PaperlayerView exact revision on arXiv

Source-linked AI summary

Conditional-Coverage Contributor Selection for Regional Digital Twins Under Mobility

Tao Yu

arXiv:2609.36459v1cs.ITmath.IT

TL;DR

Mobile regional digital twins need contributor admission that responds to changing regional coverage, but per-sender mitigation cannot use the host’s region-wide view. The paper proposes host-coordinated, congestion-priced admission of timely non-redundant coverage. In a corridor simulation, the deficit fell from 5.6% under CPM-RMR to 1.4% at comparable load, and conditional scoring reduced deficit 26% versus random admission at matched tight-budget load.

  • Problem

    Per-sender redundancy mitigation cannot resolve regional contributor selection because only the fusion host knows what receivers lack, even though all agents receive the fused state while contributors bear reporting costs.

    Source: §I · p. 1

  • Method

    A host uses conditional coverage value and a regional congestion price to admit candidates whose timely coverage gain exceeds airtime cost within a cycle deadline.

    Source: §I · p. 1, C. Conditional Incremental Coverage… · p. 2, D. Network Load and Regional… · p. 3

  • Results

    5.6% to 1.4% timely actionable coverage deficit versus CPM-RMR at comparable load; conditional scoring reduced deficit 26% versus random admission at matched tight-budget load.

    Source: D. Protocol Performance · pp. 4–5

  • Takeaways & Limitations

    In the evaluated 80%-penetration corridor, the hosted twin cut the timely actionable coverage deficit fourfold versus CPM-RMR at comparable load.

    Source: §V · p. 5

    Evaluation is reported for an 80%-penetration corridor; broader density and penetration regimes, adaptive hysteresis thresholds, non-uniform-traffic host placement, and lower-rate digests remain future work.

    Source: §V · p. 5

Abstract

from arXiv · show

Regional digital twins (DTs) under mobility must coordinate contributor admission over congested broadcast networks without fixed infrastructure. Per-sender redundancy mitigation cannot make this set-level decision because only the host has a region-wide coverage map. Treating regional state as a public good, this letter develops a host-coordinated protocol admitting contributors whose timely, non-redundant coverage gain exceeds a congestion price within a cycle deadline. A 13-byte beacon digest enables pre-transmission scoring. Standards-based simulation shows the protocol reduces the timely actionable coverage deficit from 5.6\% to 1.4\% at load comparable to ETSI redundancy mitigation and by 26\% versus random admission at matched tight-budget load.

I. INTRODUCTION

Mobile regional digital twins must select contributors as agents, observations, and fusion hosts change within seconds. The paper frames this as a host-coordinated set-level decision based on regional coverage and airtime cost.

I. INTRODUCTION

  • Existing mobile DT designs assume fixed scopes, information sources, and infrastructure hosts, while mobility changes who observes, needs, and can fuse regional state.Periodic beacons consume shared-channel capacity, and nearby observations overlap heavily.Source: §I · p. 1
  • The decision differs from per-sender mitigation, ego-vehicle helper selection, and report-content control by selecting a sender set for all regional receivers.The paper also distinguishes its training-free airtime pricing on a standardized channel from learned ego-centric collaborative perception.Source: §I · p. 1
  • Per-sender redundancy mitigation cannot make the regional decision because only the fusion point knows what each receiver lacks.The fused state is a regional public good: all agents in the footprint receive it, while contributors bear the upload cost.Source: §I · p. 1
  • The protocol admits a candidate only when its timely, non-redundant coverage gain exceeds regional airtime price within a hard service-cycle deadline.A compact coverage digest in periodic beacons enables scoring before detailed transmission; the protocol is instantiated with CAM beacons and CPM reports.Source: §I · p. 1

II. SYSTEM MODEL

Each regional twin has a host that fuses contributor reports and broadcasts fused state to a changing service scope. The benefit reaches all participating agents in the footprint, while upload and fusion costs fall on contributors.

A. Scopes and Roles

  • The contributor scope Dk(t) contains the host and admitted agents uploading detailed reports, while the service scope Fk(t) contains participants in the host’s broadcast footprint.Agents in the service scope receive fused state whether or not they contribute.Source: A. Scopes and Roles · pp. 1–2
  • Fig. 1 depicts the host broadcasting fused states to contributors and passive listeners, with a unique-view candidate admitted while redundant candidates are not.The sketch marks already-held cells with MD ≈1 and an occluded cell with MD ≈0 until the unique-view contributor is admitted.Source: Figure 1 · p. 2
  • The host fuses reports and broadcasts once per fixed-period cycle Ts, with airtime increasing as the contributor set grows.The benefit is summed over Fk, whereas uplink and broadcast costs are borne by Dk.Source: A. Scopes and Roles · p. 2
  • Task relevance determines an agent’s weight in the benefit sum, but contributor eligibility depends on whether its observations add to what the twin already holds.A high-risk vehicle may be ineligible when its view is already covered, while a low-risk vehicle may uniquely see an occluded lane.Source: A. Scopes and Roles · p. 2

B. Information Available for Admission

Admission is scored from information already available to the host, including a beacon-borne coverage digest. The score estimates a candidate’s conditional reduction in timely task loss across the service scope.

B. Information Available for Admission

  • The host’s information includes local sensing, received beacons, overheard DT broadcasts, and contributor reports delivered by transport.The coverage grid uses fixed square cells of 20 m aligned to a global origin so hosts index cells identically.Source: B. Information Available for Admission · p. 2
  • A 12-byte digest header plus a variable-length run-length encoded visible-cell bitmap averages about 13 B under heavy occlusion.The digest reveals where a candidate can observe without revealing object states, allowing value computation before admission and detailed transmission.Source: B. Information Available for Admission · p. 2

C. Conditional Incremental Coverage Value

  • Candidate value is the reduction in task loss across the service scope, conditioned on information already available to the twin.Task loss is risk-weighted conflict-course objects whose state an agent has not received within its reaction deadline; benefits include passive listeners.Source: C. Conditional Incremental Coverage… · p. 2
  • The factorized surrogate combines novelty, observability, task risk, and timely delivery, using the host’s soft coverage-confidence map and candidate visibility digest.Timely delivery depends on whether receiver information age remains within its reaction deadline given post-admission cycle completion time; the score decreases as Tcyc grows.Source: C. Conditional Incremental Coverage… · p. 2
  • The host ages its previous coverage map and fuses current-cycle coverage evidence, accounting for report decoding and overheard neighboring-host broadcasts.A decoded report removes the delivery discount applied to its digest’s expected coverage, avoiding double counting.Source: C. Conditional Incremental Coverage… · pp. 2–3

D. Network Load and Regional Congestion Control

Regional load is controlled through a per-host allowance and a regional congestion price, while a hard cycle deadline preserves real-time feasibility. The controls bind in different regimes.

D. Network Load and Regional Congestion Control

  • Orthogonal uplink accounting sums contributor report airtime and fused-state broadcast airtime, while contention and collisions are handled by the channel model.This accounting abstracts distributed radio scheduling.Source: D. Network Load and Regional… · p. 3
  • The per-host allowance b is a hard cap on one twin’s capacity use, while regional target ΓR prices total tile occupancy, including traffic the host cannot control.Mandatory beacons are excluded because admission cannot change them.Source: D. Network Load and Regional… · p. 3
  • A projected dual-ascent update adjusts λR from measured regional occupancy, pricing each admission against the congestion target.The update uses step size µ=0.5 each beacon interval.Source: D. Network Load and Regional… · p. 3
  • The hard constraint Tcyc(D, h) ≤ Tmax guarantees real-time feasibility.Source: D. Network Load and Regional… · p. 3
  • With small b, the allowance binds first and λR stays near zero; with generous b, the price rises until load settles just below ΓR.Source: D. Network Load and Regional… · p. 3

III. HOST-COORDINATED CONTRIBUTOR ADMISSION

The incumbent host scores candidate contributors by net marginal value, balancing conditional coverage gain against priced airtime. Admission and removal are constrained by cycle feasibility, the per-host allowance, and hysteresis.

III. HOST-COORDINATED CONTRIBUTOR ADMISSION

  • Each cycle, the host updates coverage, service scope, risk, link statistics, and congestion price, then scores candidates by net marginal value.Source: §III · p. 3
  • A candidate is admitted only if its score reaches ηadd and the resulting cycle and per-host allowance remain feasible.Source: §III · p. 3
  • Non-host contributors are removed when their retention score is at or below ηrem, with ηrem < ηadd creating hysteresis against oscillation.Source: §III · p. 3
  • Admissions and removals share a per-cycle quota nmax=1 in evaluation, with scores recomputed after each committed change and single marginal steps separating changes.The thresholds mainly filter near-zero-value candidates; score ordering selects contributors, and the dead band determines when the best candidate is near the threshold.Source: §III · p. 3
  • With MD as a group-level surrogate, host evaluation costs O((|J|+|F|)|G|) plus O(|J| log |J|) candidate sorting per cycle.Source: §III · p. 3

A. Vehicular Instantiation

The vehicular instantiation uses standard CAM, CPM, and PC5 dissemination with a coverage digest as its sole non-standard element. Host replacement preserves the fixed tile footprint, while migration and cross-host consistency remain out of scope.

A. Vehicular Instantiation

  • The framework uses 10 Hz ETSI CAMs with a coverage digest as the only non-standard container, ordinary CPMs for admitted contributors, and LTE-V2X PC5 mode 4.The digest is placed in CAMs because candidates do not send CPMs.Source: A. Vehicular Instantiation · p. 3
  • Contributor CPMs are received and fused directly by vehicles in range; host broadcasts extend reach, carry changed states each cycle, and refresh fully every 1 s.Listeners deduplicate multiple hosts by timestamp, confidence, and deterministic tie-break.Source: A. Vehicular Instantiation · p. 3
  • When a host leaves, the nearest-center rule replaces it without handover and keeps the service scope fixed, but may place the host at a platoon edge under strongly non-uniform traffic.Migration and cross-host consistency are out of scope.Source: A. Vehicular Instantiation · p. 4

B. Setup

The evaluation uses a congested, signalized Tokyo corridor and standards-based channel modeling, with paired-seed comparisons. Its primary metric measures whether decision-relevant objects are unknown or stale beyond a vehicle’s reaction deadline.

B. Setup

  • The simulation uses a 3.7 km Tamagawa-dōri corridor with 22 signalized intersections at density 1.5 kc and 80% connected penetration.The density is held by closed-population control, with kc=28.72 veh/km/lane measured from the corridor fundamental diagram.Source: B. Setup · p. 4
  • Maximum-rate CAMs alone occupy 64% of capacity on average, while the digest adds about 0.07 capacity units to the proposed scheme’s cooperative load.The channel model abstracts PC5 mode 4 using urban link states, published block-error curves, load-dependent collision loss, and hard capacity enforcement.Source: B. Setup · p. 4
  • The primary metric is the pooled ratio of decision-relevant objects unknown or older than the reaction deadline across vehicles and cycles.An object is decision-relevant when its ground-truth state lies on a map-derived conflict course with the vehicle’s predicted path.Source: B. Setup · p. 4
  • Fig. 2 plots timely actionable coverage deficit against normalized cooperative offered load at 1.5 kc density and 80% penetration; its shaded band is the oracle information ceiling.Error bars are arm-wise 95% confidence intervals, not pairwise comparisons.Source: Figure 2 · p. 4
  • Results use 95% seed-level bootstrap confidence intervals over 17 paired seeds, after 40 s warm-up and during 40 s measurement.Source: B. Setup · p. 4

C. Baselines

The evaluation compares admission strategies over a shared sensing, communication, fusion, and passive-listening stack, varying who reports and what is filtered.

C. Baselines

  • CPM-ETSI follows collective-perception message rules without optional redundancy mitigation, while CPM-RMR adds TR 103 562 mitigation and a 32-object cap.CPM-RMR uses object self-announcement, frequency- and dynamics-based omission, and an always-active CBR gate of 0.4.Source: C. Baselines · p. 4
  • Random top-k keeps the hard cycle and per-host allowance constraints but selects randomly, without responding to λR.The criterion ablations truncate the score to unconditional coverage, conditioning alone, or conditioning with risk weighting.Source: C. Baselines · p. 4
  • The oracle-awareness variant assumes lossless, delay-free delivery of all equipped observations, while CAM-only is the no-cooperation reference.Source: C. Baselines · p. 4

D. Protocol Performance

The proposed selection lowers timely coverage deficit as the allowance rises and compares favorably with CPM-RMR at similar load. Paired comparisons and score ablations indicate that conditional scoring improves how the allowance is spent.

D. Protocol Performance

  • At comparable load, the proposed scheme achieves a fourfold lower deficit than CPM-RMR while using 27% less reporting airtime.It spends 0.22 capacity units on reports net of the digest versus 0.30 for CPM-RMR; the deficit falls from 5.6% to 1.4%.Source: D. Protocol Performance · pp. 4–5
  • 26% lower deficit at b=0.2 versus random admission isolates the selection criterion at identical offered load.Paired deficit reductions are 33% at b=0.4 and 10% at b=0.8; the latter mixes criterion and price effects because random admission overshoots ΓR.Source: D. Protocol Performance · p. 5, Figure 3 · p. 5
  • 0.65 pp is the paired per-seed deficit reduction from conditioning on the twin’s existing coverage, the largest single ablation step at b=0.4.Risk weighting adds 0.39 pp; the timeliness factor adds 0.21 pp with an interval including zero, consistent with the cycle coupling being nonbinding at this allowance.Source: D. Protocol Performance · p. 5, Figure 4 · p. 5
  • 0.29 to 0.26 is the offered-load reduction across score refinements, with each refinement achieving more coverage using less airtime.The conditioning term carries the largest share of the criterion gain.Source: D. Protocol Performance · p. 5

V. CONCLUSION

The conclusion frames mobile regional-DT contributor admission as a host-coordinated, set-level conditional-coverage decision. In the evaluated corridor, the hosted twin improves deficit over CPM-RMR at comparable load and over random admission at a matched tight budget.

V. CONCLUSION

  • In the evaluated 80%-penetration corridor, the hosted twin cuts timely actionable coverage deficit fourfold versus CPM-RMR at comparable load.Source: §V · p. 5
  • At the matched tight-budget point, the criterion alone improves deficit by 26% over random admission.Source: §V · p. 5
  • The paper identifies broader density and penetration regimes, adaptive hysteresis thresholds, host placement under non-uniform traffic, and lower-rate digests as future work.Source: §V · p. 5
Loading 2609.36459v1…