PaperlayerView exact revision on arXiv

Source-linked AI summary

Study of the Impact of PHY and MAC Parameters in 3GPP C-V2V Mode 4

Alessandro Bazzi, Giammarco Cecchini, Barbara M. Masini, Alberto Zanella

arXiv:1807.10699v2cs.NI

TL;DR

Mode 4 relies on distributed resource selection when network scheduling is unavailable, but its performance depends on finding unoccupied orthogonal resources. This paper evaluates the standardized algorithm’s PHY and MAC parameters through large-scale simulations and finds that parameter choices can have negligible effects, improve PRR, or trade reliability against update delay.

  • Problem

    Mode 4 must select unoccupied resources without network scheduling in some coverage conditions, and evidence on the effects of its parameters has been limited or focused on only a few settings.

    Source: §I · pp. 1–2

  • Method

    The study varies PHY- and MAC-layer parameters of the standardized Mode 4 algorithm through large-scale simulations in three realistic scenarios.

    Source: §I · p. 2, 2) PHY layer · p. 4

  • Results

    Optimized parameters offer a small PRR improvement limited to crowded Bologna, while higher pk can improve PRR at the cost of higher update delay in some conditions.

    Source: 4) Mode 4 with Optimized Parameters · p. 11, 3) Standard Mode 4 · p. 11

  • Takeaways & Limitations

    The parameter evaluation identifies settings and trade-offs that can inform future improvements to the Mode 4 algorithm.

    Source: §I · p. 2

Abstract

from arXiv · show

In the latest years, 3GPP has added short range cellular-vehicle-to-anything (C-V2X) to the features of LTE and 5G in order to make vehicles, roadside devices, and vulnerable users directly exchange information using the same chipset as for classical long range connections. C-V2X is based on the use of advanced physical layer techniques and orthogonal resources, and one of the main aspects affecting its performance is the way resources are allocated. Allocations can be either managed by the network or in a distributed way, directly by the nodes. The latter case, called Mode 4, is defined to manage those situations where the network cannot be involved in the scheduling process, for example due to a lack of coverage, but could also be adopted in order to reduce the processing burden of eNodeB. An algorithm, defined in the standards, makes nodes sense the medium and identify the best time-frequency combination to allocate their messages. Focusing on C-V2V Mode 4, in this work we analyze the parameters of the algorithm designed by 3GPP and their impact on the system performance. Through simulations in different large scale scenarios, we show that modifying some parameters have negligible effect, that the proper choice of others can indeed improve the quality of service, and that a group of parameters allows to trade-off reliability with update delay. The provided results can also be exploited to guide future work.

I. INTRODUCTION

Because Mode 4 remains necessary when network coverage is intermittent or unavailable, the paper examines how its PHY and MAC settings affect resource allocation. It evaluates those settings in realistic scenarios and identifies optimal choices to guide future improvements.

I. INTRODUCTION

  • Orthogonal time-frequency resources can avoid reciprocal interference, but only when nodes can identify unoccupied resources amid scarce spectrum and spatial interference.Source: §I · p. 1
  • Mode 3 allocates resources through the network, whereas distributed Mode 4 remains an option when coverage is intermittent or unavailable.Mode 3 is expected to outperform Mode 4 because its scheduler has more information,.Source: §I · pp. 1–2
  • The study evaluates main PHY and MAC parameters in three realistic scenarios, identifies optimal settings, and offers guidelines for future Mode 4 improvements.Earlier studies focused on specific aspects or a few parameters, while often assuming fixed settings.Source: §I · p. 2

A. LTE-V2V

  • LTE-V2V uses SC-FDMA and orthogonal resource blocks; each beacon’s transport block is paired with sidelink control information in the same subframe.Resource blocks are allocated in pairs, and a transport block can span multiple subchannels.Source: A. LTE-V2V · p. 2
  • Cooperative-awareness beacons carry periodic vehicle status and movement updates, with allocations made on a semi-persistent basis.Source: A. LTE-V2V · pp. 2–3

B. THE 3GPP MODE 4 ALGORITHM IN BRIEF

Mode 4 selects beacon resources through a PHY sensing stage followed by MAC allocation. The process filters candidate resources and then either reuses or reallocates a selected resource over time.

1) Beacon resources (BRs)

  • A beacon resource (BR) is a group of subchannels available for one beacon in the next beacon period, and the algorithm seeks a resource likely to be received correctly by neighbours.BRs correspond to 3GPP single-subframe resources.Source: 1) Beacon resources (BRs) · pp. 3–4

2) PHY layer

  • The PHY layer continuously reads decodable control information and measures interference to estimate BR occupancy, discarding measurements older than the sensing interval.Source: 2) PHY layer · p. 4
  • Within the T1-to-T2 window, the PHY retains monitored BRs estimated as unused, ranks them by average S-RSSI, and passes the lowest-ranked nR candidates to MAC.If too few candidates remain, Pth increases by 3 dB and candidate filtering repeats.Source: 2) PHY layer · p. 4

3) MAC layer

  • MAC randomly selects among the nR BRs and reserves the chosen resource for a random number of beacon periods between nmin and nmax.After the reservation interval, reallocation occurs with probability 1 − pk; otherwise the same BR is retained.Source: 3) MAC layer · p. 4

III. MODELLING AND SIMULATION SETTINGS

The simulations model periodic cooperative-awareness beacons under shared message and transmission assumptions. Performance is assessed by reception success and the time between successful updates from a neighbour.

III. MODELLING AND SIMULATION SETTINGS

  • The simulations use LTEV2Vsim, with common settings summarized alongside the study’s default values.Source: §III · p. 4

A. COOPERATIVE AWARENESS SETTINGS AND DEFINITION OF NEIGHBOUR

C. SCENARIOS

The evaluation spans medium-density urban, congested urban, and high-density highway scenarios, with distinct propagation conditions. PHY results show that the best MCS depends on scenario load and the trade-off between collisions and reception threshold.

C. SCENARIOS

  • The scenarios range from medium-density Cologne to congested Bologna and a high-density highway, with urban links often NLOS and highway links LOS.The upper 20% of vehicles in Bologna and Highway have similar neighbour counts, despite the highway range being twice as large.Source: C. SCENARIOS · p. 5

D. PROPAGATION AND SETTINGS AT THE PHY LAYER

E. PRELIMINARY CONSIDERATIONS: HIDDEN NODE

Hidden-node probability rises with source–destination distance, but its level and pattern vary by scenario because road topology shapes LOS/NLOS conditions. It remains limited over the distances studied, suggesting sensing can be effective against most interferers.

E. PRELIMINARY CONSIDERATIONS: HIDDEN NODE

  • Hidden-node probability is negligible near the destination, then increases with distance; its absolute values and pattern vary across scenarios with road topology and LOS/NLOS conditions.Cologne and Bologna show non-monotonic behavior, especially at short distances.Source: E. PRELIMINARY CONSIDERATIONS: HIDDEN… · p. 6
  • Below 10% up to 50 m in urban scenarios and almost 150 m on the highway, hidden-node probability does not exceed 30% within the distances considered.The considered ranges are 100 m in urban scenarios and 200 m on the highway.Source: E. PRELIMINARY CONSIDERATIONS: HIDDEN… · p. 6
  • The limited probability over reasonable distances implies that sensing procedures are potentially effective against most interferers.Source: E. PRELIMINARY CONSIDERATIONS: HIDDEN… · p. 6

IV. IMPACT OF MODE 4 PHY SETTINGS

The PHY-parameter simulations find that some settings have little effect, while others matter under particular conditions. Shorter sensing periods can improve PRR, but the standard value remains relevant when beacon periods may be longer.

IV. IMPACT OF MODE 4 PHY SETTINGS

  • Up to 10% higher PRR is observed when Tsense falls below 1 s, except in sparse Cologne; reducing it below 1 s is not generally viable because higher beacon periods are possible.The authors suggest adapting Tsense to neighbor settings or identifying and discarding outdated information.Source: 1) Sensing period · p. 6, Figure 5 · p. 7
  • About 5% improvement from a small Pth versus a high Pth appears only in crowded Bologna; the impact is irrelevant in Cologne and Highway.Pth matters when a very large number of resources exceed the interference threshold.Source: 2) Power threshold · pp. 6–7, Figure 5 · p. 7
  • Changing T1 and T2 produces negligible PRR differences; UD increases very slightly only for T1 = 4 and T2 = 20, favoring the largest feasible window subject to processing and delay constraints.Source: Figure 5 · p. 7, 4) Time window · p. 8
  • Rsel = 0.1 brings a small 2–3% improvement in the experiments, while the standard value 0.2 is an acceptable compromise against collision risk from a smaller resource pool.The collision risk matters especially when neighbors greatly outnumber resources.Source: 3) Portion of resources · pp. 7–8

V. IMPACT OF MODE 4 MAC SETTINGS

The MAC settings control how long SPS allocations persist and when reallocation is considered. Simulations show an unavoidable PRR–update-delay trade-off: increasing the settings improves PRR but worsens delay, while changing keep probability can control the trade-off without changing the evaluation window.

V. IMPACT OF MODE 4 MAC SETTINGS

  • nmin and nmax set the minimum and maximum beacon periods before reallocation is considered, while pk determines whether the allocation is kept for another such interval.Together, they determine the time before a different scheduling is performed.Source: §V · p. 8
  • Increasing nmin, nmax, or pk improves PRR but worsens update delay, and changing pk alone can control this trade-off without significant improvement from changing the window.The effects are attributed to more stable resource use and statistically longer durations of wrong allocations.Source: Figure 7 · p. 9, B. SIMULATIONS IN THE INVESTIGATED… · pp. 9–10
  • With nmin = 5, nmax = 15, and pk = 0.8, the probability of keeping the same resource for more than 10 seconds is approximately 0.1.The passage notes this can leave two neighboring vehicles hidden to each other for over 10 seconds in 1 case every 100.Source: A. FOCUS ON SPECIFIC METRICS · p. 9
  • For the standard with pk = 0, the probability of an allocation change during Tsense = 1 s is 0.9, making the sensed interference view inaccurate.Measurements before the change alter the correct view of interference.Source: A. FOCUS ON SPECIFIC METRICS · p. 9

VI. SUMMARY RESULTS AND DISCUSSION

Across realistic scenarios, LTE-V2V Mode 4 generally improves PRR over IEEE 802.11p and random allocation, but can increase update delay. Parameter optimization offers limited PRR gains, mainly in crowded settings, while the keep probability trades PRR against update delay.

2) Random allocation

  • Mode 4 identifies free resources and enables spatial reuse relative to random allocation, but its errors are more correlated and its UD can be double or more.The update-delay comparison is stated for the cases considered in the section.Source: 2) Random allocation · p. 10, Figure 9 · p. 11

4) Mode 4 with Optimized Parameters

  • Optimized Mode 4 yields small PRR gains limited to crowded Bologna, with slight UD improvement over standard pk=0.8 in congested scenarios and negligible worsening in Cologne.The optimized settings use pk=0.8, Pth=−128 dBm, Tsense=0.1 s, and Rsel=0.1.Source: Figure 8 · p. 10, §VI · p. 10, Figure 9 · p. 11

3) Standard Mode 4

1) LTE-V2V vs. IEEE 802.11p

  • LTE-V2V Mode 4 outperforms IEEE 802.11p in PRR in almost all situations, but IEEE 802.11p has lower UD in some cases, including Bologna at target 0.999.The comparison includes IEEE 802.11p with CSMA/CA, hidden terminals, capture effect, and QPSK with 1/2 coding.Source: Figure 8 · p. 10, 1) LTE-V2V vs. IEEE 802.11p · p. 10, §VI · p. 10

5) Comparison with related work

  • The results are consistent with related work: increasing pk improves highway PRR, while the Mode 4–random allocation gap shrinks as vehicle density falls.The cited comparisons include [11] for the pk trend and for the density-related gap.Source: Figure 8 · p. 10, 5) Comparison with related work · p. 11

VII. CONCLUSION

Simulations across three realistic scenarios show that parameter changes matter little at low-to-medium vehicle counts but can matter under congestion. PHY settings generally have minor effects, while MAC keep probability provides a PRR–UD trade-off.

VII. CONCLUSION

  • Parameter modifications are almost irrelevant with low-to-medium vehicle counts but may become significant during congestion.The conclusion draws on simulations in three realistic scenarios.Source: §VII · p. 11
  • Most PHY parameters have minor performance impact; set the sensing threshold and first next-allocation TTI to their allowed minima and the last TTI to its maximum.The recommendations concern the extremes allowed by the specifications.Source: §VII · pp. 11–12
  • Reducing the portion of beacon resources passed to MAC from the specification’s 20% can yield a minimum improvement, with the best value varying slightly by vehicle density.Source: §VII · p. 12
  • Reducing the 1 s sensing period can improve performance, but requires rethinking channel sensing because some nodes may transmit at 1 Hz.Source: §VII · p. 12
  • Changing keep probability can trade packet reception probability against update delay, and its variability appears sufficient to control performance while minimum and maximum beacon-period counts remain fixed.Source: §VII · p. 12

APPENDIX A: SINR CALCULATION

This appendix defines average SINR using received power, noise, and time- and frequency-overlap effects. It also describes how hidden-node probability is constructed and averaged over time.

APPENDIX A: SINR CALCULATION

  • A message is decodable only when transmitter and receiver use different subframes and average SINR exceeds γmin.Source: Appendix A · p. 12
  • Average SINR accounts for received power, noise, temporal overlap, and frequency leakage between signals.The IBE coefficient is 1 for frequency overlap and lower otherwise.Source: Appendix A · p. 12

APPENDIX B: CALCULATION OF THE HIDDEN NODE PROBABILITY

  • Hidden-node probability construction identifies each source’s reachable destinations, SINR-degrading interferers, and interferers the source cannot hear.Source: Appendix B · p. 12
  • Overall hidden-node probability averages instant-level Phn at periodic samples spaced by the 100 ms beacon interval.Source: Appendix B · p. 12

APPENDIX C: DEMONSTRATION OF EQUATION (3)

The appendix derives Equation (3), the probability that a resource allocation persists for a specified number of beacon periods. It builds the distribution by conditioning on the number of TBE composing a TBC.

APPENDIX C: DEMONSTRATION OF EQUATION (3)

  • Equation (3) gives the probability that the same allocation is kept for a specified number of beacon periods.Source: Appendix C · p. 13
  • For a TBC containing one TBE, the persistence distribution is conditioned on that single-TBE composition.Source: Appendix C · p. 13
  • For two independent TBE lengths, the persistence distribution uses convolution, which discrete Fourier transforms convert into element-wise products.The convolution construction generalizes to any number of TBE composing a TBC.Source: Appendix C · p. 13
  • The probability of a TBC containing k TBE is the probability of keeping the allocation k−1 times and then not keeping it, summed over k to obtain Equation (3).Source: Appendix C · p. 13

APPENDIX D: DEMONSTRATION OF EQUATION (4)

The appendix derives Equation (4), the probability that a TBC ends within a specified observation interval. It conditions on TBC length and then combines that conditional probability with the TBC-length distribution.

APPENDIX D: DEMONSTRATION OF EQUATION (4)

  • Equation (4) gives the probability that a TBC ends within an observation interval of a given number of beacon intervals.Source: Appendix D · p. 13
  • For a TBC of length l and interval length w, a change is certain when l≤w; otherwise its conditional probability depends on the TBC start position.Under the stated equal-probability assumption for its start, the probability when l>w is (l−w)/l.Source: Appendix D · p. 13
  • The overall probability of a change within the interval is obtained by combining the conditional probability with the TBC-length distribution from Equation (3).Source: Appendix D · p. 13
Loading 1807.10699v2…