitly considered; simplified immediate-feedback assumptions yield several orders of magnitude DVP underestimation.
• Effective Capacity: Describes maximum sustainable arrival rate under QoS exponent $Q_{max}$6, capturing cross-layer performance. For renewal-type HARQ, $Q_{max}$7 is obtained by solving a root equation depending on the reward and renewal distributions, with explicit monotonicity and bounds in $Q_{max}$8 (Shi et al., 2018). VR-HARQ and XP-HARQ attain the same optimal $Q_{max}$9 and outperform fixed-rate schemes.
• Throughput and Fairness: In coordinated multi-user or multi-packet scenarios, SNR gains of several dB are realized over classical HARQ. Fairness and diversity increase with user collaboration (coordinated frequency/resource allocation), with theoretical diversity order scaling as $N$0 under $N$1 helpers (Makki et al., 2014).

6. Practical System Design and Implementation Considerations

Implementation of extended HARQ processes is guided by constraints on latency, feedback, complexity, and feedback granularity:

• The number of parallel HARQ processes should at least match the HARQ RTT (in slots) to eliminate transmitter idle slots (Moothedath et al., 12 Feb 2025).
• Resource allocation per packet (measured in “symbols per bit”) is the primary figure of merit for DVP compliance.
• Arrival rates must be matched to the link’s sustainable rate; surplus load increases queueing delay and DVP precipitously.
• Special node-enabled SC decoding for polar IR-HARQ achieves low area overhead and deterministic, minimum-latency operation (Jalaleddine et al., 4 Dec 2025).
• Cross-packet and multi-packet HARQ necessitate joint coding/decoding machinery (e.g., parallel turbo encoding, or multi-stage polar decoding with bit relocation maps) and, in the most aggressive feedback scenarios, look-up tables for MDP policies indexed by mutual information accumulations (Jabi et al., 2016Jabi et al., 2014).

A plausible implication is that deployment of extended HARQ in URLLC-type services or ultra-high-throughput 6G applications will require per-link and per-service parameterization, leveraging design flexibility (rate-matching, blocklength, code structure) to balance delay, reliability, and decoding complexity.

7. Comparative Analysis and Conclusions

Across ARQ/HARQ variants, parallelization and joint coding extensions uniformly reduce latency, improve throughput, and approach information-theoretic limits. Numerical studies demonstrate:

• HARQ with parallel processes brings DVP and throughput metrics into alignment with event-driven simulation, distancing itself from naive queueing models through precise finite-blocklength and latency analysis (Moothedath et al., 12 Feb 2025).
• Cross-packet/multi-packet extensions close much of the SNR gap to ergodic capacity, especially under moderate-to-high spectral efficiencies, with observed 4–8 dB gains (Jabi et al., 2016Jabi et al., 2014).
• Polar-based extended HARQ achieves sub-dB gaps to direct designs and outperforms turbo-based references, with deterministic complexity and bounded delay (1307.28002512.04418Chen et al., 2018).

Guidelines anchored in the referenced analyses are now central to system-level configuration in both standards-compliant and research platforms for latency-sensitive wireless services.

---

References:

(Moothedath et al., 12 Feb 2025) Delay Analysis of 5G HARQ in the Presence of Decoding and Feedback Latencies (Chen et al., 2013) A Hybrid ARQ Scheme Based on Polar Codes (Jabi et al., 2016) Adaptive Cross-Packet HARQ (Shi et al., 2018) Effective Capacity for Renewal Service Processes with Applications to HARQ Systems (Makki et al., 2014) Coordinated Hybrid Automatic Repeat Request; Extended Version (Jalaleddine et al., 4 Dec 2025) Enabling Fast Polar SC Decoding with IR-HARQ (Ma et al., 2017) An Incremental Redundancy HARQ Scheme for Polar Code (Jabi et al., 2014) Multi-packet Hybrid ARQ: Closing gap to the ergodic capacity (Chen et al., 2018) ARUM: Polar Coded HARQ Scheme based on Incremental Channel Polarization (Trillingsgaard et al., 2018) Generalized HARQ Protocols with Delayed Channel State Information and Average Latency Constraints